{ "schema_version": 1, "source": "https://dion-jy.github.io/spotlight-todai", "license": "Curation free to use; papers follow their original sources.", "count": 1167, "fields": { "uid": "stable global id: ---, lowercase", "conference": "ICLR | ICML | NeurIPS", "year": "publication year (int)", "track": "Oral | Spotlight", "title": "paper title", "authors": "full author list (array of strings); may be [first author] only", "affiliation": "first-author affiliation, '' if unknown", "summary": "1-2 line TLDR/abstract, '' if not yet enriched", "links": "{openreview, arxiv, detail} — '' when absent" }, "papers": [ { "uid": "iclr-2026-oral-1", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Common Corpus: The Largest Collection of Ethical Data for LLM Pre-Training", "authors": [ "Pierre-Carl Langlais", "Pavel Chizhov", "Catherine Arnett", "Carlos Rosas Hinostroza", "Mattia Nee", "Eliot Krzysztof Jones", "Irène Girard", "David Mach", "Anastasia Stasenko", "Ivan P. Yamshchikov" ], "affiliation": "Pleias", "summary": "We assemble and release the largest truly open multilingual dataset for LLM pre-training consisting of 2 trillion tokens", "links": { "openreview": "https://openreview.net/forum?id=0wSlFpMsGb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-2", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Q-RAG: Long Context Multi‑Step Retrieval via Value‑Based Embedder Training", "authors": [ "Artyom Sorokin", "Nazar Buzun", "Aleksandr Anokhin", "Egor KONSTANTINOVICH VEDERNIKOV", "Petr Anokhin", "Mikhail Burtsev", "Evgeny Burnaev" ], "affiliation": "Applied AI Institute", "summary": "Retrieval-Augmented Generation (RAG) methods enhance LLM performance by efficiently filtering relevant context for LLMs, reducing hallucinations and inference cost. However, most existing RAG methods focus on single-step retrieval, which is often insufficient for answering complex questions that require multi-step search.", "links": { "openreview": "https://openreview.net/forum?id=MS9nWFY7LG", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-3", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "From movement to cognitive maps: recurrent neural networks reveal how locomotor development shapes hippocampal spatial coding", "authors": [ "Marco P Abrate", "Laurenz Muessig", "Joshua P Bassett", "Hui Min Tan", "Francesca Cacucci", "Thomas Joseph Wills", "Caswell Barry" ], "affiliation": "University College London, University of London", "summary": "The hippocampus contains neurons whose firing correlates with an animal's location and orientation in space. Collectively, these neurons are held to support a cognitive map of the environment, enabling the recall of and navigation to specific locations.", "links": { "openreview": "https://openreview.net/forum?id=8bM7MkxJee", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-4", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "FIRE: Frobenius-Isometry Reinitialization for Balancing the Stability–Plasticity Tradeoff", "authors": [ "Isaac Han", "Sangyeon Park", "Seungwon Oh", "Donghu Kim", "Hojoon Lee", "KyungJoong Kim" ], "affiliation": "Gwangju Institute of Science and Technology", "summary": "We present FIRE, a principled reinitialization approach that balances stability and plasticity through constrained optimization.", "links": { "openreview": "https://openreview.net/forum?id=CfZLxT3zIZ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-5", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Exchangeability of GNN Representations with Applications to Graph Retrieval", "authors": [ "Kartik Nair", "Indradyumna Roy", "Soumen Chakrabarti", "Anirban Dasgupta", "Abir De" ], "affiliation": "Carnegie Mellon University", "summary": "It shows that graph representations are exchangeable random variables which can help in LSH in graphs", "links": { "openreview": "https://openreview.net/forum?id=HQcCd0laFq", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-6", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Enhancing Generative Auto-bidding with Offline Reward Evaluation and Policy Search", "authors": [ "Zhiyu Mou", "Yiqin Lv", "Miao Xu", "Cheems Wang", "Yixiu Mao", "Jinghao Chen", "Qichen Ye", "Chao Li", "Rongquan Bai", "Chuan Yu", "Jian Xu", "Bo Zheng" ], "affiliation": "Alibaba Group", "summary": "Auto-bidding is a critical tool for advertisers to improve advertising performance. Recent progress has demonstrated that AI-Generated Bidding (AIGB), which learns a conditional generative planner from offline data, achieves superior performance compared to typical offline reinforcement learning (RL)-based auto-bidding methods.", "links": { "openreview": "https://openreview.net/forum?id=kMuQBgPIdg", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-7", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Why DPO is a Misspecified Estimator and How to Fix It", "authors": [ "Aditya Gopalan", "Sayak Ray Chowdhury", "Debangshu Banerjee" ], "affiliation": "Indian Institute of Science", "summary": "DPO is not sound by design and can fail due to misspecification, we fix it with careful analysis.", "links": { "openreview": "https://openreview.net/forum?id=btEiAfnLsX", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-8", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "WebDevJudge: Evaluating (M)LLMs as Critiques for Web Development Quality", "authors": [ "Chunyang Li", "Yilun Zheng", "Xinting Huang", "Tianqing Fang", "Jiahao Xu", "Lihui Chen", "Yangqiu Song", "Han Hu" ], "affiliation": "Department of Computer Science and Engineering, Hong Kong University of Science and Technology", "summary": "A meta-evaluation benchmark for assessing LLM-as-a-judge in the context of web development.", "links": { "openreview": "https://openreview.net/forum?id=CCSPm6V5EF", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-9", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "SafeDPO: A Simple Approach to Direct Preference Optimization with Enhanced Safety", "authors": [ "Geon-Hyeong Kim", "Yu Jin Kim", "Byoungjip Kim", "Honglak Lee", "Kyunghoon Bae", "Youngsoo Jang", "Moontae Lee" ], "affiliation": "LG AI Research", "summary": "This work introduces a simple yet principled approach for directly optimizing the safety alignment objective during policy learning", "links": { "openreview": "https://openreview.net/forum?id=PJdw4VBsXD", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-10", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MedAgentGym: A Scalable Agentic Training Environment for Code-Centric Reasoning in Biomedical Data Science", "authors": [ "Ran Xu", "Yuchen Zhuang", "Yishan Zhong", "Yue Yu", "Zifeng Wang", "Xiangru Tang", "Hang Wu", "May Dongmei Wang", "Peifeng Ruan", "Donghan Yang", "Tao Wang", "Guanghua Xiao", "Xin Liu", "Carl Yang", "Yang Xie", "Wenqi Shi" ], "affiliation": "Google DeepMind", "summary": "MedAgentGym is a scalable and interactive training environment designed to enhance coding-based biomedical reasoning capabilities in LLM agents.", "links": { "openreview": "https://openreview.net/forum?id=jHDZEUgS4r", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-11", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Optimistic Task Inference for Behavior Foundation Models", "authors": [ "Thomas Rupf", "Marco Bagatella", "Marin Vlastelica", "Andreas Krause" ], "affiliation": "ETHZ - ETH Zurich", "summary": "We propose an algorithm for fast online task inference in behavior foundation models.", "links": { "openreview": "https://openreview.net/forum?id=m5byThUSNE", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-12", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "CounselBench: A Large-Scale Expert Evaluation and Adversarial Benchmarking of Large Language Models in Mental Health Question Answering", "authors": [ "Yahan Li", "Jifan Yao", "John Bosco S. Bunyi", "Adam C Frank", "Angel Hsing-Chi Hwang", "Ruishan Liu" ], "affiliation": "", "summary": "Medical question answering (QA) benchmarks often focus on multiple-choice or fact-based tasks, leaving open-ended answers to real patient questions underexplored. This gap is particularly critical in mental health, where patient questions often mix symptoms, treatment concerns, and emotional needs, requiring answers that balance clinical caution...", "links": { "openreview": "https://openreview.net/forum?id=8MBYRZHVWT", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-13", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "AstaBench: Rigorous Benchmarking of AI Agents with a Scientific Research Suite", "authors": [ "Jonathan Bragg", "Mike D'Arcy", "Nishant Balepur", "Dan Bareket", "Bhavana Dalvi Mishra", "Sergey Feldman", "Dany Haddad", "Jena D. Hwang", "Peter Jansen", "Varsha Kishore", "Bodhisattwa Prasad Majumder", "Aakanksha Naik", "Sigal Rahamimov", "Kyle Richardson", "Amanpreet Singh", "Harshit Surana", "Aryeh Tiktinsky", "Rosni Vasu", "Guy Wiener", "Chloe Anastasiades", "Stefanus Candra", "Jason Dunkelberger", "Daniel Emery", "Rob Evans", "Malachi Hamada", "Regan Huff", "Rodney Kinney", "Matt Latzke", "Jaron Lochner", "Ruben Lozano-Aguilera", "Ngoc-Uyen Nguyen", "Smita Rao", "Amber Tanaka", "Brooke Vlahos", "Peter Clark", "Doug Downey", "Yoav Goldberg", "Ashish Sabharwal", "Daniel S Weld" ], "affiliation": "Allen Institute for Artificial Intelligence", "summary": "We present principles and tooling for rigorous AI agent benchmarking, instantiated in AstaBench—the first holistic measure of agentic ability for scientific research—plus experiments showing AI remains far from solving research assistance.", "links": { "openreview": "https://openreview.net/forum?id=M7TNf5J26u", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-14", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Neon: Negative Extrapolation From Self-Training Improves Image Generation", "authors": [ "Sina Alemohammad", "Zhangyang Wang", "Richard Baraniuk" ], "affiliation": "University of Texas at Austin", "summary": "Instead of simply fine-tuning a generative model on its own synthetic outputs, briefly fine-tune it to find the direction of model collapse, then apply the reverse of that update to the original model for a major performance boost.", "links": { "openreview": "https://openreview.net/forum?id=kpLRYtPGt3", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-15", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Compositional Diffusion with Guided search for Long-Horizon Planning", "authors": [ "Utkarsh Aashu Mishra", "David He", "Yongxin Chen", "Danfei Xu" ], "affiliation": "Georgia Institute of Technology", "summary": "We integrate search into compositional diffusion to scale short-horizon models into long-horizon plans, supporting motion planning, panoramic image synthesis, and long-video generation.", "links": { "openreview": "https://openreview.net/forum?id=b8avf4F2hn", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-16", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Visual symbolic mechanisms: Emergent symbol processing in Vision Language Models", "authors": [ "Rim Assouel", "Declan Iain Campbell", "Yoshua Bengio", "Taylor Whittington Webb" ], "affiliation": "", "summary": "We describe a set of symbolic-like mechanisms that VLMs use to bind to visual entities in context", "links": { "openreview": "https://openreview.net/forum?id=3RQ863cRbx", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-17", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Addressing divergent representations from causal interventions on neural networks", "authors": [ "Satchel Grant", "Simon Jerome Han", "Alexa R. Tartaglini", "Christopher Potts" ], "affiliation": "MATS", "summary": "We show empirical representational divergence between native and causally intervened latent states, we show that this can be pernicious and propose a solution.", "links": { "openreview": "https://openreview.net/forum?id=cZrTMqYVL6", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-18", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Cross-Domain Lossy Compression via Rate- and Classification-Constrained Optimal Transport", "authors": [ "Nam Nguyen", "Thinh Nguyen", "Bella Bose" ], "affiliation": "Oregon State University", "summary": "We study cross-domain lossy compression via constrained optimal transport with rate and classification constraints, derive closed-form tradeoffs, extend to perception divergences, and validate with deep restoration and inpainting experiments.", "links": { "openreview": "https://openreview.net/forum?id=mUIGdUTtk2", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-19", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Latent Fourier Transform", "authors": [ "Mason Long Wang", "Cheng-Zhi Anna Huang" ], "affiliation": "Massachusetts Institute of Technology", "summary": "We introduce novel frequency-domain controls for generative music models by applying the Fourier transform to the latent space of a diffusion autoencoder.", "links": { "openreview": "https://openreview.net/forum?id=ogMxCjdCCq", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-20", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "GLASS Flows: Efficient Inference for Reward Alignment of Flow and Diffusion Models", "authors": [ "Peter Holderrieth", "Uriel Singer", "Tommi Jaakkola", "Ricky T. Q. Chen", "Yaron Lipman", "Brian Karrer" ], "affiliation": "MIT", "summary": "We improve inference-time reward alignment of flow matching and diffusion models by proposing a novel sampling paradigm that enables more efficient exploration.", "links": { "openreview": "https://openreview.net/forum?id=vH7OAPZ2dR", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-21", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Latent Speech-Text Transformer", "authors": [ "Yen-Ju Lu", "Yashesh Gaur", "Wei Zhou", "Benjamin Muller", "Jesus Villalba", "Najim Dehak", "Luke Zettlemoyer", "Gargi Ghosh", "Mike Lewis", "Srini Iyer", "Duc Le" ], "affiliation": "Johns Hopkins University", "summary": "We introduce Latent Speech-Text Transformer, which patches long speech token sequences into latent units, improving text–speech transfer while cutting pre-training and inference compute, and significantly outperforming existing speech-text LLMs.", "links": { "openreview": "https://openreview.net/forum?id=krGpQzo8Mz", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-22", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "LoongRL: Reinforcement Learning for Advanced Reasoning over Long Contexts", "authors": [ "Siyuan Wang", "Gaokai Zhang", "Li Lyna Zhang", "Ning Shang", "Fan Yang", "Dongyao Chen", "Mao Yang" ], "affiliation": "Shanghai Jiaotong University", "summary": "Reasoning over long contexts is essential for large language models. While reinforcement learning (RL) enhances short-context reasoning by inducing \"Aha\" moments in chain-of-thought, the advanced thinking patterns required for long-context reasoning remain largely unexplored, and high-difficulty RL data are scarce.", "links": { "openreview": "https://openreview.net/forum?id=o29E01Q6bv", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-23", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Speculative Actions: A Lossless Framework for Faster AI Agents", "authors": [ "Naimeng Ye", "Arnav Ahuja", "Georgios Liargkovas", "Yunan Lu", "Kostis Kaffes", "Tianyi Peng" ], "affiliation": "Columbia University", "summary": "We introduce speculative actions—a lossless framework that predicts likely actions using faster models, enabling multiple API calls to be executed in parallel and thus yields substantial acceleration.", "links": { "openreview": "https://openreview.net/forum?id=P0GOk5wslg", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-24", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Gaussian certified unlearning in high dimensions: A hypothesis testing approach", "authors": [ "Aaradhya Pandey", "Arnab Auddy", "Haolin Zou", "Arian Maleki", "Sanjeev Kulkarni" ], "affiliation": "Princeton University", "summary": "We introduce the canonical dimension free notion of certifiability suitable to high dimensions and show its utility via a Newton based unlearning algorithm", "links": { "openreview": "https://openreview.net/forum?id=0FJYicpOj0", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-25", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "RedTeamCUA: Realistic Adversarial Testing of Computer-Use Agents in Hybrid Web-OS Environments", "authors": [ "Zeyi Liao", "Jaylen Jones", "Linxi Jiang", "Yuting Ning", "Eric Fosler-Lussier", "Yu Su", "Zhiqiang Lin", "Huan Sun" ], "affiliation": "Ohio State University, Columbus", "summary": "We provide a realistic, controlled and hybrid sandbox for systematic adversarial testings against computer-use agents.", "links": { "openreview": "https://openreview.net/forum?id=yWwrgcBoK3", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-26", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Exploratory Causal Inference in SAEnce", "authors": [ "Tommaso Mencattini", "Riccardo Cadei", "Francesco Locatello" ], "affiliation": "EPFL - EPF Lausanne", "summary": "New method to uncover causal treatment effects directly from trial data using foundation models, SAE and recursive stratification, without any prior and supervision.", "links": { "openreview": "https://openreview.net/forum?id=Ml8t8kQMUP", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-27", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Every Language Model Has a Forgery-Resistant Signature", "authors": [ "Matthew Finlayson", "Xiang Ren", "Swabha Swayamdipta" ], "affiliation": "University of Southern California", "summary": "We show that all language models impose elliptical constraints on their outputs, which can be used as a hard-to-fake signature to identify a model from its outputs.", "links": { "openreview": "https://openreview.net/forum?id=vLFqOoMBol", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-28", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Pre-training under infinite compute", "authors": [ "Konwoo Kim", "Suhas Kotha", "Percy Liang", "Tatsunori Hashimoto" ], "affiliation": "", "summary": "Since compute grows faster than the web, we design simple recipes that improve the asymptote of compute scaling laws to be 5x data efficient, offering better performance with sufficient compute.", "links": { "openreview": "https://openreview.net/forum?id=ck0aZTAnwK", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-29", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning", "authors": [ "Lakshya A Agrawal", "Shangyin Tan", "Dilara Soylu", "Noah Ziems", "Rishi Khare", "Krista Opsahl-Ong", "Arnav Singhvi", "Herumb Shandilya", "Michael J Ryan", "Meng Jiang", "Christopher Potts", "Koushik Sen", "Alex Dimakis", "Ion Stoica", "Dan Klein", "Matei Zaharia", "Omar Khattab" ], "affiliation": "University of California, Berkeley", "summary": "GEPA uses natural language reflection to optimize prompts, outperforming GRPO and MIPROv2 while needing far fewer rollouts.", "links": { "openreview": "https://openreview.net/forum?id=RQm2KQTM5r", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-30", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "p-less Sampling: A Robust Hyperparameter-Free Approach for LLM Decoding", "authors": [ "Runyan Tan", "Shuang Wu", "Phillip Howard" ], "affiliation": "AI Labs, Thoughtworks", "summary": "P-less Sampling: A parameterless sampling strategy grounded in information theory, where the truncation threshold adapts to the entire token probability distribution, is bounded and valid, and dynamically adjusts with temperature.", "links": { "openreview": "https://openreview.net/forum?id=ItFuNJQGH4", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-31", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "EigenBench: A Comparative Behavioral Measure of Value Alignment", "authors": [ "Jonathn Chang", "Leonhard Piff", "Suvadip Sana", "Jasmine Xinze Li", "Lionel Levine" ], "affiliation": "Cornell University", "summary": "Aligning AI with human values is a pressing unsolved problem. To address the lack of quantitative metrics for value alignment, we propose EigenBench: a black-box method for comparatively benchmarking language models’ values.", "links": { "openreview": "https://openreview.net/forum?id=fm79KXJIUQ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-32", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "A Representer Theorem for Hawkes Processes via Penalized Least Squares Minimization", "authors": [ "Hideaki Kim", "Tomoharu Iwata" ], "affiliation": "NTT", "summary": "The representer theorem is a cornerstone of kernel methods, which aim to estimate latent functions in reproducing kernel Hilbert spaces (RKHSs) in a nonparametric manner. Its significance lies in converting inherently infinite-dimensional optimization problems into finite-dimensional ones over dual coefficients, thereby enabling practical and co...", "links": { "openreview": "https://openreview.net/forum?id=gJjRdLG5MY", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-33", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "UALM: Unified Audio Language Model for Understanding, Generation and Reasoning", "authors": [ "Jinchuan Tian", "Sang-gil Lee", "Zhifeng Kong", "Sreyan Ghosh", "Arushi Goel", "Chao-Han Huck Yang", "Wenliang Dai", "Zihan Liu", "Hanrong Ye", "Shinji Watanabe", "Mohammad Shoeybi", "Bryan Catanzaro", "Rafael Valle", "Wei Ping" ], "affiliation": "CMU, Carnegie Mellon University", "summary": "This paper introduces UALM, an audio language model designed to unify audio understanding, generation, and reasoning", "links": { "openreview": "https://openreview.net/forum?id=TsdlOjcQNu", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-34", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Agent Data Protocol: Unifying Datasets for Diverse, Effective Fine-tuning of LLM Agents", "authors": [ "Yueqi Song", "Ketan Ramaneti", "Zaid Sheikh", "Ziru Chen", "Boyu Gou", "Tianbao Xie", "Yiheng Xu", "Danyang Zhang", "Apurva Gandhi", "Fan Yang", "Joseph Liu", "Tianyue Ou", "Zhihao Yuan", "Frank F. Xu", "Shuyan Zhou", "Xingyao Wang", "Xiang Yue", "Tao Yu", "Huan Sun", "Yu Su", "Graham Neubig" ], "affiliation": "Carnegie Mellon University", "summary": "We propose Agent Data Protocol (ADP), a lightweight \"interlingua\" schema that standardizes heterogeneous agent trajectories so datasets can plug into multiple agent SFT pipelines without per-dataset engineering.", "links": { "openreview": "https://openreview.net/forum?id=tG6301ORHd", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-35", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "OpenApps: Simulating Environment Variations to Measure UI Agent Reliability", "authors": [ "Karen Ullrich", "Jingtong Su", "Claudia Shi", "Arjun Subramonian", "Amir Bar", "Ivan Evtimov", "Nikolaos Tsilivis", "Randall Balestriero", "Julia Kempe", "Mark Ibrahim" ], "affiliation": "Meta AI", "summary": "We introduce a new environment, OpenApps, for generating thousands of versions of apps to test UI agent reliability.", "links": { "openreview": "https://openreview.net/forum?id=cj1MAx7lKs", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-36", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Overparametrization bends the landscape: BBP transitions at initialization in simple Neural Networks", "authors": [ "Brandon Livio Annesi", "Dario Bocchi", "Chiara Cammarota" ], "affiliation": "University of Roma \"La Sapienza\"", "summary": "We quantitatively analyze how overparametrization reshapes the high-dimensional loss landscape of a teacher–student setup in random positions, showing it can anticipate and qualitatively alter transitions between successful and failed signal recovery", "links": { "openreview": "https://openreview.net/forum?id=xDLE5n3x9Y", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-37", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Softmax Transformers are Turing-Complete", "authors": [ "Hongjian Jiang", "Michael Hahn", "Georg Zetzsche", "Anthony Widjaja Lin" ], "affiliation": "Rheinland-Pfälzische Technische Universität", "summary": "Hard attention Chain-of-Thought (CoT) transformers are known to be Turing-complete. However, it is an open problem whether softmax attention Chain-of-Thought (CoT) transformers are Turing-complete.", "links": { "openreview": "https://openreview.net/forum?id=FdkPOHlChS", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-38", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Steering the Herd: A Framework for LLM-based Control of Social Learning", "authors": [ "Raghu Arghal", "Kevin He", "Shirin Saeedi Bidokhti", "Saswati Sarkar" ], "affiliation": "University of Pennsylvania, University of Pennsylvania", "summary": "We introduce, analyze, and simulate (via LLMs) a model of controlled social learning to study how algorithms can influence social beliefs, decisions, and welfare via information design.", "links": { "openreview": "https://openreview.net/forum?id=RtS4UqSmNt", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-39", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Quantitative Bounds for Length Generalization in Transformers", "authors": [ "Zachary Izzo", "Eshaan Nichani", "Jason D. Lee" ], "affiliation": "NEC Labs America", "summary": "We provide an upper bound on the length of training sequences required for a transformer to generalize to sequences of arbitrary lengths.", "links": { "openreview": "https://openreview.net/forum?id=TLSUIyBIfs", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-40", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "AgentGym-RL: An Open-Source Framework to Train LLM Agents for Long-Horizon Decision Making via Multi-Turn RL", "authors": [ "Zhiheng Xi", "Jixuan Huang", "Chenyang Liao", "Baodai Huang", "Jiaqi Liu", "Honglin Guo", "yajie yang", "Rui Zheng", "Junjie Ye", "Jiazheng Zhang", "Wenxiang Chen", "Wei He", "Yiwen Ding", "Guanyu Li", "Zehui Chen", "Zhengyin Du", "Xuesong Yao", "Yufei Xu", "Jiecao Chen", "Tao Gui", "Zuxuan Wu", "Qi Zhang", "Xuanjing Huang", "Yu-Gang Jiang" ], "affiliation": "Fudan University", "summary": "We present AgentGym-RL, a unified open-source framework for training LLM agents from scratch across diverse and realistic environments, and propose ScalingInter-RL, a staged training strategy for stable long-horizon RL training.", "links": { "openreview": "https://openreview.net/forum?id=ZgCCDwcGwn", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-41", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "LLM Fingerprinting via Semantically Conditioned Watermarks", "authors": [ "Thibaud Gloaguen", "Robin Staab", "Nikola Jovanović", "Martin Vechev" ], "affiliation": "ETH Zurich", "summary": "We introduce a robust LLM fingerprinting method based on semantically conditioned watermarks", "links": { "openreview": "https://openreview.net/forum?id=t38nZqqi3Z", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-42", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Watch your steps: Dormant Adversarial Behaviors that Activate upon LLM Finetuning", "authors": [ "Thibaud Gloaguen", "Mark Vero", "Robin Staab", "Martin Vechev" ], "affiliation": "ETH Zurich", "summary": "We show that adversaries can implant hidden adversarial behaviors in LLM that are inadvertently triggered by users finetuning the model.", "links": { "openreview": "https://openreview.net/forum?id=yfM2e8Icsw", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-43", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Universal Inverse Distillation for Matching Models with Real-Data Supervision (No GANs)", "authors": [ "Nikita Maksimovich Kornilov", "David Li", "Tikhon Mavrin", "Aleksei Leonov", "Nikita Gushchin", "Evgeny Burnaev", "Iaroslav Sergeevich Koshelev", "Alexander Korotin" ], "affiliation": "Applied AI Institute", "summary": "While achieving exceptional generative quality, modern diffusion, flow, and other matching models suffer from slow inference, as they require many steps of iterative generation. Recent distillation methods address this problem by training efficient one-step generators under the guidance of a pre-trained teacher model.", "links": { "openreview": "https://openreview.net/forum?id=8NuN5UzXLC", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-44", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource", "authors": [ "Houyi Li", "Ka Man Lo", "Shijie Xuyang", "Ziqi Wang", "Wenzhen Zheng", "Haocheng Zhang", "Zhao Li", "Shuigeng Zhou", "Xiangyu Zhang", "Daxin Jiang" ], "affiliation": "Fudan University", "summary": "Mixture-of-Experts (MoE) language models dramatically expand model capacity and achieve remarkable performance without increasing per-token compute. However, can MoEs surpass dense architectures under strictly equal resource constraints — that is, when the total parameter count, training compute, and data budget are identical?", "links": { "openreview": "https://openreview.net/forum?id=oIdzliJAeA", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-45", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Pareto-Conditioned Diffusion Models for Offline Multi-Objective Optimization", "authors": [ "Jatan Shrestha", "Santeri Heiskanen", "Kari Hepola", "Severi Rissanen", "Pekka Jääskeläinen", "Joni Pajarinen" ], "affiliation": "Aalto University", "summary": "We propose Pareto-Conditioned Diffusion (PCD), a novel framework for Offline Multi-Objective Optimization", "links": { "openreview": "https://openreview.net/forum?id=S2Q00li155", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-46", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "The Coverage Principle: How Pre-Training Enables Post-Training", "authors": [ "Fan Chen", "Audrey Huang", "Noah Golowich", "Sadhika Malladi", "Adam Block", "Jordan T. Ash", "Akshay Krishnamurthy", "Dylan J Foster" ], "affiliation": "Massachusetts Institute of Technology", "summary": "We introduce the coverage profile, which captures the relationship between pre- and post-training performance and admits a rich statistical theory", "links": { "openreview": "https://openreview.net/forum?id=AUXvYQlQLZ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-47", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Transformers are Inherently Succinct", "authors": [ "Pascal Bergsträßer", "Ryan Cotterell", "Anthony Widjaja Lin" ], "affiliation": "Universität Kaiserslautern", "summary": "We study succinctness as a measure of the expressive power of transformers. Succinctness---how compactly a formalism can describe a language relative to other formalisms---is a classical notion in logic and automata theory.", "links": { "openreview": "https://openreview.net/forum?id=Yxz92UuPLQ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-48", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Conformal Robustness Control: A New Strategy for Robust Decision", "authors": [ "Yang Hu", "Jieren Tan", "Changliang Zou", "Yajie Bao", "Haojie Ren" ], "affiliation": "Shanghai Jiaotong University", "summary": "This paper develops a new strategy for robust decision problems via conformal robustness control.", "links": { "openreview": "https://openreview.net/forum?id=bt4Ahpemmi", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-49", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "HATSolver: Learning Gröbner Bases with Hierarchical Attention Transformers", "authors": [ "Mohamed Malhou", "Ludovic Perret", "Kristin E. Lauter" ], "affiliation": "", "summary": "Efficient hierarchical attention transformers for learning to solve non-linear equations through by computing groebner bases.", "links": { "openreview": "https://openreview.net/forum?id=5C3LljOEGC", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-50", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "It's All Just Vectorization: einx, a Universal Notation for Tensor Operations", "authors": [ "Florian Fervers", "Sebastian Bullinger", "Christoph Bodensteiner", "Michael Arens" ], "affiliation": "Fraunhofer IOSB", "summary": "We introduce einx, a universal notation for tensor operations, and provide a Python implementation.", "links": { "openreview": "https://openreview.net/forum?id=QqvQ3iAdpC", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-51", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Fast training of accurate physics-informed neural networks without gradient descent", "authors": [ "Chinmay Datar", "Taniya Kapoor", "Abhishek Chandra", "Qing Sun", "Erik Lien Bolager", "Iryna Burak", "Anna Veselovska", "Massimo Fornasier", "Felix Dietrich" ], "affiliation": "Technische Universität München", "summary": "Our approach - Frozen-PINNs addresses longstanding training and accuracy bottlenecks of Physics-Informed Neural Networks (PINNs) and makes PINNs highly realize high-precision, temporal causality, and extremely fast training.", "links": { "openreview": "https://openreview.net/forum?id=3VdSuh3sie", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-52", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "In-Place Test-Time Training", "authors": [ "Guhao Feng", "Shengjie Luo", "Kai Hua", "Ge Zhang", "Wenhao Huang", "Di He", "Tianle Cai" ], "affiliation": "Peking University", "summary": "The static \"train then deploy\" paradigm fundamentally limits Large Language Models (LLMs) from dynamically adapting their weights in response to continuous streams of new information inherent in real-world tasks. Test-Time Training (TTT) offers a compelling alternative by updating a subset of model parameters (fast weights) at inference time, ye...", "links": { "openreview": "https://openreview.net/forum?id=dTWfCLSoyl", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-53", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format", "authors": [ "Zhehao Huang", "Yuhang Liu", "Baijiong Lin", "Yixin Lou", "Zhengbao He", "Hanling Tian", "Tao Li", "Xiaolin Huang" ], "affiliation": "Shanghai Jiaotong University", "summary": "Large reasoning models (LRMs) excel at a long chain of reasoning but often fail to faithfully follow instructions regarding output format, constraints, or specific requirements. We investigate whether this gap can be closed by integrating an instruction-tuned model (ITM) into an LRM.", "links": { "openreview": "https://openreview.net/forum?id=PO2iULmu5e", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-54", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "To Infinity and Beyond: Tool-Use Unlocks Length Generalization in State Space Models", "authors": [ "Eran Malach", "Omid Saremi", "Sinead Williamson", "Arwen Bradley", "Aryo Lotfi", "Emmanuel Abbe", "Joshua M. Susskind", "Etai Littwin" ], "affiliation": "Apple", "summary": "State Space Models (SSMs) have become the leading alternative to Transformers for sequence modeling tasks. Their primary advantage is efficiency in long-context and long-form generation, enabled by fixed-size memory and linear scaling of computational complexity.", "links": { "openreview": "https://openreview.net/forum?id=sSfep4udCb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-55", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "TileLang: Bridge Programmability and Performance in Modern Neural Kernels", "authors": [ "Lei Wang", "Yu Cheng", "Yining Shi", "Zhiwen Mo", "Zhengju Tang", "Wenhao Xie", "Tong Wu", "Lingxiao Ma", "Yuqing Xia", "Jilong Xue", "Fan Yang", "Zhi Yang" ], "affiliation": "Peking University", "summary": "We introduce TileLang, a controllable programming system for fused neural kernels.", "links": { "openreview": "https://openreview.net/forum?id=Jb1WkNSfUB", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-56", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Navigating the Latent Space Dynamics of Neural Models", "authors": [ "Marco Fumero", "Luca Moschella", "Emanuele Rodolà", "Francesco Locatello" ], "affiliation": "", "summary": "Neural networks transform high-dimensional data into compact, structured representations, often modeled as elements of a lower dimensional latent space. In this paper, we present an alternative interpretation of neural models as dynamical systems acting on the latent manifold.", "links": { "openreview": "https://openreview.net/forum?id=Zunww3FHPU", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-57", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Non-Asymptotic Analysis of (Sticky) Track-and-Stop", "authors": [ "Riccardo Poiani", "Martino Bernasconi", "Andrea Celli" ], "affiliation": "Bocconi University", "summary": "We derive non-asymptotic guarantees for the Track-and-Stop and Sticky Track-and-Stop algorithms.", "links": { "openreview": "https://openreview.net/forum?id=vebqP5aioj", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-58", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "The Shape of Adversarial Influence: Characterizing LLM Latent Spaces with Persistent Homology", "authors": [ "Aideen Fay", "Inés García-Redondo", "Qiquan Wang", "Haim Dubossarsky", "Anthea Monod" ], "affiliation": "Imperial College London", "summary": "We use persistent homology to interpret how adversarial inputs reshape LLM representation spaces, resulting in a robust signature that provides multiscale, geometry-aware insights complementary to standard interpretability methods.", "links": { "openreview": "https://openreview.net/forum?id=v2PglvLLKT", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-59", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Hyperparameter Trajectory Inference with Conditional Lagrangian Optimal Transport", "authors": [ "Harry Amad", "Mihaela van der Schaar" ], "affiliation": "University of Cambridge", "summary": "Neural networks (NNs) often have critical behavioural trade-offs that are set at design time with hyperparameters—such as reward weights in reinforcement learning or quantile targets in regression. Post-deployment, however, user preferences can evolve, making initial settings undesirable, necessitating potentially expensive retraining.", "links": { "openreview": "https://openreview.net/forum?id=P5B97gZwRb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-60", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Vid-LLM: A Compact Video-based 3D Multimodal LLM with Reconstruction–Reasoning Synergy", "authors": [ "Haijier Chen", "Bo Xu", "Shoujian zhang", "Haoze Liu", "Jiaxuan Lin", "Jingrong Wang" ], "affiliation": "Wuhan University", "summary": "Recent developments in Multimodal Large Language Models (MLLMs) have significantly improved Vision–Language (VL) reasoning in 2D domains. However, extending these capabilities to 3D scene understanding remains a major challenge.", "links": { "openreview": "https://openreview.net/forum?id=l1cLdEjESj", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-61", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "From Markov to Laplace: How Mamba In-Context Learns Markov Chains", "authors": [ "Marco Bondaschi", "Nived Rajaraman", "Xiuying Wei", "Razvan Pascanu", "Caglar Gulcehre", "Michael Gastpar", "Ashok Vardhan Makkuva" ], "affiliation": "EPFL - EPF Lausanne", "summary": "We uncover an interesting phenomenon where a single-layer Mamba represents the Bayes optimal Laplacian smoothing estimator when trained on Markov chains and we demonstrate it theoretically and empirically.", "links": { "openreview": "https://openreview.net/forum?id=kmK3WSCOCT", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-62", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Huxley-G\\\"odel Machine: Human-Level Coding Agent Development by an Approximation of the Optimal Self-Improving Machine", "authors": [ "Wenyi Wang", "Piotr Piękos", "Li Nanbo", "Firas Laakom", "Yimeng Chen", "Mateusz Ostaszewski", "Mingchen Zhuge", "Jürgen Schmidhuber" ], "affiliation": "King Abdullah University of Science and Technology", "summary": "We propose Huxley-G\\\"odel Machine, an algorithm guideing self-improvements following an estimation of the value function of G\\\"odel Machines.", "links": { "openreview": "https://openreview.net/forum?id=T0EiEuhOOL", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-63", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Gaia2: Benchmarking LLM Agents on Dynamic and Asynchronous Environments", "authors": [ "Romain Froger", "Pierre Andrews", "Matteo Bettini", "Amar Budhiraja", "Ricardo Silveira Cabral", "Virginie Do", "Emilien Garreau", "Jean-Baptiste Gaya", "Hugo Laurençon", "Maxime Lecanu", "Kunal Malkan", "Dheeraj Mekala", "Pierre Menard", "Gerard Moreno-Torres Bertran", "Ulyana Piterbarg", "Mikhail Plekhanov", "Mathieu Rita", "Andrey Rusakov", "Vladislav Vorotilov", "Mengjue Wang", "Ian Yu", "Amine Benhalloum", "Grégoire Mialon", "Thomas Scialom" ], "affiliation": "Facebook", "summary": "Gaia2 evaluates LLM agents in asynchronous, dynamic environments with action-level verification, revealing fundamental trade-offs between reasoning, speed, and robustness.", "links": { "openreview": "https://openreview.net/forum?id=9gw03JpKK4", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-64", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Fast Escape, Slow Convergence: Learning Dynamics of Phase Retrieval under Power-Law Data", "authors": [ "Guillaume Braun", "Bruno Loureiro", "Minh Ha Quang", "Masaaki Imaizumi" ], "affiliation": "RIKEN", "summary": "Scaling laws describe how learning performance improves with data, compute, or training time, and have become a central theme in modern deep learning. We study this phenomenon in a canonical nonlinear model: phase retrieval with anisotropic Gaussian inputs whose covariance spectrum follows a power law.", "links": { "openreview": "https://openreview.net/forum?id=Ae4eZpkXBX", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-65", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "AutoEP: LLMs-Driven Automation of Hyperparameter Evolution for Metaheuristic Algorithms", "authors": [ "Zhenxing Xu", "Yizhe Zhang", "Weidong Bao", "Hao Wang", "Ming Chen", "Haoran Ye", "Wenzheng Jiang", "Hui Yan", "Ji Wang" ], "affiliation": "National University of Defense Technology", "summary": "Dynamically configuring algorithm hyperparameters is a fundamental challenge in computational intelligence. While learning-based methods offer automation, they suffer from prohibitive sample complexity and poor generalization.", "links": { "openreview": "https://openreview.net/forum?id=hit3hGBheP", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-66", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "On the Wasserstein Geodesic Principal Component Analysis of probability measures", "authors": [ "Nina Vesseron", "Elsa Cazelles", "Alice Le Brigant", "Klein" ], "affiliation": "Ecole Nationale de la Statistique et de l'Administration Economique", "summary": "This paper focuses on Geodesic Principal Component Analysis (GPCA) on a collection of probability distributions using the Otto-Wasserstein geometry. The goal is to identify geodesic curves in the space of probability measures that best capture the modes of variation of the underlying dataset.", "links": { "openreview": "https://openreview.net/forum?id=OJupg4mDjS", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-67", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Discount Model Search for Quality Diversity Optimization in High-Dimensional Measure Spaces", "authors": [ "Bryon Tjanaka", "Henry Chen", "Matthew Christopher Fontaine", "Stefanos Nikolaidis" ], "affiliation": "University of Southern California", "summary": "We present a method that enhances exploration in quality diversity (QD) optimization and show how this method enables new applications for QD. Project page: https://discount-models.github.io/", "links": { "openreview": "https://openreview.net/forum?id=m6Hv0yZO3n", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-68", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series", "authors": [ "Guoqi Yu", "Juncheng Wang", "Chen Yang", "Jing Qin", "Angelica I Aviles-Rivero", "Shujun Wang" ], "affiliation": "", "summary": "We propose a centralized module to replace decentralized attention in Transformer for centralized medical time series like EEG and ECG.", "links": { "openreview": "https://openreview.net/forum?id=oZJFY2BQt2", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-69", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention", "authors": [ "Haiquan Qiu", "Quanming Yao" ], "affiliation": "Tsinghua University, Tsinghua University", "summary": "For the first time, we mechanistically explain why low-precision training with flash attention fails, identifying a vicious cycle of rounding errors and proposing a simple, effective fix.", "links": { "openreview": "https://openreview.net/forum?id=0jHyEKHDyx", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-70", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Overcoming Joint Intractability with Lossless Hierarchical Speculative Decoding", "authors": [ "Yuxuan Zhou", "Fei Huang", "Heng Li", "Fengyi Wu", "Tianyu Wang", "jianwei zhang", "Junyang Lin", "Zhi-Qi Cheng" ], "affiliation": "Baidu", "summary": "Hierarchical speculative decoding enables lossless long-prefix verification with substantially improved decoding efficiency.", "links": { "openreview": "https://openreview.net/forum?id=LaVrNaBNwM", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-71", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "In-the-Flow Agentic System Optimization for Effective Planning and Tool Use", "authors": [ "Zhuofeng Li", "Haoxiang Zhang", "Seungju Han", "Sheng Liu", "Jianwen Xie", "Yu Zhang", "Yejin Choi", "James Zou", "Pan Lu" ], "affiliation": "Texas A&M University - College Station", "summary": "We introduce AgentFlow, a trainable agentic system, and Flow-GRPO, an on-policy RL algorithm that optimizes the planner \"in-the-flow\" by broadcasting a final outcome reward to all steps, enabling effective long-horizon planning and tool use.", "links": { "openreview": "https://openreview.net/forum?id=Mf5AleTUVK", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-72", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "What's In My Human Feedback? Learning Interpretable Descriptions of Preference Data", "authors": [ "Rajiv Movva", "Smitha Milli", "Sewon Min", "Emma Pierson" ], "affiliation": "University of California, Berkeley", "summary": "We present WIMHF, a method to describe the preferences encoded by human feedback; produce insights from seven widely-used datasets; and show that the method enables new approaches to data curation and personalization.", "links": { "openreview": "https://openreview.net/forum?id=sC6A1bFDUt", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-73", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "SimuHome: A Temporal- and Environment-Aware Benchmark for Smart Home LLM Agents", "authors": [ "Gyuhyeon Seo", "Jungwoo Yang", "Junseong Pyo", "Nalim Kim", "Jonggeun Lee", "Yohan Jo" ], "affiliation": "Seoul National University", "summary": "We introduce $\\textbf{SimuHome}$, a high-fidelity smart home simulator and a benchmark of 600 episodes for LLM-based smart home agents. Existing smart home benchmarks treat the home as a static system, neither simulating how device operations affect environmental variables over time nor supporting workflow scheduling of device commands.", "links": { "openreview": "https://openreview.net/forum?id=LCS1WsGvha", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-74", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Reasoning with Sampling: Your Base Model is Smarter Than You Think", "authors": [ "Aayush Karan", "Yilun Du" ], "affiliation": "Harvard University", "summary": "We find a training-free sampling algorithm that achieves reasoning boosts on base models comparable to those obtained by RL techniques.", "links": { "openreview": "https://openreview.net/forum?id=Vsgq2ldr4K", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-75", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Learning to Segment for Vehicle Routing Problems", "authors": [ "Wenbin Ouyang", "Sirui Li", "Yining Ma", "Cathy Wu" ], "affiliation": "Massachusetts Institute of Technology", "summary": "Iterative heuristics are widely recognized as state-of-the-art for Vehicle Routing Problems (VRPs). In this work, we exploit a critical observation: a large portion of the solution remains stable, i.e., unchanged across search iterations, causing redundant computations, especially for large-scale VRPs with long subtours.", "links": { "openreview": "https://openreview.net/forum?id=pN261iTKvr", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-76", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "WAVE: Learning Unified & Versatile Audio-Visual Embeddings with Multimodal LLM", "authors": [ "Changli Tang", "Qinfan Xiao", "Ke Mei", "Tianyi Wang", "Fengyun Rao", "Chao Zhang" ], "affiliation": "Tsinghua University, Tsinghua University", "summary": "This paper builds a versatile audio-visual embedding LLM, which can not only achieve any-to-any retrieval but also generate prompt-aware embeddings.", "links": { "openreview": "https://openreview.net/forum?id=MiV3WXDYJb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-77", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "mCLM: A Modular Chemical Language Model that Generates Functional and Makeable Molecules", "authors": [ "Carl Edwards", "Chi Han", "Gawon Lee", "Thao Nguyen", "Sara Szymkuć", "Chetan Kumar Prasad", "Bowen Jin", "Jiawei Han", "Ying Diao", "Ge Liu", "Hao Peng", "Bartosz Andrzej Grzybowski", "Martin D. Burke", "Heng Ji" ], "affiliation": "Genentech", "summary": "We propose mCLM: a bilingual, modular Chemical-Language Model that understands both natural language descriptions of functions and molecular blocks; mCLM front-loads synthesizability while improving the functions of molecules in a principled manner.", "links": { "openreview": "https://openreview.net/forum?id=r2HG3xOMJI", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-78", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Seeing Through the Brain: New Insights from Decoding Visual Stimuli with fMRI", "authors": [ "Zheng Huang", "Enpei Zhang", "Weikang Qiu", "Yinghao Cai", "Carl Yang", "Elynn Chen", "Xiang Zhang", "Rex Ying", "Dawei Zhou", "Yujun Yan" ], "affiliation": "", "summary": "We present PRISM, a framework to decode visual stimuli from fMRI with language model alignment", "links": { "openreview": "https://openreview.net/forum?id=88ZLp7xYxw", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-79", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Temporal Sparse Autoencoders: Leveraging the Sequential Nature of Language for Interpretability", "authors": [ "Usha Bhalla", "Alex Oesterling", "Claudio Mayrink Verdun", "Himabindu Lakkaraju", "Flavio Calmon" ], "affiliation": "", "summary": "We propose that using contextual information to train SAEs will improve their representation of semantic and high-level features.", "links": { "openreview": "https://openreview.net/forum?id=bojVI4l9Kn", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-80", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "VibeVoice: Expressive Podcast Generation with Next-Token Diffusion", "authors": [ "Zhiliang Peng", "Jianwei Yu", "Wenhui Wang", "Yaoyao Chang", "Yutao Sun", "Li Dong", "Yi Zhu", "Weijiang Xu", "Hangbo Bao", "Zehua Wang", "Shaohan Huang", "Yan Xia", "Furu Wei" ], "affiliation": "Microsoft", "summary": "VibeVoice can synthesize long-form speech for up to 90 minutes (in a 64K context window length) with a maximum of 4 speakers, capturing the authentic conversational vibe and surpassing open-source and proprietary dialogue models.", "links": { "openreview": "https://openreview.net/forum?id=FihSkzyxdv", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-81", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "AdAEM: An Adaptively and Automated Extensible Measurement of LLMs' Value Difference", "authors": [ "Jing Yao", "Shitong Duan", "Xiaoyuan Yi", "Dongkuan Xu", "Peng Zhang", "Tun Lu", "Ning Gu", "Zhicheng Dou", "Xing Xie" ], "affiliation": "Microsoft", "summary": "This paper proposes aa novel dynamic and automated evaluation framework to probe LLMs' value orientations and value differences", "links": { "openreview": "https://openreview.net/forum?id=qNlTH4kYJZ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-82", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Omni-Reward: Towards Generalist Omni-Modal Reward Modeling with Free-Form Preferences", "authors": [ "Zhuoran Jin", "Hongbang Yuan", "Kejian Zhu", "Jiachun Li", "Pengfei Cao", "Yubo Chen", "Kang Liu", "Jun Zhao" ], "affiliation": "Institute of Automation, Chinese Academy of Sciences", "summary": "We propose Omni-Reward, a step towards universal omni-modal reward modeling with free-form preferences.", "links": { "openreview": "https://openreview.net/forum?id=9C4gVbPqSy", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-83", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Spherical Watermark: Encryption-Free, Lossless Watermarking for Diffusion Models", "authors": [ "Xiaoxiao Hu", "Jiaqi Jin", "Sheng Li", "Wanli Peng", "Xinpeng Zhang", "Zhenxing Qian" ], "affiliation": "Fudan University", "summary": "Employing a novel spherical mapping mechanism, we propose a novel lossless watermarking scheme for text-to-image diffusion models.", "links": { "openreview": "https://openreview.net/forum?id=2eAGrunxVz", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-84", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Rodrigues Network for Learning Robot Actions", "authors": [ "Jialiang Zhang", "Haoran Geng", "Yang You", "Congyue Deng", "Pieter Abbeel", "Jitendra Malik", "Leonidas Guibas" ], "affiliation": "Peking University", "summary": "We design a new neural network, the Rodrigues Network (RodriNet), that addresses the kinematic structural priors in articulated robot action learning.", "links": { "openreview": "https://openreview.net/forum?id=IZHk6BXBST", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-85", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "CyberGym: Evaluating AI Agents' Real-World Cybersecurity Capabilities at Scale", "authors": [ "Zhun Wang", "Tianneng Shi", "Jingxuan He", "Matthew Cai", "Jialin Zhang", "Dawn Song" ], "affiliation": "", "summary": "AI agents have significant potential to reshape cybersecurity, making a thorough assessment of their capabilities critical. However, existing evaluations fall short, because they are based on small-scale benchmarks and only measure static outcomes, failing to capture the full, dynamic range of real-world security challenges.", "links": { "openreview": "https://openreview.net/forum?id=2YvbLQEdYt", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-86", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Energy-Based Transformers are Scalable Learners and Thinkers", "authors": [ "Alexi Gladstone", "Ganesh Nanduru", "Md Mofijul Islam", "Peixuan Han", "Hyeonjeong Ha", "Aman Chadha", "Yilun Du", "Heng Ji", "Jundong Li", "Tariq Iqbal" ], "affiliation": "Flapping Airplanes", "summary": "We introduce Energy-Based Transformers, a scalable new approach for learning how to think from unsupervised learning, generalizing current System 2 Thinking/reasoning approaches.", "links": { "openreview": "https://openreview.net/forum?id=ZBj3Qp1bYg", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-87", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Global Resolution: Optimal Multi-Draft Speculative Sampling via Convex Optimization", "authors": [ "Rahul Krishna Thomas", "Arka Pal" ], "affiliation": "Columbia University", "summary": "We reduce optimal multi-draft speculative sampling to a convex minimization problem, and solve a truncated version to achieve state-of-the-art acceptance with negligible performance degradation.", "links": { "openreview": "https://openreview.net/forum?id=gpsczXOsHn", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-88", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "FALCON: Few-step Accurate Likelihoods for Continuous Flows", "authors": [ "Danyal Rehman", "Tara Akhound-Sadegh", "Artem Gazizov", "Yoshua Bengio", "Alexander Tong" ], "affiliation": "Mila - Quebec Artificial Intelligence Institute", "summary": "Few-step Flow Matching with Accurate Likelihoods for Scalable Boltzmann Generators", "links": { "openreview": "https://openreview.net/forum?id=FbssShlI4N", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-89", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "How Reliable is Language Model Micro-Benchmarking?", "authors": [ "Gregory Yauney", "Shahzaib Saqib Warraich", "Swabha Swayamdipta" ], "affiliation": "Amazon", "summary": "Micro-benchmarking offers a solution to the often prohibitive time and cost of language model development: evaluate on a very small subset of existing benchmarks. Can these micro-benchmarks, however, rank models as consistently as the full benchmarks they replace?", "links": { "openreview": "https://openreview.net/forum?id=cReExMQLiK", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-90", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "EDIT-Bench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits", "authors": [ "Wayne Chi", "Valerie Chen", "Ryan Shar", "Aditya Mittal", "Jenny Liang", "Wei-Lin Chiang", "Anastasios Nikolas Angelopoulos", "Ion Stoica", "Graham Neubig", "Ameet Talwalkar", "Chris Donahue" ], "affiliation": "Carnegie Mellon University", "summary": "We propose a new benchmark for evaluating an LLM's ability to perform code edits. Our data is gathered from in-the-wild code edits, leading to more realistic problems.", "links": { "openreview": "https://openreview.net/forum?id=FtL9eEmU6v", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-91", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Online Learning and Equilibrium Computation with Ranking Feedback", "authors": [ "Mingyang Liu", "Yongshan Chen", "Zhiyuan Fan", "Gabriele Farina", "Asuman E. Ozdaglar", "Kaiqing Zhang" ], "affiliation": "Massachusetts Institute of Technology", "summary": "Hardness and positive results for online learning with ranking feedback. Together with equilibrium computation with ranking feedback.", "links": { "openreview": "https://openreview.net/forum?id=lg6H2oJPky", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-92", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains", "authors": [ "Xuying Ning", "Dongqi Fu", "Tianxin Wei", "Mengting Ai", "Jiaru Zou", "Ting-Wei Li", "Hanghang Tong", "Yada Zhu", "Hendrik Hamann", "Jingrui He" ], "affiliation": "University of Illinois at Urbana-Champaign", "summary": "With the increasing demand for step-wise, cross-modal, and knowledge-grounded reasoning, multimodal large language models (MLLMs) are evolving beyond the traditional fixed retrieve-then-generate paradigm toward more sophisticated agentic multimodal retrieval-augmented generation (MM-RAG). Existing benchmarks, however, mainly focus on simplified...", "links": { "openreview": "https://openreview.net/forum?id=JEGDp1E4OH", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-93", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "True Self-Supervised Novel View Synthesis is Transferable", "authors": [ "Thomas Mitchel", "Hyunwoo Ryu", "Vincent Sitzmann" ], "affiliation": "", "summary": "The key criterion for determining whether a models is capable of NVS is transferability, and we present the first fully geometry-free and self-supervised model capable of it.", "links": { "openreview": "https://openreview.net/forum?id=aJJppqAm6r", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-94", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Revela: Dense Retriever Learning via Language Modeling", "authors": [ "Fengyu Cai", "Tong Chen", "Xinran Zhao", "Sihao Chen", "Hongming Zhang", "Tongshuang Wu", "Iryna Gurevych", "Heinz Koeppl" ], "affiliation": "Technische Universität Darmstadt", "summary": "Dense retrievers play a vital role in accessing external and specialized knowledge to augment language models (LMs). Training dense retrievers typically requires annotated query-document pairs, which are costly to create and scarce in specialized domains (e.g., code) or in complex settings (e.g., requiring reasoning).", "links": { "openreview": "https://openreview.net/forum?id=e7pAjJZJWb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-95", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "LLMs Get Lost In Multi-Turn Conversation", "authors": [ "Philippe Laban", "Hiroaki Hayashi", "Yingbo Zhou", "Jennifer Neville" ], "affiliation": "Microsoft", "summary": "We discover that when LLMs take a wrong turn in a conversation, they get lost and do not recover.", "links": { "openreview": "https://openreview.net/forum?id=VKGTGGcwl6", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-96", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "TTSDS2: Resources and Benchmark for Evaluating Human-Quality Text to Speech Systems", "authors": [ "Christoph Minixhofer", "Ondrej Klejch", "Peter Bell" ], "affiliation": "University of Edinburgh, University of Edinburgh", "summary": "With TTSDS2, we introduce a metric and benchmark for TTS, covering 14 languages, which consistently correlates with human judgements.", "links": { "openreview": "https://openreview.net/forum?id=uGai5lYHlV", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-97", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Mamba-3: Improved Sequence Modeling using State Space Principles", "authors": [ "Aakash Lahoti", "Kevin Li", "Berlin Chen", "Caitlin Wang", "Aviv Bick", "J Zico Kolter", "Tri Dao", "Albert Gu" ], "affiliation": "CMU, Carnegie Mellon University", "summary": "Mamba-3, an inference-first SSM that pushes on core SSM principles: improved discretization for better quality, complex dynamics for new capabilities, and MIMO updates for efficient inference.", "links": { "openreview": "https://openreview.net/forum?id=HwCvaJOiCj", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-98", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Is it Thinking or Cheating? Detecting Implicit Reward Hacking by Measuring Reasoning Effort", "authors": [ "Xinpeng Wang", "Nitish Joshi", "Barbara Plank", "Rico Angell", "He He" ], "affiliation": "Ludwig-Maximilians-Universität München", "summary": "TRACE detects implicit reward hacking by measuring how quickly truncated reasoning suffices to pass verification, outperforming CoT monitoring and enabling hidden loopholes discovery.", "links": { "openreview": "https://openreview.net/forum?id=Gk7gLAtVDO", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-99", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Causal Structure Learning in Hawkes Processes with Complex Latent Confounder Networks", "authors": [ "Songyao Jin", "Biwei Huang" ], "affiliation": "University of California, San Diego", "summary": "We propose a method to uncover causal relationships in partially observed multivariate Hawkes processes, despite the presence of latent subprocesses, using a discrete-time representation and a two-phase iterative algorithm.", "links": { "openreview": "https://openreview.net/forum?id=mA78uXqcnl", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-100", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Mean Flow Policy with Instantaneous Velocity Constraint for One-step Action Generation", "authors": [ "Guojian Zhan", "Letian Tao", "Pengcheng Wang", "Yixiao Wang", "Yuxin Chen", "Yiheng Li", "Hongyang Li", "Masayoshi Tomizuka", "Shengbo Eben Li" ], "affiliation": "Tsinghua University", "summary": "We introduce the mean velocity policy, a new RL policy that, along with a novel instantaneous velocity constraint, achieves state-of-the-art performance and the fastest training and inference speed.", "links": { "openreview": "https://openreview.net/forum?id=mIeKe74W43", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-101", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Radiometrically Consistent Gaussian Surfels for Inverse Rendering", "authors": [ "Kyu Beom Han", "Jaeyoon Kim", "Woo Jae Kim", "Jinhwan Seo", "Sung-eui Yoon" ], "affiliation": "Korea Advanced Institute of Science & Technology", "summary": "Radiometric Consistency for Gaussian Surfels provide accurate indirect illumination for inverse rendering", "links": { "openreview": "https://openreview.net/forum?id=lKqE7UuMvp", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-102", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Hubble: a Model Suite to Advance the Study of LLM Memorization", "authors": [ "Johnny Wei", "Ameya Godbole", "Mohammad Aflah Khan", "Ryan Yixiang Wang", "Xiaoyuan Zhu", "James Flemings", "Nitya Kashyap", "Krishna P. Gummadi", "Willie Neiswanger", "Robin Jia" ], "affiliation": "University of Southern California", "summary": "Hubble is a suite of paired LLMs (largest 8B), where the perturbed models are trained in the same way as standard models but with text (e.g. book passages, biographies, and test sets) inserted and designed to emulate key memorization risks.", "links": { "openreview": "https://openreview.net/forum?id=ZfdnZhOP0k", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-103", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Differentially Private Domain Discovery", "authors": [ "Vinod Raman", "Travis Dick", "Matthew Joseph" ], "affiliation": "Google DeepMind", "summary": "We study several problems in differentially private domain discovery, where each user holds a subset of items from a shared but unknown domain, and the goal is to output an informative subset of items. For set union, we show that the simple baseline Weighted Gaussian Mechanism (WGM) has a near-optimal $\\ell_1$ missing mass guarantee on Zipfian d...", "links": { "openreview": "https://openreview.net/forum?id=yBpzF8hp3J", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-104", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "CauKer: Classification Time Series Foundation Models Can Be Pretrained on Synthetic Data", "authors": [ "Shifeng Xie", "Vasilii Feofanov", "Jianfeng Zhang", "Themis Palpanas", "Ievgen Redko" ], "affiliation": "Télécom Paris", "summary": "Time series foundation models (TSFMs) have recently gained significant attention due to their strong zero-shot capabilities and widespread real-world applications. Such models typically require a computationally costly pretraining on large-scale, carefully curated collections of real-world sequences.", "links": { "openreview": "https://openreview.net/forum?id=xBW2FIfswU", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-105", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Non-Convex Federated Optimization under Cost-Aware Client Selection", "authors": [ "Xiaowen Jiang", "Anton Rodomanov", "Sebastian U Stich" ], "affiliation": "CISPA Helmholtz Center for Information Security", "summary": "We propose a composite gradient method with SAGA and recursive gradient estimator for non-convex federated optimization that can exploit second-order similarity and achieve high communication and local computation efficiency.", "links": { "openreview": "https://openreview.net/forum?id=FnaDv6SMd9", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-106", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Verifying Chain-of-Thought Reasoning via Its Computational Graph", "authors": [ "Zheng Zhao", "Yeskendir Koishekenov", "Xianjun Yang", "Naila Murray", "Nicola Cancedda" ], "affiliation": "University of Edinburgh, University of Edinburgh", "summary": "We introduce CRV, a white-box methodology that treats attribution graphs as execution traces, and use it to provide evidence that flawed reasoning has a verifiable computational structure.", "links": { "openreview": "https://openreview.net/forum?id=CxiNICq0Rr", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-107", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "How Learning Rate Decay Wastes Your Best Data in Curriculum-Based LLM Pretraining", "authors": [ "Kairong Luo", "Zhenbo Sun", "Haodong Wen", "Xinyu Shi", "Jiarui Cui", "Chenyi Dang", "Kaifeng Lyu", "Wenguang Chen" ], "affiliation": "Tsinghua University", "summary": "Use model weight average to enhance curriculum learning in LLM pretraining.", "links": { "openreview": "https://openreview.net/forum?id=T5wkZJqzkz", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-108", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Probabilistic Kernel Function for Fast Angle Testing", "authors": [ "Kejing Lu", "Chuan Xiao", "Yoshiharu Ishikawa" ], "affiliation": "Yamanashi University", "summary": "We propose two probabilistic kernel functions for angle testing, which can be used to accelerate vector-based similarity search.", "links": { "openreview": "https://openreview.net/forum?id=nCsF3Bsn2n", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-109", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "RefineStat: Efficient Exploration for Probabilistic Program Synthesis", "authors": [ "Madhav Kanda", "Shubham Ugare", "Sasa Misailovic" ], "affiliation": "University of Illinois at Urbana-Champaign", "summary": "Probabilistic programming offers a powerful framework for modeling uncertainty, yet statistical model discovery in this domain entails navigating an immense search space under strict domain‐specific constraints. When small language models are tasked with generating probabilistic programs, they frequently produce outputs that suffer from both syn...", "links": { "openreview": "https://openreview.net/forum?id=SAl337ZX5d", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-110", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Scaling Atomistic Protein Binder Design with Generative Pretraining and Test-Time Compute", "authors": [ "Kieran Didi", "Zuobai Zhang", "Guoqing Zhou", "Danny Reidenbach", "Zhonglin Cao", "Sooyoung Cha", "Tomas Geffner", "Christian Dallago", "Jian Tang", "Michael M. Bronstein", "Martin Steinegger", "Emine Kucukbenli", "Arash Vahdat", "Karsten Kreis" ], "affiliation": "NVIDIA", "summary": "We introduce a novel method for state-of-the-art structure-based protein binder design that combines flow matching-based generative pretraining with inference-time compute scaling techniques.", "links": { "openreview": "https://openreview.net/forum?id=qmCpJtFZra", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-111", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Structured Flow Autoencoders: Learning Structured Probabilistic Representations with Flow Matching", "authors": [ "Yidan Xu", "Yixin Wang", "XuanLong Nguyen" ], "affiliation": "", "summary": "A framework that composes any probabilistic graphical model with flow matching, jointly learning structured latent representations and high-fidelity generative models through a single objective.", "links": { "openreview": "https://openreview.net/forum?id=KYdfvF2SZN", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-112", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Plug-and-Play Compositionality for Boosting Continual Learning with Foundation Models", "authors": [ "Weiduo Liao", "Fei Han", "Hisao Ishibuchi", "Qingfu Zhang", "Ying Wei" ], "affiliation": "City University of Hong Kong", "summary": "We introduce CompSLOT, a universal concept learning method to continual learning with foundation models system to establish a concept-level understanding of class prediction for alternative continual learners.", "links": { "openreview": "https://openreview.net/forum?id=22hBwIf7OC", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-113", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Generating metamers of human scene understanding", "authors": [ "Ritik Raina", "Abe Leite", "Alexandros Graikos", "Seoyoung Ahn", "Dimitris Samaras", "Greg Zelinsky" ], "affiliation": "State University of New York at Stony Brook", "summary": "Human vision combines low-resolution “gist” information from the visual periphery with sparse but high-resolution information from fixated locations to construct a coherent understanding of a visual scene. In this paper, we introduce MetamerGen, a tool for generating scenes that are aligned with latent human scene representations.", "links": { "openreview": "https://openreview.net/forum?id=cSDXx8V6K9", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-114", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Efficient Resource-Constrained Training of Transformers via Subspace Optimization", "authors": [ "Le-Trung Nguyen", "Enzo Tartaglione", "Van-Tam Nguyen" ], "affiliation": "Télécom Paris", "summary": "We propose a novel method that enables training vision transformer models within a low-rank subspace to optimize computational resources, making on-device learning practically feasible.", "links": { "openreview": "https://openreview.net/forum?id=0nvQ5kHXf4", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-115", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Learning with Dual-level Noisy Correspondence for Multi-modal Entity Alignment", "authors": [ "Haobin Li", "Yijie Lin", "Peng Hu", "Mouxing Yang", "Xi Peng" ], "affiliation": "School of Software Engineering, Sichuan University, Sichuan University", "summary": "Multi-modal entity alignment (MMEA) aims to identify equivalent entities across heterogeneous multi-modal knowledge graphs (MMKGs), where each entity is described by attributes from various modalities. Existing methods typically assume that both intra-entity and inter-graph correspondences are faultless, which is often violated in real-world MMK...", "links": { "openreview": "https://openreview.net/forum?id=mytIKuRsSE", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-116", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Pinet: Optimizing hard-constrained neural networks with orthogonal projection layers", "authors": [ "Panagiotis D. Grontas", "Antonio Terpin", "Efe C. Balta", "Raffaello D'Andrea", "John Lygeros" ], "affiliation": "International Business Machines", "summary": "We introduce an output layer for neural networks that ensures satisfaction of convex constraints. Our approach, $\\Pi$net, leverages operator splitting for rapid and reliable projections in the forward pass, and the implicit function theorem for backpropagation.", "links": { "openreview": "https://openreview.net/forum?id=EJ680UQeZG", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-117", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "ParaRNN: Unlocking Parallel Training of Nonlinear RNNs for Large Language Models", "authors": [ "Federico Danieli", "Pau Rodriguez", "Miguel Sarabia", "Xavier Suau", "Luca Zappella" ], "affiliation": "Apple", "summary": "We break the sequential bottleneck of nonlinear RNNs, enabling training of billion-scale LSTM/GRU models, competitive with modern architectures", "links": { "openreview": "https://openreview.net/forum?id=mX8b64iUaa", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-118", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Exploring Synthesizable Chemical Space with Iterative Pathway Refinements", "authors": [ "Seul Lee", "Karsten Kreis", "Srimukh Prasad Veccham", "Meng Liu", "Danny Reidenbach", "Saee Gopal Paliwal", "Weili Nie", "Arash Vahdat" ], "affiliation": "Korea Advanced Institute of Science and Technology", "summary": "A well-known pitfall of molecular generative models is that they are not guaranteed to generate synthesizable molecules. Existing solutions for this problem often struggle to effectively navigate exponentially large combinatorial space of synthesizable molecules and suffer from poor coverage.", "links": { "openreview": "https://openreview.net/forum?id=aQKVfKOkR5", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-119", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "On The Surprising Effectiveness of a Single Global Merging in Decentralized Learning", "authors": [ "Tongtian Zhu", "Tianyu Zhang", "Mingze Wang", "Zhanpeng Zhou", "Can Wang" ], "affiliation": "Zhejiang University", "summary": "We discover and theoretically explain why and when a single global parameter merging in decentralized learning can recover the performance of federated learning, even in highly heterogeneous and communication-constrained environments.", "links": { "openreview": "https://openreview.net/forum?id=zrFnwRHuQo", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-120", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Visual Planning: Let's Think Only with Images", "authors": [ "Yi Xu", "Chengzu Li", "Han Zhou", "Xingchen Wan", "Caiqi Zhang", "Anna Korhonen", "Ivan Vulić" ], "affiliation": "Mistral AI", "summary": "Recent advancements in Large Language Models (LLMs) and their multimodal extensions (MLLMs) have substantially enhanced machine reasoning across diverse tasks. However, these models predominantly rely on pure text as the medium for both expressing and structuring reasoning, even when visual information is present.", "links": { "openreview": "https://openreview.net/forum?id=wsnse46kRO", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-121", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Optimal Sparsity of Mixture-of-Experts Language Models for Reasoning Tasks", "authors": [ "Taishi Nakamura", "Satoki Ishikawa", "Masaki Kawamura", "Takumi Okamoto", "Daisuke Nohara", "Jun Suzuki", "Rio Yokota" ], "affiliation": "Institute of Science Tokyo", "summary": "Memorization skills consistently benefit from higher sparsity, while reasoning skills require balancing active FLOPs with total tokens per parameter; the optimal point shifts with the compute budget.", "links": { "openreview": "https://openreview.net/forum?id=XFw2EPRUUR", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-122", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Exploratory Diffusion Model for Unsupervised Reinforcement Learning", "authors": [ "Chengyang Ying", "Huayu Chen", "Xinning Zhou", "Zhongkai Hao", "Hang Su", "Jun Zhu" ], "affiliation": "Tsinghua University, Tsinghua University", "summary": "We propose Exploratory Diffusion Model (ExDM), boosting unsupervised exploration and few-shot fine-tuning by diffusion models.", "links": { "openreview": "https://openreview.net/forum?id=k0Kb1ynFbt", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-123", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "DCFold: Efficient Protein Structure Generation with Single Forward Pass", "authors": [ "Zhe Zhang", "Yuanning Feng", "Yuxuan Song", "Keyue Qiu", "Hao Zhou", "Wei-Ying Ma" ], "affiliation": "Tsinghua University", "summary": "AlphaFold3 introduces a diffusion-based architecture that elevates protein structure prediction to all-atom resolution with improved accuracy. This state-of-the-art performance has established AlphaFold3 as a foundation model for diverse generation and design tasks.", "links": { "openreview": "https://openreview.net/forum?id=LMsdys7t1L", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-124", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Triple-BERT: Do We Really Need MARL for Order Dispatch on Ride-Sharing Platforms?", "authors": [ "Zijian Zhao", "Sen Li" ], "affiliation": "Huawei Technologies Ltd.", "summary": "This paper proposes a novel centralized reinforcement learning framework for large-scale order dispatching tasks in ride-sharing scenarios, achieving better cooperation among workers compared to previous multi-agent methods.", "links": { "openreview": "https://openreview.net/forum?id=symgW6FhA6", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-125", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "InfoNCE Induces Gaussian Distribution", "authors": [ "Roy Betser", "Eyal Gofer", "Meir Yossef Levi", "Guy Gilboa" ], "affiliation": "Technion - Israel Institute of Technology, Technion", "summary": "Contrastive learning based representations can be well approximated by a multivariate Gaussian distribution.", "links": { "openreview": "https://openreview.net/forum?id=BlSH7gNQSq", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-126", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Coupling Experts and Routers in Mixture-of-Experts via an Auxiliary Loss", "authors": [ "Ang Lv", "Jin Ma", "Yiyuan Ma", "Siyuan Qiao" ], "affiliation": "Renmin University of China", "summary": "Mixture-of-Experts (MoE) models lack explicit constraints to ensure the router's decisions align well with the experts' capabilities, which ultimately limits model performance. To address this, we propose expert-router coupling (ERC) loss, a lightweight auxiliary loss that tightly couples the router's decisions with expert capabilities.", "links": { "openreview": "https://openreview.net/forum?id=MpeyjgWbKt", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-127", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "P-GenRM: Personalized Generative Reward Model with Test-time User-based Scaling", "authors": [ "Pinyi Zhang", "Ting-En Lin", "Yuchuan Wu", "Jingyang Chen", "Zongqi Wang", "Hua Yang", "Xu Ze", "Fei Huang", "Yongbin Li", "Kai Zhang" ], "affiliation": "East China Normal University", "summary": "The first personalized generative reward model with test-time user-based scaling for preference alignment", "links": { "openreview": "https://openreview.net/forum?id=hXNApWLBZG", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-128", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Modality-free Graph In-context Alignment", "authors": [ "Wei Zhuo", "Siqiang Luo" ], "affiliation": "Nanyang Technological University", "summary": "A pretraining framework that enables in-context learning on graph-structured data without modality assumptions.", "links": { "openreview": "https://openreview.net/forum?id=cDc95lucVL", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-129", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "TRACE: Your Diffusion Model is Secretly an Instance Edge Detector", "authors": [ "Sanghyun Jo", "Ziseok Lee", "Wooyeol Lee", "Jonghyun Choi", "Jaesik Park", "Kyungsu Kim" ], "affiliation": "OGQ", "summary": "TRACE turns pretrained diffusion models into annotation-free instance edge generators for instance and panoptic segmentation.", "links": { "openreview": "https://openreview.net/forum?id=BjElYlJKMj", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-130", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Scaling Laws and Spectra of Shallow Neural Networks in the Feature Learning Regime", "authors": [ "Leonardo Defilippis", "Yizhou Xu", "Julius Girardin", "Vittorio Erba", "Emanuele Troiani", "Lenka Zdeborová", "Bruno Loureiro", "Florent Krzakala" ], "affiliation": "Ecole Normale Supérieure, Ecole Normale Supérieure de Paris", "summary": "We derive a phase diagram of scaling laws for diagonal and quadratic neural networks via a bridge to LASSO and matrix compressed sensing, predicting both generalization and the emergence of power-law weight spectra.", "links": { "openreview": "https://openreview.net/forum?id=Q3yLIIkt7z", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-131", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent", "authors": [ "Hongli Yu", "Tinghong Chen", "Jiangtao Feng", "Jiangjie Chen", "Weinan Dai", "Qiying Yu", "Ya-Qin Zhang", "Wei-Ying Ma", "Jingjing Liu", "Mingxuan Wang", "Hao Zhou" ], "affiliation": "Tsinghua University", "summary": "We propose MemAgent, a novel agent workflow for long-text processing, demonstrating exceptional extrapolation and performance in large-scale tasks after RL Training.", "links": { "openreview": "https://openreview.net/forum?id=k5nIOvYGCL", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-132", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "TD-JEPA: Latent-predictive Representations for Zero-Shot Reinforcement Learning", "authors": [ "Marco Bagatella", "Matteo Pirotta", "Ahmed Touati", "Alessandro Lazaric", "Andrea Tirinzoni" ], "affiliation": "Max Planck Institute for Intelligent Systems, Max Planck Institute for Intelligent Systems", "summary": "We propose a temporal-difference latent-predictive method for zero-shot unsupervised RL.", "links": { "openreview": "https://openreview.net/forum?id=SzXDuBN8M1", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-133", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale", "authors": [ "Chunrui Han", "Guopeng Li", "Jingwei Wu", "Quan Sun", "Yan Cai", "Yuang Peng", "Zheng Ge", "Deyu Zhou", "Haomiao Tang", "Hongyu Zhou", "Kenkun Liu", "Shu-Tao Xia", "Binxing Jiao", "Daxin Jiang", "Xiangyu Zhang", "Yibo Zhu" ], "affiliation": "Stepfun", "summary": "Prevailing autoregressive (AR) models for text-to-image generation either rely on heavy, computationally-intensive diffusion models to process continuous image tokens, or employ vector quantization (VQ) to obtain discrete tokens with quantization loss. In this paper, we push the autoregressive paradigm forward with NextStep-1, a 14B autoregressi...", "links": { "openreview": "https://openreview.net/forum?id=Ndnwg9oOQO", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-134", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Differentiable Model Predictive Control on the GPU", "authors": [ "Emre Adabag", "Marcus Greiff", "John Subosits", "Thomas Jonathan Lew" ], "affiliation": "University of Michigan - Ann Arbor", "summary": "Differentiable model predictive control (MPC) offers a powerful framework for combining learning and control. However, its adoption has been limited by the inherently sequential nature of traditional optimization algorithms, which are challenging to parallelize on modern computing hardware like GPUs.", "links": { "openreview": "https://openreview.net/forum?id=bFYfV6c9zu", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-135", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Planner Aware Path Learning in Diffusion Language Models Training", "authors": [ "Fred Zhangzhi Peng", "Zachary Bezemek", "Jarrid Rector-Brooks", "Shuibai Zhang", "Michael M. Bronstein", "Anru Zhang", "Alexander Tong", "Joey Bose" ], "affiliation": "Duke University", "summary": "We propose Planner Aware Path Learning (PAPL), a simple planner-aligned training method for Diffusion Language Models that resolves the training–inference mismatch and consistently improves generation quality.", "links": { "openreview": "https://openreview.net/forum?id=lAlI5FuIf7", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-136", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "PateGAIL++: Utility Optimized Private Trajectory Generation with Imitation Learning", "authors": [ "Yingjie Ma", "Bijal Bharadva", "Xin Zhang", "Joann Qiongna Chen" ], "affiliation": "", "summary": "Human mobility trajectory data supports a wide range of applications, including urban planning, intelligent transportation systems, and public safety monitoring. However, large-scale, high-quality mobility datasets are difficult to obtain due to privacy concerns.", "links": { "openreview": "https://openreview.net/forum?id=Oyfz6G0hmc", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-137", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction", "authors": [ "Zilin Xiao", "Qi Ma", "Mengting Gu", "Chun-cheng Jason Chen", "Xintao Chen", "Vicente Ordonez", "Vijai Mohan" ], "affiliation": "Rice University", "summary": "Universal multimodal embedding models have achieved great success in capturing semantic relevance between queries and candidates. However, current methods either condense queries and candidates into a single vector, potentially limiting the expressiveness for fine-grained information, or produce too many vectors that are prohibitively expensive...", "links": { "openreview": "https://openreview.net/forum?id=yKDqg9HwZX", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-138", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Monocular Normal Estimation via Shading Sequence Estimation", "authors": [ "Zongrui Li", "Xinhua Ma", "Minghui Hu", "Yunqing Zhao", "Yingchen Yu", "Qian Zheng", "Chang Liu", "Xudong Jiang", "Song Bai" ], "affiliation": "Nanyang Technological University", "summary": "Monocular normal estimation aims to estimate the normal map from a single RGB image of an object under arbitrary lights. Existing methods rely on deep models to directly predict normal maps.", "links": { "openreview": "https://openreview.net/forum?id=d7itDxMD1n", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-139", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Shoot First, Ask Questions Later? Building Rational Agents that Explore and Act Like People", "authors": [ "Gabriel Grand", "Valerio Pepe", "Joshua B. Tenenbaum", "Jacob Andreas" ], "affiliation": "Massachusetts Institute of Technology", "summary": "We introduce a collaborative Battleship task to evaluate information-seeking in humans and agents; insights from Bayesian Experimental Design (BED) yield inference-time strategies for building resource-rational agents in discovery settings.", "links": { "openreview": "https://openreview.net/forum?id=EQhUvWH78U", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-140", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Intrinsic Entropy of Context Length Scaling in LLMs", "authors": [ "Jingzhe Shi", "Qinwei Ma", "Hongyi Liu", "Hang Zhao", "Jenq-Neng Hwang", "Lei Li" ], "affiliation": "Tsinghua University", "summary": "We propose to use Intrinsic Entropy for understanding impact of context length on Language Modeling, and conduct experiments to validate theoretical assumptions and deductions with language and synthetic datasets.", "links": { "openreview": "https://openreview.net/forum?id=vnipyA8c9V", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-141", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling", "authors": [ "Tal Daniel", "Carl Qi", "Dan Haramati", "Amir Zadeh", "Chuan Li", "Aviv Tamar", "Deepak Pathak", "David Held" ], "affiliation": "Carnegie Mellon University", "summary": "a self-supervised object-centric world model that learns keypoints, and masks directly from videos, supports multi-modal conditioning, scaled to real-world multi-object datasets", "links": { "openreview": "https://openreview.net/forum?id=lTaPtGiUUc", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-142", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Beyond Prompt-Induced Lies: Investigating LLM Deception on Benign Prompts", "authors": [ "Zhaomin Wu", "Mingzhe Du", "See-Kiong Ng", "Bingsheng He" ], "affiliation": "National University of Singapore", "summary": "We detected the widespread deception of LLM under benign prompts and found its tendency increases with task difficulty.", "links": { "openreview": "https://openreview.net/forum?id=PDBBYwd1LY", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-143", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "LLM DNA: Tracing Model Evolution via Functional Representations", "authors": [ "Zhaomin Wu", "Haodong Zhao", "Ziyang Wang", "Jizhou Guo", "Qian Wang", "Bingsheng He" ], "affiliation": "National University of Singapore", "summary": "We introduce LLM DNA, a low-dimensional representation of LLMs, uncovers undocumented relations and constructs phylogenetic tree for LLM.", "links": { "openreview": "https://openreview.net/forum?id=UIxHaAqFqQ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-144", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Generative Universal Verifier as Multimodal Meta-Reasoner", "authors": [ "Xinchen Zhang", "Xiaoying Zhang", "Youbin Wu", "Yanbin Cao", "Renrui Zhang", "Ruihang Chu", "Ling Yang", "Yujiu Yang", "Guang Shi" ], "affiliation": "ByteDance Seed", "summary": "We introduce *Generative Universal Verifier*, a novel concept and plugin designed for next-generation multimodal reasoning in vision-language models and unified multimodal models, providing the fundamental capability of reflection and refinement on visual outcomes during the reasoning and generation process. This work makes three main contributi...", "links": { "openreview": "https://openreview.net/forum?id=DM0Y0oL33T", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-145", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "$PhyWorldBench$: A Comprehensive Evaluation of Physical Realism in Text-to-Video Models", "authors": [ "Jing Gu", "Xian Liu", "Yu Zeng", "Ashwin Nagarajan", "Fangrui Zhu", "Daniel Hong", "Yue Fan", "Qianqi Yan", "Kaiwen Zhou", "Ming-Yu Liu", "Xin Eric Wang" ], "affiliation": "xAI", "summary": "Large-scale, multidimensional video generation for physics", "links": { "openreview": "https://openreview.net/forum?id=rlZeILv3fm", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-146", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "World-In-World: World Models in a Closed-Loop World", "authors": [ "Jiahan Zhang", "Muqing Jiang", "Nanru Dai", "TaiMing Lu", "Arda Uzunoglu", "Shunchi Zhang", "Yana Wei", "Jiahao Wang", "Vishal M. Patel", "Paul Pu Liang", "Daniel Khashabi", "Cheng Peng", "Rama Chellappa", "Tianmin Shu", "Alan Yuille", "Yilun Du", "Jieneng Chen" ], "affiliation": "", "summary": "By grounding assessment in embodied task success instead of video metrics, World-In-World provides a principled yardstick for future research on generative world models in the context of embodiment", "links": { "openreview": "https://openreview.net/forum?id=yDmb7xAfeb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-147", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Multimodal Aligned Semantic Knowledge for Unpaired Image-text Matching", "authors": [ "Laiguo Yin", "Yixin Zhang", "YUQING SUN", "Lizhen Cui" ], "affiliation": "Shandong University", "summary": "We propose multimodal aligned semantic knowledge, which leverages word embeddings as bridges to associate words with prototypes, capturing semantic relationships between words and further utilizing information from OOD words.", "links": { "openreview": "https://openreview.net/forum?id=d3CISVVO6v", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-148", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "DiffusionNFT: Online Diffusion Reinforcement with Forward Process", "authors": [ "Kaiwen Zheng", "Huayu Chen", "Haotian Ye", "Haoxiang Wang", "Qinsheng Zhang", "Kai Jiang", "Hang Su", "Stefano Ermon", "Jun Zhu", "Ming-Yu Liu" ], "affiliation": "Tsinghua University", "summary": "We propose a new online reinforcement learning (RL) algorithm for diffusion and flow models based on forward process.", "links": { "openreview": "https://openreview.net/forum?id=VJZ477R89F", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-149", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "The Spacetime of Diffusion Models: An Information Geometry Perspective", "authors": [ "Rafal Karczewski", "Markus Heinonen", "Alison Pouplin", "Søren Hauberg", "Vikas K Garg" ], "affiliation": "Aalto University", "summary": "We present a novel geometric perspective on the latent space of diffusion models. We first show that the standard pullback approach, utilizing the deterministic probability flow ODE decoder, is fundamentally flawed.", "links": { "openreview": "https://openreview.net/forum?id=qCsbYJZRA5", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-150", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Reducing Belief Deviation in Reinforcement Learning for Active Reasoning of LLM Agents", "authors": [ "Deyu Zou", "Yongqiang Chen", "Jianxiang Wang", "Garry YANG", "Mufei Li", "Qing Da", "James Cheng", "Pan Li", "Yu Gong" ], "affiliation": "Department of Computer Science and Engineering, The Chinese University of Hong Kong", "summary": "Active reasoning requires large language model (LLM) agents to interact with external sources and strategically gather information to solve problems in multiple turns. Central to this process is belief tracking: maintaining an accurate representation of the underlying state and uncertainty in understanding and solving the problem.", "links": { "openreview": "https://openreview.net/forum?id=r8hzDA3pUY", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-151", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Information Shapes Koopman Representation", "authors": [ "Xiaoyuan Cheng", "Wenxuan Yuan", "Yiming Yang", "Yuanzhao Zhang", "Sibo Cheng", "Yi He", "Zhuo Sun" ], "affiliation": "University College London, University of London", "summary": "Because the Koopman operator is infinite-dimensional, identifying tractable finite-dimensional subspaces is challenging. We aim to construct these subspaces through information theory.", "links": { "openreview": "https://openreview.net/forum?id=Szh0ELyQxL", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-152", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Sequences of Logits Reveal the Low Rank Structure of Language Models", "authors": [ "Noah Golowich", "Allen Liu", "Abhishek Shetty" ], "affiliation": "Microsoft", "summary": "We exploit the low-rank structure of the logit matrices of LLMs to draw new empirical and theoretical conclusions.", "links": { "openreview": "https://openreview.net/forum?id=gdZ6J5hZzF", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-153", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "BIRD-INTERACT: Re-imagining Text-to-SQL Evaluation via Lens of Dynamic Interactions", "authors": [ "Nan Huo", "Xiaohan Xu", "Jinyang Li", "Per Jacobsson", "Shipei Lin", "Bowen Qin", "Binyuan Hui", "Xiaolong Li", "Ge Qu", "Shuzheng Si", "Linheng Han", "Edward Alexander", "Xintong Zhu", "Rui Qin", "Ruihan Yu", "Yiyao Jin", "Feige Zhou", "Weihao Zhong", "Yun Chen", "Hongyu Liu", "Chenhao Ma", "Fatma Ozcan", "Yannis Papakonstantinou", "Reynold Cheng" ], "affiliation": "the University of Hong Kong, University of Hong Kong", "summary": "Large language models (LLMs) have demonstrated remarkable performance on single-turn text-to-SQL tasks, but real-world database applications predominantly require multi-turn interactions to handle ambiguous queries, execution errors, and evolving user requirements. Existing multi-turn benchmarks fall short of capturing this complexity, either by...", "links": { "openreview": "https://openreview.net/forum?id=nHrYBGujps", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-154", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "EditVerse: Unifying Image and Video Editing and Generation with In-Context Learning", "authors": [ "Xuan Ju", "Tianyu Wang", "Yuqian Zhou", "He Zhang", "Qing Liu", "Nanxuan Zhao", "Zhifei Zhang", "Yijun Li", "Yuanhao Cai", "Shaoteng Liu", "Daniil Pakhomov", "Zhe Lin", "Soo Ye Kim", "Qiang Xu" ], "affiliation": "Chinese University of Hong Kong", "summary": "Recent advances in foundation models highlight a clear trend toward unification and scaling, showing emergent capabilities across diverse domains. While image generation and editing have rapidly transitioned from task-specific to unified frameworks, video generation and editing remain fragmented due to architectural limitations and data scarcity.", "links": { "openreview": "https://openreview.net/forum?id=blJXE07r7I", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-155", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "The Art of Scaling Reinforcement Learning Compute for LLMs", "authors": [ "Fnu Devvrit", "Lovish Madaan", "Rishabh Tiwari", "Rachit Bansal", "Sai Surya Duvvuri", "Manzil Zaheer", "Inderjit S Dhillon", "David Brandfonbrener", "Rishabh Agarwal" ], "affiliation": ", University of Texas at Austin", "summary": "We study compute scaling properties of RL methods on LLMs", "links": { "openreview": "https://openreview.net/forum?id=FMjeC9Msws", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-156", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Partition Generative Modeling: Masked Modeling Without Masks", "authors": [ "Justin Deschenaux", "Lan Tran", "Caglar Gulcehre" ], "affiliation": "EPFL - EPF Lausanne", "summary": "We show that it is possible to train masked generative models without using MASK tokens, resulting in efficiency gains at inference.", "links": { "openreview": "https://openreview.net/forum?id=vEh1ceS154", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-157", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability", "authors": [ "Shawn Im", "Changdae Oh", "Zhen Fang", "Sharon Li" ], "affiliation": "Department of Computer Science, University of Wisconsin - Madison", "summary": "Semantic associations such as the link between \"bird\" and \"flew\" are foundational for language modeling as they enable models to go beyond memorization and instead generalize and generate coherent text. Understanding how these associations are learned and represented in language models is essential for connecting deep learning with linguistic th...", "links": { "openreview": "https://openreview.net/forum?id=A4Us8jxVGq", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-158", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "On the Reasoning Abilities of Masked Diffusion Language Models", "authors": [ "Anej Svete", "Ashish Sabharwal" ], "affiliation": "Department of Computer Science, ETHZ - ETH Zurich", "summary": "We prove that masked text diffusion models are equivalent to padded looped transformers, can solve all problems that chain-of-thought transformers can, and are more efficient on certain problem classes due to their parallel generation mechanism.", "links": { "openreview": "https://openreview.net/forum?id=BVnIsh4Nz1", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-159", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "FRABench and UFEval: Unified Fine-grained Evaluation with Task and Aspect Generalization", "authors": [ "Shibo Hong", "Jiahao Ying", "Haiyuan Liang", "Mengdi Zhang", "Jun Kuang", "Jiazheng Zhang", "Yixin Cao" ], "affiliation": "Fudan University", "summary": "Evaluating open-ended outputs of Multimodal Large Language Models has become a bottleneck as model capabilities, task diversity, and modality rapidly expand. Existing ``MLLM-as-a-Judge'' evaluators, though promising, remain constrained to specific tasks and aspects (i.e., specific evaluation criteria such as fluency for text and image quality fo...", "links": { "openreview": "https://openreview.net/forum?id=7WdY3Cojy9", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-160", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Quotient-Space Diffusion Models", "authors": [ "Yixian Xu", "Yusong Wang", "Shengjie Luo", "Kaiyuan Gao", "Tianyu He", "Di He", "Chang Liu" ], "affiliation": "Peking University", "summary": "We propose a principled way to leverage group symmetry of the target distribution by defining a diffusion model on the quotient space, which achieves both easier learning and correct sampling for the first time.", "links": { "openreview": "https://openreview.net/forum?id=3JPAkwSVc4", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-161", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "SWINGARENA: Adversarial Programming Arena for Long-context GitHub Issue Solving", "authors": [ "Wendong XU", "Jing Xiong", "Chenyang Zhao", "Qiujiang Chen", "Haoran Wang", "Hui Shen", "Zhongwei Wan", "Jianbo Dai", "Taiqiang Wu", "He Xiao", "Chaofan Tao", "Zhuoqing Mao", "Ying Sheng", "Zhijiang Guo", "Hongxia Yang", "Bei Yu", "Lingpeng Kong", "Quanquan Gu", "Ngai Wong" ], "affiliation": "University of Hong Kong", "summary": "We present \\textsc{SwingArena}, a adversarial evaluation framework for Large Language Models (LLMs) that closely mirrors real-world software development workflows. Unlike traditional static benchmarks, \\textsc{SwingArena} models the collaborative process of software iteration by pairing LLMs as \\textit{submitters}, who generate patches, and \\tex...", "links": { "openreview": "https://openreview.net/forum?id=YuxgSGFaqb", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-162", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "AnyUp: Universal Feature Upsampling", "authors": [ "Thomas Wimmer", "Prune Truong", "Marie-Julie Rakotosaona", "Michael Oechsle", "Federico Tombari", "Bernt Schiele", "Jan Eric Lenssen" ], "affiliation": "Max-Planck Institute", "summary": "A universal feature upsampling model that can be used to upsample any feature from any to any resolution and generalizes to features unseen during training.", "links": { "openreview": "https://openreview.net/forum?id=Y9UAgPehqo", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-163", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "High-dimensional Analysis of Synthetic Data Selection", "authors": [ "Parham Rezaei", "Filip Kovačević", "Francesco Locatello", "Marco Mondelli" ], "affiliation": "Institute of Science and Technology Austria", "summary": "We give a precise analysis for the problem of synthetic data selection through the lens of high-dimensional regression, and we translate the theoretical insights into a method that performs well in practice.", "links": { "openreview": "https://openreview.net/forum?id=Y54P2BBPPh", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-164", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Task-free Adaptive Meta Black-box Optimization", "authors": [ "Chao Wang", "Licheng Jiao", "Lingling Li", "Jiaxuan Zhao", "Guanchun Wang", "Fang Liu", "Shuyuan Yang" ], "affiliation": "Xidian University", "summary": "Handcrafted optimizers become prohibitively inefficient for complex black-box optimization (BBO) tasks. MetaBBO addresses this challenge by meta-learning to automatically configure optimizers for low-level BBO tasks, thereby eliminating heuristic dependencies.", "links": { "openreview": "https://openreview.net/forum?id=AufVSUgMUo", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-165", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Token-Importance Guided Direct Preference Optimization", "authors": [ "Ning Yang", "Hai Lin", "Yibo Liu", "Baoliang Tian", "Guoqing Liu", "Haijun Zhang" ], "affiliation": "Institute of automation, Chinese academy of science, Chinese Academy of Sciences", "summary": "We proposes Token-Importance Guided Direct Preference Optimization (TI-DPO) to better align LLMs with human preferences by using a hybrid weighting mechanism to identify key tokens and a triplet loss to guide the optimization process.", "links": { "openreview": "https://openreview.net/forum?id=cMEnMVvMw9", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-166", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "DTO-KD: Dynamic Trade-off Optimization for Effective Knowledge Distillation", "authors": [ "Zeeshan Hayder", "Ali Cheraghian", "Lars Petersson", "Mehrtash Harandi", "Richard Hartley" ], "affiliation": "Google", "summary": "DTO-KD", "links": { "openreview": "https://openreview.net/forum?id=QMItTyQW92", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-167", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Distributional Equivalence in Linear Non-Gaussian Latent-Variable Cyclic Causal Models: Characterization and Learning", "authors": [ "Haoyue Dai", "Immanuel Albrecht", "Peter Spirtes", "Kun Zhang" ], "affiliation": "Carnegie Mellon University", "summary": "Causal discovery with latent variables is a fundamental task. Yet most existing methods rely on strong structural assumptions, such as enforcing specific indicator patterns for latents or restricting how they can interact with others.", "links": { "openreview": "https://openreview.net/forum?id=b8TlYh6PN6", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-168", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Reliable Weak-to-Strong Monitoring of LLM Agents", "authors": [ "Neil Kale", "Chen Bo Calvin Zhang", "Kevin Zhu", "Ankit Aich", "Paula Rodriguez", "Christina Q Knight", "Zifan Wang" ], "affiliation": "School of Computer Science, Carnegie Mellon University", "summary": "This paper introduces a monitor red teaming workflow to stress test systems for detecting covert misbehavior in LLM agents, finding that a well-designed monitor scaffold enables weaker models to oversee strong aware attackers.", "links": { "openreview": "https://openreview.net/forum?id=WV7xIboTDK", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-169", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Multiplayer Nash Preference Optimization", "authors": [ "Fang Wu", "Xu Huang", "Weihao Xuan", "Zhiwei Zhang", "Yijia Xiao", "Guancheng Wan", "Xiaomin Li", "Bing Hu", "Peng Xia", "Jure Leskovec", "Yejin Choi" ], "affiliation": "", "summary": "Reinforcement learning from human feedback (RLHF) has emerged as the standard paradigm for aligning large language models with human preferences. However, reward-based methods grounded in the Bradley–Terry assumption struggle to capture the nontransitivity and heterogeneity of real-world preferences.", "links": { "openreview": "https://openreview.net/forum?id=x7aLhLMVn1", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-170", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Multi-Domain Riemannian Graph Gluing for Building Graph Foundation Models", "authors": [ "Li Sun", "Zhenhao Huang", "Silei Chen", "Lanxu Yang", "Junda Ye", "Sen Su", "Philip S. Yu" ], "affiliation": "Beijing University of Post and Telecommunications", "summary": "From differential geometry perspective, we present a novel framework that merges multi-domain graphs into a unified, smooth manifold with geometric consistency, enabling quantifiable transferability and geometric scaling behavior.", "links": { "openreview": "https://openreview.net/forum?id=G3uNHQpP7J", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-171", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "TROLL: Trust Regions Improve Reinforcement Learning for Large Language Models", "authors": [ "Philipp Becker", "Niklas Freymuth", "Serge Thilges", "Fabian Otto", "Gerhard Neumann" ], "affiliation": "Facebook", "summary": "Replacing PPO's clipping objective with more principled trust regions improves RL from verifiable rewards.", "links": { "openreview": "https://openreview.net/forum?id=X9D5MVpPJ9", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-172", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "FlashVID: Efficient Video Large Language Models via Training-free Tree-based Spatiotemporal Token Merging", "authors": [ "Ziyang Fan", "Keyu Chen", "Ruilong Xing", "Yulin Li", "Li Jiang", "Zhuotao Tian" ], "affiliation": "Harbin Institute of Technology, Shenzhen", "summary": "We introduce FlashVID, a training-free and plug-and-play inference acceleration framework for Video LLMs, enabling a satisfactory speedup with negligible performance degradation.", "links": { "openreview": "https://openreview.net/forum?id=H6rDX4w6Al", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-173", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "WSM: Decay-Free Learning Rate Schedule via Checkpoint Merging for LLM Pre-training", "authors": [ "Changxin Tian", "jiapeng wang", "Qian Zhao", "Kunlong Chen", "Jia Liu", "Ziqi Liu", "Jiaxin Mao", "Xin Zhao", "Zhiqiang Zhang", "JUN ZHOU" ], "affiliation": "Ant Group", "summary": "Recent advances in learning rate~(LR) scheduling have demonstrated the effectiveness of decay-free approaches that eliminate the traditional decay phase while maintaining competitive performance. Model merging techniques have emerged as particularly promising solutions in this domain.", "links": { "openreview": "https://openreview.net/forum?id=HhThhjKyfw", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-174", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MrRoPE: Mixed-radix Rotary Position Embedding", "authors": [ "Qingyuan Tian", "Wenhong Zhu", "Xiaoran Liu", "Xiaofeng Wang", "Rui Wang" ], "affiliation": "Shanghai Jiaotong University", "summary": "We present a unified theory MrRoPE linking major RoPE-extension methods to radix conversion. Based on this, we propose MrRoPE-Pro, a training-free context window extension method..", "links": { "openreview": "https://openreview.net/forum?id=1J63FJYJKg", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-175", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "OpenThoughts: Data Recipes for Reasoning Models", "authors": [ "Etash Kumar Guha", "Ryan Marten", "Sedrick Keh", "Negin Raoof", "Georgios Smyrnis", "Hritik Bansal", "Marianna Nezhurina", "Jean Mercat", "Trung Vu", "Zayne Rea Sprague", "Ashima Suvarna", "Benjamin Feuer", "Leon Liangyu Chen", "Zaid Khan", "Eric Frankel", "Sachin Grover", "Caroline Choi", "Niklas Muennighoff", "Shiye Su", "Wanjia Zhao", "John Yang", "Shreyas Pimpalgaonkar", "Kartik sharma", "Charlie Cheng-Jie Ji", "Yichuan Deng", "Sarah M Pratt", "Vivek Ramanujan", "Jon Saad-Falcon", "Stutee Acharya", "Jeffrey Li", "Achal Dave", "Alon Albalak", "Kushal Arora", "Blake Wulfe", "Chinmay Hegde", "Greg Durrett", "Sewoong Oh", "Mohit Bansal", "Saadia Gabriel", "Aditya Grover", "Kai-Wei Chang", "Vaishaal Shankar", "Aaron Gokaslan", "Mike A Merrill", "Tatsunori Hashimoto", "Yejin Choi", "Jenia Jitsev", "Reinhard Heckel", "Maheswaran Sathiamoorthy", "Alex Dimakis", "Ludwig Schmidt" ], "affiliation": "RIKEN", "summary": "Data pipeline analysis for training reasoning models", "links": { "openreview": "https://openreview.net/forum?id=7xjoTuaNmN", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-176", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Text-to-3D by Stitching a Multi-view Reconstruction Network to a Video Generator", "authors": [ "Hyojun Go", "Dominik Narnhofer", "Goutam Bhat", "Prune Truong", "Federico Tombari", "Konrad Schindler" ], "affiliation": "ETHZ - ETH Zurich", "summary": "Text-to-3D scene generative modelling by unifying a video generative model with a foundational 3D model via model stitching and alignment.", "links": { "openreview": "https://openreview.net/forum?id=kI27Niy4xY", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-177", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "A Scalable Distributed Framework for Multimodal GigaVoxel Image Registration", "authors": [ "Rohit Jena", "Vedant Zope", "Pratik Chaudhari", "James Gee" ], "affiliation": "University of Pennsylvania", "summary": "we propose non-GEMM CUDA kernels and distributed primitives to scale multimodal image registration to arbitrary image sizes", "links": { "openreview": "https://openreview.net/forum?id=8dLexnao2h", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-178", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Invisible Safety Threat: Malicious Finetuning for LLM via Steganography", "authors": [ "Guangnian Wan", "Xinyin Ma", "Gongfan Fang", "Xinchao Wang" ], "affiliation": "National University of Singapore", "summary": "We highlight an insidious safety threat: a compromised LLM can maintain a facade of proper safety alignment while covertly generating harmful content through steganography.", "links": { "openreview": "https://openreview.net/forum?id=6cEPDGaShH", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-179", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Temporal superposition and feature geometry of RNNs under memory demands", "authors": [ "Pratyaksh Sharma", "Alexandra Maria Proca", "Lucas Prieto", "Pedro A. M. Mediano" ], "affiliation": "G-Research", "summary": "We study the feature geometry of RNNs under memory demands and characterize their representational strategies using a novel framework of temporal superposition.", "links": { "openreview": "https://openreview.net/forum?id=7cMzTpbJHC", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-180", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "InfoTok: Adaptive Discrete Video Tokenizer via Information-Theoretic Compression", "authors": [ "Haotian Ye", "Qiyuan He", "Jiaqi Han", "Puheng Li", "Jiaojiao Fan", "Zekun Hao", "Fitsum Reda", "Yogesh Balaji", "Huayu Chen", "Sheng Liu", "Angela Yao", "James Zou", "Stefano Ermon", "Haoxiang Wang", "Ming-Yu Liu" ], "affiliation": "Stanford University", "summary": "This paper introduces InfoTok, an adaptive video tokenizer guided by information theory, which significantly boosts video compression efficiency and reduces computational overhead without degrading visual quality.", "links": { "openreview": "https://openreview.net/forum?id=JEYWpFGzvn", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-181", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Actions Speak Louder than Prompts: A Large-Scale Study of LLMs for Graph Inference", "authors": [ "Ben Finkelshtein", "Silviu Cucerzan", "Sujay Kumar Jauhar", "Ryen W White" ], "affiliation": "University of Oxford", "summary": "A comprehensive study of LLMs for node classification, providing a principled understanding of their capabilities in processing graph information that practitioners can apply in real-world tasks", "links": { "openreview": "https://openreview.net/forum?id=MgJUj9Sk3C", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-182", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs", "authors": [ "Zhiyu Pan", "Yizheng Wu", "Jiashen Hua", "Junyi Feng", "Shaotian Yan", "Bing Deng", "Zhiguo Cao", "Jieping Ye" ], "affiliation": "Alibaba Group", "summary": "Reasoning has emerged as a key capability of large language models. In linguistic tasks, this capability can be enhanced by self-improving techniques that refine reasoning paths for subsequent fine-tuning.", "links": { "openreview": "https://openreview.net/forum?id=ZymCPON45y", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-183", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "SAFETY-GUIDED FLOW (SGF): A UNIFIED FRAMEWORK FOR NEGATIVE GUIDANCE IN SAFE GENERATION", "authors": [ "Mingyu Kim", "Young-Heon Kim", "Mijung Park" ], "affiliation": "Kookmin University", "summary": "We introduced a unified probabilistic framework for safe generation in diffusion and flow models, using Maximum Mean Discrepancy-based energy potentials.", "links": { "openreview": "https://openreview.net/forum?id=EA80Zib9UI", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-184", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Depth Anything 3: Recovering the Visual Space from Any Views", "authors": [ "Haotong Lin", "Sili Chen", "Jun Hao Liew", "Donny Y. Chen", "Zhenyu Li", "Yang Zhao", "Sida Peng", "Hengkai Guo", "Xiaowei Zhou", "Guang Shi", "Jiashi Feng", "Bingyi Kang" ], "affiliation": "Zhejiang University", "summary": "Depth Anything 3 uses a single vanilla DINOv2 transformer to take arbitrary input views and outputs consistent depth and ray maps, delivering leading pose, geometry, and visual rendering performance.", "links": { "openreview": "https://openreview.net/forum?id=yirunib8l8", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-185", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Taming Momentum: Rethinking Optimizer States Through Low-Rank Approximation", "authors": [ "Zhengbo Wang", "Jian Liang", "Ran He", "Zilei Wang", "Tieniu Tan" ], "affiliation": "University of Science and Technology of China", "summary": "Modern optimizers like Adam and Muon are central to training large language models, but their reliance on first- and second-order momenta introduces significant memory overhead, which constrains scalability and computational efficiency. In this work, we reframe the exponential moving average (EMA) used in these momenta as the training of a linea...", "links": { "openreview": "https://openreview.net/forum?id=9Q0dNBYeEY", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-186", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Premise Selection for a Lean Hammer", "authors": [ "Thomas Zhu", "Joshua Clune", "Jeremy Avigad", "Albert Q. Jiang", "Sean Welleck" ], "affiliation": "ByteDance Inc.", "summary": "LeanHammer integrates neural premise selection with symbolic reasoning to automate theorem proving in Lean.", "links": { "openreview": "https://openreview.net/forum?id=m04JJNeRK6", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-187", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "DepthLM: Metric Depth from Vision Language Models", "authors": [ "Zhipeng Cai", "Ching-Feng Yeh", "Hu Xu", "Zhuang Liu", "Gregory P. Meyer", "Xinjie Lei", "Changsheng Zhao", "Shang-Wen Li", "Vikas Chandra", "Yangyang Shi" ], "affiliation": "Meta", "summary": "The first proof that VLMs can have expert model level depth estimation accuracy without architecture or loss change", "links": { "openreview": "https://openreview.net/forum?id=ObFVZGnSFN", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-188", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Hallucination Begins Where Saliency Drops", "authors": [ "Xiaofeng Zhang", "Yuanchao Zhu", "Chaochen Gu", "Xiaosong Yuan", "Qiyan Zhao", "Jiawei Cao", "Feilong Tang", "Sinan Fan", "Yaomin Shen", "Chen Shen", "Hao Tang" ], "affiliation": "Shanghai Jiao Tong University", "summary": "Recent studies have investigated attention dynamics in large vision language models (LVLMs), yet existing methods remain limited in reliably distinguishing hallucinated from correct outputs — primarily because they rely solely on forward-pass attention, ignoring gradient-based signals that reveal how token influence propagates through the model....", "links": { "openreview": "https://openreview.net/forum?id=sjnErRHXf3", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-189", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Overthinking Reduction with Decoupled Rewards and Curriculum Data Scheduling", "authors": [ "Shuyang Jiang", "Yusheng Liao", "Ya Zhang", "Yanfeng Wang", "Yu Wang" ], "affiliation": "Fudan University", "summary": "theoretical unveil the underlying limitations of length reward and propose D$^2$yOR to achieve supreme efficiency without performance degradation", "links": { "openreview": "https://openreview.net/forum?id=kdeiRledV6", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-190", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "On the Generalization Capacities of MLLMs for Spatial Intelligence", "authors": [ "Gongjie Zhang", "Wenhao Li", "Quanhao Qian", "Jiuniu Wang", "Deli Zhao", "Shijian Lu", "Ran Xu" ], "affiliation": "Alibaba Group", "summary": "We show that RGB-only MLLMs are fundamentally flawed for spatial reasoning due to an inherent geometric ambiguity, and propose a camera-aware MLLM framework that incorporates camera intrinsics for robust, generalizable spatial intelligence.", "links": { "openreview": "https://openreview.net/forum?id=DE5ZJtR4bg", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-191", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "The Polar Express: Optimal Matrix Sign Methods and their Application to the Muon Algorithm", "authors": [ "Noah Amsel", "David Persson", "Christopher Musco", "Robert M. Gower" ], "affiliation": "NYU, New York University", "summary": "We introduce a GPU-friendly algorithm for computing the polar decomposition of a matrix to low accuracy that is optimal in its class. This improves Muon.", "links": { "openreview": "https://openreview.net/forum?id=yRtgZ1K8hO", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-192", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Characterizing the Discrete Geometry of ReLU Networks", "authors": [ "Blake B. Gaines", "Jinbo Bi" ], "affiliation": "University of Connecticut", "summary": "We describe the geometry of the polyhedral complexes defined by the linear regions of ReLU networks, both by theoretically bounding their connectivity and diameter and by empirically characterizing it using experiments on trained networks.", "links": { "openreview": "https://openreview.net/forum?id=TgLW2DiRDG", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-193", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "ThinKV: Thought-Adaptive KV Cache Compression for Efficient Reasoning Models", "authors": [ "Akshat Ramachandran", "Marina Neseem", "Charbel Sakr", "Rangharajan Venkatesan", "Brucek Khailany", "Tushar Krishna" ], "affiliation": "Georgia Institute of Technology", "summary": "The long-output context generation of large reasoning models enables extended chain of thought (CoT) but also drives rapid growth of the key–value (KV) cache, quickly overwhelming GPU memory. To address this challenge, we propose ThinKV, a thought-adaptive KV cache compression framework.", "links": { "openreview": "https://openreview.net/forum?id=M3CeHnZKNC", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-194", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Difficult Examples Hurt Unsupervised Contrastive Learning: A Theoretical Perspective", "authors": [ "Yi-Ge Zhang", "Jingyi Cui", "Qiran Li", "Yisen Wang" ], "affiliation": "Peking University", "summary": "We introduce a similarity-based theoretical framework that shows how difficult boundary examples impair generalization in unsupervised contrastive learning, and we design mechanisms that address this issue and boost downstream accuracy.", "links": { "openreview": "https://openreview.net/forum?id=5LMdnUdAoy", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-195", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers", "authors": [ "Shikang Zheng", "Guantao Chen", "Qinming Zhou", "Yuqi Lin", "Lixuan He", "Chang Zou", "Peiliang Cai", "Jiacheng Liu", "Linfeng Zhang" ], "affiliation": "", "summary": "Diffusion Transformers offer state-of-the-art fidelity in image and video synthesis, but their iterative sampling process remains a major bottleneck due to the high cost of transformer forward passes at each timestep. To mitigate this, feature caching has emerged as a training-free acceleration technique that reuses hidden representations.", "links": { "openreview": "https://openreview.net/forum?id=URbsHlTK8c", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-196", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MotionStream: Real-Time Video Generation with Interactive Motion Controls", "authors": [ "Joonghyuk Shin", "Zhengqi Li", "Richard Zhang", "Jun-Yan Zhu", "Jaesik Park", "Eli Shechtman", "Xun Huang" ], "affiliation": "Seoul National University", "summary": "We present MotionStream, a streaming (real-time, infinite length) video generation system with motion controls, unlocking new possibilities for interactive content creation.", "links": { "openreview": "https://openreview.net/forum?id=v1DKz5Vxr7", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-197", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "BioX-Bridge: Model Bridging for Unsupervised Cross-Modal Knowledge Transfer across Biosignals", "authors": [ "Chenqi Li", "Yu Liu", "Timothy Denison", "Tingting Zhu" ], "affiliation": "University of Oxford", "summary": "Biosignals offer valuable insights into the physiological states of the human body. Although biosignal modalities differ in functionality, signal fidelity, sensor comfort, and cost, they are often intercorrelated, reflecting the holistic and interconnected nature of human physiology.", "links": { "openreview": "https://openreview.net/forum?id=1448q0s3zZ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-198", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Compactness and Consistency: A Conjoint Framework for Deep Graph Clustering", "authors": [ "Wei Ju", "Siyu Yi", "Kangjie Zheng", "Yifan Wang", "Ziyue Qiao", "Li Shen", "Yongdao Zhou", "Xiaochun Cao", "Jiancheng Lv" ], "affiliation": "", "summary": "Graph clustering is a fundamental task in data analysis, aiming at grouping nodes with similar characteristics in the graph into clusters. This problem has been widely explored using graph neural networks (GNNs) due to their ability to leverage node attributes and graph topology for effective cluster assignments.", "links": { "openreview": "https://openreview.net/forum?id=9jdQLmPUHW", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-199", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "WAFT: Warping-Alone Field Transforms for Optical Flow", "authors": [ "Yihan Wang", "Jia Deng" ], "affiliation": "Princeton University", "summary": "We introduce Warping-Alone Field Transforms (WAFT), a simple and effective method for optical flow. WAFT is similar to RAFT but replaces cost volume with high-resolution warping, achieving better accuracy with lower memory cost.", "links": { "openreview": "https://openreview.net/forum?id=HTqGE0KcuF", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-200", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "FlashWorld: High-quality 3D Scene Generation within Seconds", "authors": [ "Xinyang Li", "Tengfei Wang", "Zixiao Gu", "Shengchuan Zhang", "Chunchao Guo", "Liujuan Cao" ], "affiliation": "Xiamen University", "summary": "a generative model that produces high-quality 3D scenes from a single image or text prompt in seconds", "links": { "openreview": "https://openreview.net/forum?id=2IftRjRB07", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-201", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "TabStruct: Measuring Structural Fidelity of Tabular Data", "authors": [ "Xiangjian Jiang", "Nikola Simidjievski", "Mateja Jamnik" ], "affiliation": "University of Cambridge", "summary": "We propose TabStruct, a comprehensive benchmark, along with a novel metric, global utility, for evaluating the structural fidelity of tabular data without requiring access to ground-truth causal structures.", "links": { "openreview": "https://openreview.net/forum?id=XOPH34Extq", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-202", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Uncover Underlying Correspondence for Robust Multi-view Clustering", "authors": [ "Haochen Zhou", "Guofeng Ding", "Mouxing Yang", "Peng Hu", "Yijie Lin", "Xi Peng" ], "affiliation": "Sichuan University", "summary": "Multi-view clustering (MVC) aims to group unlabeled data into semantically meaningful clusters by leveraging cross-view consistency. However, real-world datasets collected from the web often suffer from noisy correspondence (NC), which breaks the consistency prior and results in unreliable alignments.", "links": { "openreview": "https://openreview.net/forum?id=a4S1nQay3b", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-203", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Instilling an Active Mind in Avatars via Cognitive Simulation", "authors": [ "Jianwen Jiang", "Weihong Zeng", "Zerong Zheng", "Jiaqi Yang", "Chao Liang", "Wang Liao", "Han Liang", "Weifeng Chen", "XING WANG", "Yuan Zhang", "Mingyuan Gao" ], "affiliation": "", "summary": "This paper introduces a novel framework that uses a Large Language Model (LLM) for semantic guidance and a Multimodal Diffusion Transformer (DiT) for fusion to generate expressive, context-aware video avatars, demonstrating competitive performance", "links": { "openreview": "https://openreview.net/forum?id=80JylHgQn1", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-204", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Diffusion Language Model Knows the Answer Before It Decodes", "authors": [ "Pengxiang Li", "Yefan Zhou", "Dilxat Muhtar", "Lu Yin", "Shilin Yan", "Li Shen", "Yi Liang", "Soroush Vosoughi", "Shiwei Liu" ], "affiliation": "Hong Kong Polytechnic University", "summary": "Diffusion language models (DLMs) have recently emerged as an alternative to autoregressive approaches, offering parallel sequence generation and flexible token orders. However, their inference remains slower than that of autoregressive models, primarily due to the cost of bidirectional attention and the large number of refinement steps required...", "links": { "openreview": "https://openreview.net/forum?id=g88nt4ieTG", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-205", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Generative Human Geometry Distribution", "authors": [ "Xiangjun Tang", "Biao Zhang", "Peter Wonka" ], "affiliation": "King Abdullah University of Science and Technology", "summary": "We introduce the first method that integrates geometry distributions into generative modeling.", "links": { "openreview": "https://openreview.net/forum?id=YsQM7sQl0j", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-206", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "EmotionThinker: Prosody-Aware Reinforcement Learning for Explainable Speech Emotion Reasoning", "authors": [ "Dingdong WANG", "Shujie LIU", "Tianhua Zhang", "Youjun Chen", "Jinyu Li", "Helen M. Meng" ], "affiliation": "Chinese University of Hong Kong, The Chinese University of Hong Kong", "summary": "Emotional information in speech plays a unique role in multimodal perception. However, current Speech Large Language Models (SpeechLLMs), similar to conventional speech emotion recognition (SER) systems, still treat emotion understanding as a simple classification problem.", "links": { "openreview": "https://openreview.net/forum?id=wbttgzp7MT", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-207", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Mastering Sparse CUDA Generation through Pretrained Models and Deep Reinforcement Learning", "authors": [ "Yaoyu Wang", "Hankun Dai", "Zhidong Yang", "Junmin Xiao", "Guangming Tan" ], "affiliation": "University of Chinese Academy of Sciences", "summary": "We propose SparseRL, a deep reinforcement learning framework that generates high-performance CUDA code for sparse matrix operations, achieving significant improvements in both correctness and execution efficiency.", "links": { "openreview": "https://openreview.net/forum?id=VdLEaGPYWT", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-208", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "LongWriter-Zero: Mastering Ultra-Long Text Generation via Reinforcement Learning", "authors": [ "Yuhao Wu", "Yushi Bai", "Zhiqiang Hu", "Roy Ka-Wei Lee", "Juanzi Li" ], "affiliation": "Singapore University of Technology and Design", "summary": "Ultra-long generation by large language models (LLMs) is a widely demanded scenario, yet it remains a significant challenge due to their maximum generation length limit and overall quality degradation as sequence length increases. Previous approaches, exemplified by LongWriter, typically rely on ''teaching'', which involves supervised fine-tunin...", "links": { "openreview": "https://openreview.net/forum?id=JWx4DI2N8k", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-209", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "RealPDEBench: A Benchmark for Complex Physical Systems with Real-World Data", "authors": [ "Peiyan Hu", "Haodong Feng", "Hongyuan Liu", "Tongtong Yan", "Wenhao Deng", "Tianrun Gao", "Rong Zheng", "Haoren Zheng", "Chenglei Yu", "Chuanrui Wang", "Kaiwen Li", "Zhi-Ming Ma", "Dezhi Zhou", "Xingcai Lu", "Dixia Fan", "Tailin Wu" ], "affiliation": "Chinese Academy of Sciences", "summary": "We propose the first benchmark for complex physical systems with paired real-world data and simulated data, and explore how to bridge simulated and real-world data.", "links": { "openreview": "https://openreview.net/forum?id=y3oHMcoItR", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-210", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Improving Diffusion Models for Class-imbalanced Training Data via Capacity Manipulation", "authors": [ "Feng Hong", "Jiangchao Yao", "Yifei Shen", "Dongsheng Li", "Ya Zhang", "Yanfeng Wang" ], "affiliation": "Shanghai Jiao Tong University", "summary": "While diffusion models have achieved remarkable performance in image generation, they often struggle with the imbalanced datasets frequently encountered in real-world applications, resulting in significant performance degradation on minority classes. In this paper, we identify model capacity allocation as a key and previously underexplored facto...", "links": { "openreview": "https://openreview.net/forum?id=wSGle6ag5I", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-211", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer", "authors": [ "Junsong Chen", "Yuyang Zhao", "Jincheng YU", "Ruihang Chu", "Junyu Chen", "Shuai Yang", "Xianbang Wang", "Yicheng Pan", "Daquan Zhou", "Huan Ling", "Haozhe Liu", "Hongwei Yi", "Hao Zhang", "Muyang Li", "Yukang Chen", "Han Cai", "Sanja Fidler", "Ping Luo", "Song Han", "Enze Xie" ], "affiliation": "University of Hong Kong", "summary": "We introduce SANA-Video, a small diffusion model that can efficiently generate videos up to 720×1280 resolution and minute-length duration. SANA-Video synthesizes high-resolution, high-quality and long videos with strong text-video alignment at a remarkably fast speed, deployable on RTX 5090 GPU.", "links": { "openreview": "https://openreview.net/forum?id=mzAchylAtf", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-212", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Stable Video Infinity: Infinite-Length Video Generation with Error Recycling", "authors": [ "Wuyang Li", "Wentao Pan", "Po-Chien Luan", "Yang Gao", "Alexandre Alahi" ], "affiliation": "EPFL - EPF Lausanne", "summary": "We propose **Stable Video Infinity (SVI)** that can generate non-looping, ultra-long videos with stable visual quality, while supporting per-clip prompt control and multi-modal conditioning. While existing long-video methods attempt to _**mitigate accumulated errors**_ via handcrafted anti-drifting (e.g., modified noise scheduler, frame anchorin...", "links": { "openreview": "https://openreview.net/forum?id=X96Ei9n34a", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-213", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Extending Sequence Length is Not All You Need: Effective Integration of Multimodal Signals for Gene Expression Prediction", "authors": [ "Zhao Yang", "Yi Duan", "Jiwei Zhu", "Ying Ba", "Chuan Cao", "Bing Su" ], "affiliation": "Renmin University of China", "summary": "Gene expression prediction, which predicts mRNA expression levels from DNA sequences, presents significant challenges. Previous works often focus on extending input sequence length to locate distal enhancers, which may influence target genes from hundreds of kilobases away.", "links": { "openreview": "https://openreview.net/forum?id=wwPSfcf5Pj", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-214", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "MomaGraph: State-Aware Unified Scene Graphs with Vision-Language Models for Embodied Task Planning", "authors": [ "Yuanchen Ju", "Yongyuan Liang", "Yen-Jen Wang", "Gireesh Nandiraju", "Yuanliang Ju", "Seungjae Lee", "Qiao Gu", "Elvis Hsieh", "Furong Huang", "Koushil Sreenath" ], "affiliation": "", "summary": "We present MomaGraph, a unified scene representation for task-oriented understanding, along with a dataset and benchmark built upon it, and MomaGraph-R1, a 7B model that constructs MomaGraph representations and generates task plans.", "links": { "openreview": "https://openreview.net/forum?id=3eTr9dGwJv", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-215", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training", "authors": [ "Junlin Han", "Shengbang Tong", "David Fan", "Yufan Ren", "Koustuv Sinha", "Philip Torr", "Filippos Kokkinos" ], "affiliation": "University of Oxford", "summary": "Explore and understand the visual priors within LLMs and thus build better MLLMs.", "links": { "openreview": "https://openreview.net/forum?id=pfw176o1YJ", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-216", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models", "authors": [ "Bartłomiej Marek", "Lorenzo Rossi", "Vincent Hanke", "Xun Wang", "Michael Backes", "Franziska Boenisch", "Adam Dziedzic" ], "affiliation": "CISPA Helmholtz Center for Information Security", "summary": "DP adaptations of LLMs can leak data in practice, with risk rising as adaptation data becomes closer to the pretraining distribution.", "links": { "openreview": "https://openreview.net/forum?id=jY7fAo9rfK", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-217", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "ScaleCUA: Scaling Open-Source Computer Use Agents with Cross-Platform Data", "authors": [ "Zhaoyang Liu", "JingJing Xie", "Zichen Ding", "Zehao Li", "Bowen Yang", "Zhenyu Wu", "Xuehui Wang", "Qiushi Sun", "Shi Liu", "Weiyun Wang", "Shenglong Ye", "Qingyun Li", "Zeyue Tian", "Gen Luo", "Xiangyu Yue", "Biqing Qi", "Kai Chen", "Bowen Zhou", "Yu Qiao", "Qifeng Chen", "Wenhai Wang" ], "affiliation": "Hong Kong University of Science and Technology", "summary": "Vision-Language Models (VLMs) have enabled computer use agents (CUAs) that operate GUIs autonomously, showing great potential, yet progress is limited by the lack of large-scale, open-source computer use data and foundation models. In this work, we introduce ScaleCUA, a step toward scaling open-source CUAs.", "links": { "openreview": "https://openreview.net/forum?id=yBFUqdJFZn", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-218", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Train-before-Test Harmonizes Language Model Rankings", "authors": [ "Guanhua Zhang", "Ricardo Dominguez-Olmedo", "Moritz Hardt" ], "affiliation": "Max Planck Institute for Intelligent Systems, Max-Planck Institute", "summary": "Existing language model benchmarks provide contradictory model rankings, even for benchmarks that aim to capture similar skills. This dilemma of conflicting rankings hampers model selection, clouds model comparisons, and adds confusion to a growing ecosystem of competing models.", "links": { "openreview": "https://openreview.net/forum?id=ORv3SAzus1", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-219", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "One for Two: A Unified Framework for Imbalanced Graph Classification via Dynamic Balanced Prototype", "authors": [ "Guanjun Wang", "Binwu Wang", "Jiaming Ma", "Zhengyang Zhou", "Pengkun Wang", "Xu Wang", "Yang Wang" ], "affiliation": "University of Science and Technology of China", "summary": "Graph Neural Networks (GNNs) have advanced graph classification, yet they remain vulnerable to graph-level imbalance, encompassing class imbalance and topological imbalance. To address both types of imbalance in a unified manner, we propose UniImb, a Unified framework for Imbalanced graph classification.", "links": { "openreview": "https://openreview.net/forum?id=MraQM41SNS", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-220", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning", "authors": [ "Hao Tan", "jun lan", "Zichang Tan", "Senyuan Shi", "Ajian Liu", "Chuanbiao Song", "Huijia Zhu", "Weiqiang Wang", "Jun Wan", "Zhen Lei" ], "affiliation": "Institute of automation, Chinese Academy of Sciences", "summary": "We introduce a MLLM-based detector for transparent deepfake detection, along with a holistic dataset for deepfake detection.", "links": { "openreview": "https://openreview.net/forum?id=5VXJPS1HoM", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-221", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Semi-Supervised Preference Optimization with Limited Feedback", "authors": [ "Seonggyun Lee", "Sungjun Lim", "Seojin Park", "Soeun Cheon", "Kyungwoo Song" ], "affiliation": "Yonsei University", "summary": "The field of preference optimization has made outstanding contributions to the alignment of language models with human preferences. Despite these advancements, recent methods still rely heavily on substantial paired (labeled) feedback data, leading to substantial resource expenditures.", "links": { "openreview": "https://openreview.net/forum?id=ghwxbTx7do", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-222", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Locality-aware Parallel Decoding for Efficient Autoregressive Image Generation", "authors": [ "Zhuoyang Zhang", "Luke J. Huang", "Chengyue Wu", "Shang Yang", "Kelly Peng", "Yao Lu", "Song Han" ], "affiliation": "Massachusetts Institute of Technology", "summary": "We present Locality-aware Parallel Decoding (LPD) to accelerate autoregressive image generation. Traditional autoregressive image generation relies on next-patch prediction, a memory-bound process that leads to high latency.", "links": { "openreview": "https://openreview.net/forum?id=h06l9w1clt", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-223", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Reasoning as Representation: Rethinking Visual Reinforcement Learning in Image Quality Assessment", "authors": [ "Shijie Zhao", "Xuanyu Zhang", "Weiqi Li", "Junlin Li", "Li zhang", "Tianfan Xue", "Jian Zhang" ], "affiliation": "ByteDance Inc.", "summary": "Reasoning-based image quality assessment (IQA) models trained through reinforcement learning (RL) exhibit exceptional generalization, yet the underlying mechanisms and critical factors driving this capability remain underexplored in current research. Moreover, despite their superior performance, these models incur inference energy usage and late...", "links": { "openreview": "https://openreview.net/forum?id=DkHt2K1g2Y", "arxiv": "", "detail": "" } }, { "uid": "iclr-2026-oral-224", "conference": "ICLR", "year": 2026, "track": "Oral", "title": "Half-order Fine-Tuning for Diffusion Model: A Recursive Likelihood Ratio Optimizer", "authors": [ "Tao Ren", "Zishi Zhang", "Jinyang Jiang", "Zehao Li", "Shentao Qin", "Yi Zheng", "Guanghao Li", "Qianyou Sun", "Yan Li", "Jiafeng Liang", "Xinping Li", "Yijie Peng" ], "affiliation": "Peking University", "summary": "The probabilistic diffusion model (DM), generating content by inferencing through a recursive chain structure, has emerged as a powerful framework for visual generation. After pre-training on enormous data, the model needs to be properly aligned to meet requirements for downstream applications.", "links": { "openreview": "https://openreview.net/forum?id=AZ6lqcvHLX", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-oral-1", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Benchmarking at the Edge of Comprehension", "authors": [ "Samuele Marro", "Jialin Yu", "Emanuele La Malfa", "Oishi Deb", "Jiawei Li", "Yibo Yang", "Ebey Abraham", "Sunando Sengupta", "Eric Sommerlade", "Michael Wooldridge", "Phil Torr" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71031" } }, { "uid": "icml-2026-oral-2", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Asymmetric Perturbation in Solving Bilinear Saddle-Point Optimization", "authors": [ "Kenshi Abe", "Mitsuki Sakamoto", "Kaito Ariu", "Atsushi Iwasaki" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71051" } }, { "uid": "icml-2026-oral-3", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: Don't Just \"Fix it in Post'': A Science of AI Must Study Learning Dynamics", "authors": [ "Stella Biderman", "Mohammad Aflah Khan", "Fatemehsadat Mireshghallah", "Catherine Arnett", "Fazl Barez", "Naomi Saphra" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71043" } }, { "uid": "icml-2026-oral-4", "conference": "ICML", "year": 2026, "track": "Oral", "title": "dnaHNet: A Scalable and Hierarchical Foundation Model for Genomic Sequence Learning", "authors": [ "Arnav Shah", "Junzhe Li", "Parsa Idehpour", "Adibvafa Fallahpour", "Brandon Wang", "Sukjun Hwang", "BO WANG", "Patrick Hsu", "Hani Goodarzi", "Albert Gu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71035" } }, { "uid": "icml-2026-oral-5", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Are VLMs Seeing or Just Saying? Uncovering the Illusion of Visual Re-examination", "authors": [ "Chufan Shi", "Cheng Yang", "Yaokang Wu", "Linghao Jin", "Bo Shui", "Taylor Berg-Kirkpatrick", "Xuezhe Ma" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71047" } }, { "uid": "icml-2026-oral-6", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Do We Need Adam? Surprisingly Strong and Sparse Reinforcement Learning with SGD in LLMs", "authors": [ "Sagnik Mukherjee", "Lifan Yuan", "Pavan Jayasinha", "Dilek Hakkani-Tür", "Hao Peng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71027" } }, { "uid": "icml-2026-oral-7", "conference": "ICML", "year": 2026, "track": "Oral", "title": "DiScoFormer: Plug-In Density and Score Estimation with Transformers", "authors": [ "Vasily Ilin", "Peter Sushko", "Ranjay Krishna" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71039" } }, { "uid": "icml-2026-oral-8", "conference": "ICML", "year": 2026, "track": "Oral", "title": "CLEAR: Context-Aware Learning with End-to-End Mask-Free Inference for Adaptive Subtitle Removal", "authors": [ "Qingdong He", "Chaoyi Wang", "Peng TANG", "Yifan Yang", "Xiaobin Hu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71048" } }, { "uid": "icml-2026-oral-9", "conference": "ICML", "year": 2026, "track": "Oral", "title": "A Systematic Study of Behavioral Cloning for Scientific Data Annotation", "authors": [ "Ishaan Singh Chandok", "Core Francisco Park" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71044" } }, { "uid": "icml-2026-oral-10", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Mixtures Closest To A Given Measure: A Semidefinite Programming Approach", "authors": [ "Srećko Ðurašinović", "Jean B Lasserre", "Victor Magron" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71052" } }, { "uid": "icml-2026-oral-11", "conference": "ICML", "year": 2026, "track": "Oral", "title": "FLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications", "authors": [ "Kieran Didi", "Sarah Alamdari", "Alex Lu", "Bruce Wittmann", "Kadina Johnston", "Ava Amini", "Ali Madani", "Maya Czeneszew", "Christian Dallago", "Kevin Yang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71036" } }, { "uid": "icml-2026-oral-12", "conference": "ICML", "year": 2026, "track": "Oral", "title": "daVinci-Dev: Agent-native Mid-training for Software Engineering", "authors": [ "Ji Zeng", "Dayuan Fu", "Tiantian Mi", "Zhuang Yumin", "Yaxing Huang", "Xuefeng Li", "Lyumanshan Ye", "Muhang Xie", "Qishuo Hua", "Zhen Huang", "Mohan Jiang", "Hanning Wang", "Shijie Xia", "Yang Xiao", "Jie Sun", "Yunze Wu", "Pengfei Liu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71032" } }, { "uid": "icml-2026-oral-13", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Learning Unmasking Policies for Diffusion Language Models", "authors": [ "Metod Jazbec", "Theo X. Olausson", "Louis Béthune", "Pierre Ablin", "Michael Kirchhof", "Joao Monteiro", "Victor Guilherme Turrisi da Costa", "Jason Ramapuram", "Marco Cuturi" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71028" } }, { "uid": "icml-2026-oral-14", "conference": "ICML", "year": 2026, "track": "Oral", "title": "LASER: Learning Active Sensing for Continuum Field Reconstruction", "authors": [ "Huayu Deng", "Jinghui Zhong", "Xiangming Zhu", "Yunbo Wang", "Xiaokang Yang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71040" } }, { "uid": "icml-2026-oral-15", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Protein Autoregressive Modeling via Multiscale Structure Generation", "authors": [ "Yanru Qu", "Cheng-Yen Hsieh", "Zaixiang Zheng", "Ge Liu", "Quanquan Gu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71037" } }, { "uid": "icml-2026-oral-16", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Multimodal Nested Learning for Decoupled and Coordinated Optimization", "authors": [ "Yanglin Feng", "Yang Qin", "Dezhong Peng", "Rui Wang", "Xiaomin Song", "Peng Hu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71041" } }, { "uid": "icml-2026-oral-17", "conference": "ICML", "year": 2026, "track": "Oral", "title": "AI Engram: In Search of Memory Traces in Artificial Intelligence", "authors": [ "Jea Kwon", "Dong-Kyum Kim", "Jiwon Kim", "Yonghyun Kim", "Woong Kook", "MEEYOUNG CHA" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71045" } }, { "uid": "icml-2026-oral-18", "conference": "ICML", "year": 2026, "track": "Oral", "title": "On the Convergence Rate of LoRA Gradient Descent", "authors": [ "Siqiao Mu", "Diego Klabjan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71053" } }, { "uid": "icml-2026-oral-19", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Motion Attribution for Video Generation", "authors": [ "Xindi Wu", "Despoina Paschalidou", "Jun Gao", "Antonio Torralba", "Laura Leal-Taixé", "Olga Russakovsky", "Sanja Fidler", "Jonathan Lorraine" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71049" } }, { "uid": "icml-2026-oral-20", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Less is Enough: Synthesizing Diverse Data in Feature Space of LLMs", "authors": [ "Zhongzhi Li", "Xuansheng Wu", "Yijiang Li", "Lijie Hu", "Ninghao Liu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71029" } }, { "uid": "icml-2026-oral-21", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Strategic Navigation or Stochastic Search? How Agents and Humans Reason Over Document Collections", "authors": [ "Lukasz Borchmann", "Jordy Van Landeghem", "Michał Turski", "Shreyansh Padarha", "Ryan Kearns", "Adam Mahdi", "Niels Rogge", "Clémentine Fourrier", "Siwei Han", "Huaxiu Yao", "Artemis Llabrés", "Yiming Xu", "Dimosthenis Karatzas", "Hao Zhang", "Anupam Datta" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71033" } }, { "uid": "icml-2026-oral-22", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Protein Fold Classification at Scale: Benchmarking and Pretraining", "authors": [ "Dexiong Chen", "Andrei Manolache", "Mathias Niepert", "Karsten Borgwardt" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71038" } }, { "uid": "icml-2026-oral-23", "conference": "ICML", "year": 2026, "track": "Oral", "title": "VenusBench-Mobile: A Challenging and User-Centric Benchmark for Mobile GUI Agents with Capability Diagnostics", "authors": [ "Yichen Gong", "Zhuohan Cai", "Sunhao Dai", "Yuqi Zhou", "Zhangxuan Gu", "Changhua Meng", "Shuheng Shen" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71034" } }, { "uid": "icml-2026-oral-24", "conference": "ICML", "year": 2026, "track": "Oral", "title": "OPUS: Towards Efficient and Principled Data Selection in Large Language Model Pre-training in Every Iteration", "authors": [ "Shaobo Wang", "Xuan Ouyang", "Tianyi Xu", "Yuzheng Hu", "Jialin Liu", "Guo Chen", "Tianyu Zhang", "Junhao Zheng", "Kexin Yang", "Xingzhang Ren", "Dayiheng Liu", "Linfeng Zhang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71030" } }, { "uid": "icml-2026-oral-25", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Guaranteed Optimal Compositional Explanations for Neurons", "authors": [ "Biagio La Rosa", "Leilani Gilpin" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71046" } }, { "uid": "icml-2026-oral-26", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Revenue Efficiency of Correlated Equilibria in First Price Auctions", "authors": [ "Anders Bo Ipsen", "Stratis Skoulakis" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71054" } }, { "uid": "icml-2026-oral-27", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Riemannian Metric Matching for Scalable Geometric Modeling of Distributions", "authors": [ "Jacob Bamberger", "Adam Gosztolai", "Pierre Vandergheynst", "Michael Bronstein", "Iolo Jones" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71042" } }, { "uid": "icml-2026-oral-28", "conference": "ICML", "year": 2026, "track": "Oral", "title": "PhotoAgent: Exploratory Visual Aesthetic Planning with Large Vision Models", "authors": [ "Mingde Yao", "Zhiyuan You", "King-Man Tam", "Menglu Wang", "Tianfan Xue" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71050" } }, { "uid": "icml-2026-oral-29", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: Anthropomorphic Misalignment Research Needs Stronger Evidence", "authors": [ "Vansh Gupta", "Peter Nutter", "Samuel Stante", "Andreas Krause", "Florian Tramer", "Lukas Fluri", "Xin Chen", "Anna Hedström" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71063" } }, { "uid": "icml-2026-oral-30", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Diffract: Spectral View of LLM Domain Adaptation", "authors": [ "Nikita Borodin", "Maria Krylova", "Artem Zabolotnyi", "Dmitry Aspisov", "Egor Shikov", "Nikita Tyuplyaev", "Oleg Travkin", "Roman Alferov", "Dmitry Vinichenko" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71079" } }, { "uid": "icml-2026-oral-31", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Learning Human-Robot Collaboration via Heterogeneous-Agent Lyapunov Policy Optimization", "authors": [ "Hao Zhang", "Yaru Niu", "Yikai Wang", "Ding Zhao", "Eric Tseng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71075" } }, { "uid": "icml-2026-oral-32", "conference": "ICML", "year": 2026, "track": "Oral", "title": "From Feasible to Practical: Pareto-Optimal Synthesis Planning", "authors": [ "Friedrich Hastedt", "Dongda Zhang", "Antonio Del rio chanona" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71067" } }, { "uid": "icml-2026-oral-33", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Controlled LLM Training on Spectral Sphere", "authors": [ "Tian Xie", "Haoming Luo", "Haoyu Tang", "Hu Yiwen", "Jason Liu", "Qingnan Ren", "Yang Wang", "Xin Zhao", "Rui Yan", "Bing Su", "Chong Luo", "Baining Guo" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71055" } }, { "uid": "icml-2026-oral-34", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Don't Force the Fit: Bounded Log-Likelihood Loss for Enhanced Reasoning in Large Language Models", "authors": [ "Feng Zhao", "Hong Zhang", "Yu Yang", "Ruilin Zhao", "Guandong Xu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71059" } }, { "uid": "icml-2026-oral-35", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Chebyshev Policies and the Mountain Car Problem: Reinforcement Learning for Low-dimensional Control Tasks", "authors": [ "Stefan Huber", "Hannes Unger", "Georg Schäfer", "Jakob Rehrl" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71071" } }, { "uid": "icml-2026-oral-36", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Monitoring Monitorability", "authors": [ "Melody Guan", "Miles Wang", "Micah Carroll", "Zehao Dou", "Annie Wei", "Marcus Williams", "Benjamin Arnav", "Joost Huizinga", "Ian Kivlichan", "Amelia Glaese", "Jakub Pachocki", "Bowen Baker" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71064" } }, { "uid": "icml-2026-oral-37", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Optimal and Scalable MAPF via Multi-Marginal Optimal Transport and Schrödinger Bridges", "authors": [ "Usman A Khan", "Joseph Durham" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71076" } }, { "uid": "icml-2026-oral-38", "conference": "ICML", "year": 2026, "track": "Oral", "title": "ReViT: Rotational-equivariant Vision Transformers for Neural PDE Solvers", "authors": [ "Hao Wei", "Björn List", "Nils Thuerey" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71068" } }, { "uid": "icml-2026-oral-39", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Evaluating Robustness of Reasoning Models on Parameterized Logical Problems", "authors": [ "Naïm Es-sebbani", "Esteban Marquer", "Yakoub Salhi", "Zied Bouraoui" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71060" } }, { "uid": "icml-2026-oral-40", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Detecting the Semantic Fixed Point: A Geometric Framework for Efficient Inference", "authors": [ "Jiawei Gu", "Ziyue Qiao", "Xiao Luo" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71056" } }, { "uid": "icml-2026-oral-41", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Maximum Likelihood Reinforcement Learning", "authors": [ "Fahim Tajwar", "Guanning Zeng", "Yueer Zhou", "Yuda Song", "Daman Arora", "Yiding Jiang", "Jeff Schneider", "Russ Salakhutdinov", "Haiwen Feng", "Andrea Zanette" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71072" } }, { "uid": "icml-2026-oral-42", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Foundations of Equivariant Deep Learning: Unifying Graph and Sheaf Neural Networks", "authors": [ "Yoshihiro Maruyama" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71080" } }, { "uid": "icml-2026-oral-43", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Rex: A Family of Reversible Exponential (Stochastic) Runge-Kutta Solvers", "authors": [ "Zander Blasingame", "Chen Liu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71069" } }, { "uid": "icml-2026-oral-44", "conference": "ICML", "year": 2026, "track": "Oral", "title": "ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios", "authors": [ "Xinyi Hu", "Yuhao Shen", "Zhang Baolin", "Hengxin Zhang", "Jun Dai", "Shuang Ge", "Chen Lei", "Yue Li", "Mingcheng Wan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71057" } }, { "uid": "icml-2026-oral-45", "conference": "ICML", "year": 2026, "track": "Oral", "title": "The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes", "authors": [ "Mohammad Taufeeque", "Stefan Heimersheim", "Adam Gleave", "Chris Cundy" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71065" } }, { "uid": "icml-2026-oral-46", "conference": "ICML", "year": 2026, "track": "Oral", "title": "On Minimum Depth and Width of Floating-Point Neural Networks for Representing Floating-Point Functions", "authors": [ "Sejun Park", "Yeachan Park", "Geonho Hwang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71081" } }, { "uid": "icml-2026-oral-47", "conference": "ICML", "year": 2026, "track": "Oral", "title": "RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies", "authors": [ "Yinpei Dai", "Hongze Fu", "Jayjun Lee", "Yuejiang Liu", "Haoran Zhang", "Jianing Yang", "Chelsea Finn", "Nima Fazeli", "Joyce Chai" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71077" } }, { "uid": "icml-2026-oral-48", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Minimax Optimal Strategy for Delayed Observations in Online Reinforcement Learning", "authors": [ "Harin Lee", "Kevin Jamieson" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71073" } }, { "uid": "icml-2026-oral-49", "conference": "ICML", "year": 2026, "track": "Oral", "title": "TG-RAG: A Retrieval-Augmented Framework for Reasoning Guidance in Specialized Domains", "authors": [ "Liang Su", "Mingyang Zhang", "Yun Xiong", "Tengfei LIU", "Siwei Zhang", "Xi Chen", "Li Sun" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71061" } }, { "uid": "icml-2026-oral-50", "conference": "ICML", "year": 2026, "track": "Oral", "title": "The Expressivity Limits of Transformers", "authors": [ "Maxime Meyer", "Mario Michelessa", "Caroline Chaux", "Vincent Tan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71082" } }, { "uid": "icml-2026-oral-51", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Optimal Decision-Making Based on Prediction Sets", "authors": [ "Tao Wang", "Edgar Dobriban" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71074" } }, { "uid": "icml-2026-oral-52", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Towards Sub-second Biological Foundation Model Infrastructure: A Quantized Consistency Diffusion Framework for Molecular Docking", "authors": [ "Kexin Zhang", "Weichen Qin", "Yue Teng", "Jiale Yu", "Yuanyuan Ma", "jinyu lin", "Liping Sun", "Jie Zheng", "Jingyi Yu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71070" } }, { "uid": "icml-2026-oral-53", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Understanding Reasoning Collapse in LLM Agent Reinforcement Learning", "authors": [ "Zihan (Zenus) Wang", "Chi Gui", "Xing Jin", "Qineng Wang", "Licheng Liu", "Kangrui Wang", "Shiqi Chen", "Linjie Li", "Zhengyuan Yang", "Pingyue Zhang", "Yiping Lu", "Jiajun Wu", "Li Fei-Fei", "Lijuan Wang", "Yejin Choi", "Manling Li" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71062" } }, { "uid": "icml-2026-oral-54", "conference": "ICML", "year": 2026, "track": "Oral", "title": "VALUEFLOW: Toward Pluralistic and Steerable Value-based Alignment in Large Language Models", "authors": [ "Woojin Kim", "Sieun Hyeon", "Jusang Oh", "Jaeyoung Do" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71066" } }, { "uid": "icml-2026-oral-55", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Unsupervised Partner Design Enables Robust Ad-hoc Teamwork", "authors": [ "Constantin Ruhdorfer", "Matteo Bortoletto", "Victor Oei", "Anna Penzkofer", "Andreas Bulling" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71078" } }, { "uid": "icml-2026-oral-56", "conference": "ICML", "year": 2026, "track": "Oral", "title": "MuonSSM: Orthogonalizing State Space Models for Sequence Modeling", "authors": [ "Thai Khanh Nguyen", "Ngoc Bich Uyen Vo", "Thieu Vo", "Tan Nguyen", "Cuong Pham" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71058" } }, { "uid": "icml-2026-oral-57", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Mitigating Reward Hacking in RLHF via Bayesian Non-negative Reward Modeling", "authors": [ "Zhibin Duan", "Guowei Rong", "Zhuo Li", "Bo Chen", "Mingyuan Zhou", "Dandan Guo" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71087" } }, { "uid": "icml-2026-oral-58", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Any-Order GPT as Masked Diffusion Model: Decoupling Formulation and Architecture", "authors": [ "Shuchen Xue", "Tianyu Xie", "Tianyang Hu", "Zijin Feng", "Jiacheng Sun", "Kenji Kawaguchi", "Zhenguo Li", "Zhi-Ming Ma" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71083" } }, { "uid": "icml-2026-oral-59", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: Irresponsible AI: big tech’s influence on AI research and associated impacts", "authors": [ "Alex Hernandez-Garcia", "Alexandra Volokhova", "Ezekiel Williams", "Dounia Shaaban Kabakibo", "Mélisande Teng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71091" } }, { "uid": "icml-2026-oral-60", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Bad Seeing or Bad Thinking? Rewarding Perception for Multimodal Reasoning", "authors": [ "WANG", "Qixin Xu", "Changpeng Wang", "Taofeng Xue", "Chong Peng", "Wenhu Chen", "Fangzhen Lin" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71095" } }, { "uid": "icml-2026-oral-61", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: Stop Automating Peer Review Without Rigorous Evaluation", "authors": [ "Joachim Baumann", "Jiaxin Pei", "Sanmi Koyejo", "Dirk Hovy" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71099" } }, { "uid": "icml-2026-oral-62", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Jailbreak Foundry: From Papers to Runnable Attacks for Reproducible Benchmarking", "authors": [ "Zhicheng Fang", "Jingjie Zheng", "Chenxu Fu", "Wei Xu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71103" } }, { "uid": "icml-2026-oral-63", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Exact Functional ANOVA Decomposition for Categorical Inputs", "authors": [ "Baptiste Ferrere", "Nicolas Bousquet", "Gamboa Fabrice", "Jean-Michel Loubes", "Joseph Muré" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71107" } }, { "uid": "icml-2026-oral-64", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Quantifying Frontier LLM Capabilities for Container Sandbox Escape", "authors": [ "Rahul Marchand", "Art Cathain", "Jerome Wynne", "Philippos Giavridis", "Sam Deverett", "John Wilkinson", "Jason Gwartz", "Harry Coppock" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71104" } }, { "uid": "icml-2026-oral-65", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Joint Learning in the Gaussian Single Index Model", "authors": [ "Loucas Pillaud-Vivien", "Adrien Schertzer" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71108" } }, { "uid": "icml-2026-oral-66", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: The AI Imperative: Scaling High-Quality Peer Review in Machine Learning", "authors": [ "Qiyao Wei", "Samuel Holt", "Jing Yang", "Markus Wulfmeier", "Mihaela van der Schaar" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71100" } }, { "uid": "icml-2026-oral-67", "conference": "ICML", "year": 2026, "track": "Oral", "title": "DroneDINO: Towards Heterogeneous Routed Mixture of Experts for Drone-based Unified Object Detection", "authors": [ "Rui Chen", "Dongdong Li", "Yan Fan", "Yan Liu", "Yangliu Kuai", "Pengfei Zhu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71096" } }, { "uid": "icml-2026-oral-68", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Error Propagation Mechanisms and Compensation Strategies for Quantized Diffusion Models", "authors": [ "Songwei Liu", "Chao Zeng", "Chenqian Yan", "Xurui Peng", "WANG", "Fangmin Chen", "Xing Mei" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71084" } }, { "uid": "icml-2026-oral-69", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers", "authors": [ "Adam Karvonen", "James Chua", "Clément Dumas", "Kit Fraser-Taliente", "Subhash Kantamneni", "Julian Minder", "Euan Ong", "Arnab Sen Sharma", "Daniel Wen", "Owain Evans", "Samuel Marks" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71092" } }, { "uid": "icml-2026-oral-70", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Reinforcement Learning with Evolving Rubrics for Deep Research", "authors": [ "Rulin Shao", "Akari Asai", "Shannon Shen", "Hamish Ivison", "Varsha Kishore", "Jingming Zhuo", "Xinran Zhao", "Molly Park", "Samuel Finlayson", "David Sontag", "Tyler Murray", "Sewon Min", "Pradeep Dasigi", "Luca Soldaini", "Faeze Brahman", "Scott Yih", "Sherry Wu", "Luke Zettlemoyer", "Yoon Kim", "Hannaneh Hajishirzi", "Pang Wei Koh" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71088" } }, { "uid": "icml-2026-oral-71", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Simultaneous Speech-to-Speech Translation Without Aligned Data", "authors": [ "Tom Labiausse", "Romain Fabre", "Yannick Estève", "Alexandre Défossez", "Neil Zeghidour" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71089" } }, { "uid": "icml-2026-oral-72", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Incentivizing Truthfulness and Collaborative Fairness in Bayesian Learning", "authors": [ "Rachael Hwee Ling Sim", "Jue Fan", "Xiao Tian", "Xinyi Xu", "Patrick Jaillet", "Bryan Kian Hsiang Low" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71101" } }, { "uid": "icml-2026-oral-73", "conference": "ICML", "year": 2026, "track": "Oral", "title": "SVRG and Beyond via Posterior Correction", "authors": [ "Nico Daheim", "Thomas Moellenhoff", "James Ming Liang Ang", "Mohammad Emtiyaz Khan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71109" } }, { "uid": "icml-2026-oral-74", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Lottery Prior: Randomized Neural Compression for Zero-Shot Inverse Problems", "authors": [ "Haotian Wu", "Di You", "Pier Luigi Dragotti", "Deniz Gunduz" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71097" } }, { "uid": "icml-2026-oral-75", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Large Language Models Develop Novel Social Biases Through Adaptive Exploration", "authors": [ "Addison J. Wu", "Ryan Liu", "Xuechunzi Bai", "Thomas Griffiths" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71093" } }, { "uid": "icml-2026-oral-76", "conference": "ICML", "year": 2026, "track": "Oral", "title": "High-accuracy and dimension-free sampling with diffusions", "authors": [ "Khashayar Gatmiry", "Sitan Chen", "Adil Salim" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71085" } }, { "uid": "icml-2026-oral-77", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Robust Harmful Features Under Jailbreak Attacks: Mechanistic Evidence from Attention Head Specialization in Large Language Models", "authors": [ "Yanchen Yin", "Dongqi Han", "Linghui Li" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71105" } }, { "uid": "icml-2026-oral-78", "conference": "ICML", "year": 2026, "track": "Oral", "title": "The Flexibility Trap: Rethinking the Value of Arbitrary Order in Diffusion Language Models", "authors": [ "Zanlin Ni", "Shenzhi Wang", "Yang Yue", "Tianyu Yu", "Weilin Zhao", "Yeguo Hua", "Tianyi Chen", "Jun Song", "YuCheng", "Bo Zheng", "Gao Huang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71086" } }, { "uid": "icml-2026-oral-79", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Is Your LLM Overcharging You? Tokenization, Transparency, and Incentives", "authors": [ "Ander Artola Velasco", "Stratis Tsirtsis", "Nastaran Okati", "Manuel Gomez-Rodriguez" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71102" } }, { "uid": "icml-2026-oral-80", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Mechanistic Data Attribution: Tracing the Training Origins of Interpretable LLM Units", "authors": [ "Jianhui Chen", "Yuzhang Luo", "Liangming Pan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71094" } }, { "uid": "icml-2026-oral-81", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Video-Based Optimal Transport for Feedback-Efficient Offline Preference-Based Reinforcement Learning", "authors": [ "Minh-Tung Luu", "Hwanhee Kim", "Younghwan Lee", "Chang Yoo" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71090" } }, { "uid": "icml-2026-oral-82", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Scalable Event Cloud Network for Event-based Classification", "authors": [ "Hongwei Ren", "Fei Ma", "Xiaopeng LIN", "Yuetong Fang", "Hongxiang Huang", "Yue Zhou", "Yulong Huang", "Haotian FU", "Ziyi Yang", "Youxin Jiang", "Xiangqian Wu", "Bojun Cheng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71098" } }, { "uid": "icml-2026-oral-83", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Towards Fair Sequential Decision-Making: A Causal Decomposition Approach", "authors": [ "Jiajun Chen", "Jin Tian", "Chris Quinn" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71110" } }, { "uid": "icml-2026-oral-84", "conference": "ICML", "year": 2026, "track": "Oral", "title": "When the Prompt Becomes Visual: Vision-Centric Jailbreak Attacks for Large Image Editing Models", "authors": [ "Jiacheng Hou", "Yining Sun", "Ruochong Jin", "Haochen Han", "Fangming Liu", "Victor Chan", "Alex Jinpeng Wang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71106" } }, { "uid": "icml-2026-oral-85", "conference": "ICML", "year": 2026, "track": "Oral", "title": "A Recursive Decomposition Framework for Causal Structure Learning in the Presence of Latent Variables", "authors": [ "Zheng Li", "Feng Xie", "Shenglan Nie", "Xichen Guo", "Ruxin Wang", "Hao Zhang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71123" } }, { "uid": "icml-2026-oral-86", "conference": "ICML", "year": 2026, "track": "Oral", "title": "ConFlux: Multivariate Time Series in Flux, One Unified Forecast in Confluence", "authors": [ "Shiyu Wang", "Yuchen Fang", "Juntong Ni", "Ziyi Zhang", "Baichuan Mo", "Xinyue Zhong", "Chengxin Wang", "Zhou Ye", "Yang Xiang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71135" } }, { "uid": "icml-2026-oral-87", "conference": "ICML", "year": 2026, "track": "Oral", "title": "CAT-Q: Cost-efficient and Accurate Ternary Quantization for LLMs", "authors": [ "Shigeng Wang", "Chao Li", "Yangyuxuan Kang", "Jiawei Fan", "Anbang Yao" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71111" } }, { "uid": "icml-2026-oral-88", "conference": "ICML", "year": 2026, "track": "Oral", "title": "CoEvol-NO: State and Coordinate Co-Evolution with an Error-Driven Predictor-Corrector Paradigm for Neural Operator Transformer", "authors": [ "Jianqiao Zeng", "Ruocheng Wang", "Yanzhi Liu", "Hao Xiong", "Junchi Yan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71127" } }, { "uid": "icml-2026-oral-89", "conference": "ICML", "year": 2026, "track": "Oral", "title": "DPO Unchained: Your Training Algorithm is Secretly Disentangled in Human Choice Theory (and Its Loss' Convexity is Dispensable)", "authors": [ "Wenxuan Zhou", "Shujian Zhang", "brice magdalou", "John Lambert", "Ehsan Amid", "Richard Nock", "Andrew Hard" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71131" } }, { "uid": "icml-2026-oral-90", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: The Alignment Community is Unintentionally Building a Censor’s Toolkit", "authors": [ "Sarah Ball", "Phil Hackemann" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71119" } }, { "uid": "icml-2026-oral-91", "conference": "ICML", "year": 2026, "track": "Oral", "title": "From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model", "authors": [ "Bing Hu", "Zaijing Li", "Rui Shao", "Junda Chen", "April Hua Liu", "Wei-Shi Zheng", "Liqiang Nie" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71115" } }, { "uid": "icml-2026-oral-92", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Information Flow Reveals When to Trust Language Models", "authors": [ "Rui Xu", "Yi Chen", "Jiujiu Chen", "Sihong Xie" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71120" } }, { "uid": "icml-2026-oral-93", "conference": "ICML", "year": 2026, "track": "Oral", "title": "From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space", "authors": [ "Lehui Li", "Yuyao Wang", "Jisheng Yan", "Wei Zhang", "Jinliang Deng", "Haoliang Sun", "Zhongyi Han", "Yongshun Gong" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71136" } }, { "uid": "icml-2026-oral-94", "conference": "ICML", "year": 2026, "track": "Oral", "title": "High-accuracy sampling for diffusion models and log-concave distributions", "authors": [ "Fan Chen", "Sinho Chewi", "Constantinos Daskalakis", "Alexander Rakhlin" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71132" } }, { "uid": "icml-2026-oral-95", "conference": "ICML", "year": 2026, "track": "Oral", "title": "From Pixels to Tokens: A Systematic Study of Latent Action Supervision for Vision-Language-Action Models", "authors": [ "Yihan Lin", "Haoyang Li", "Yang Li", "Haitao Shen", "Yihan Zhao", "Chao Shao", "Jing Zhang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71116" } }, { "uid": "icml-2026-oral-96", "conference": "ICML", "year": 2026, "track": "Oral", "title": "POET-X: Memory-efficient LLM Training by Scaling Orthogonal Transformation", "authors": [ "Zeju Qiu", "Lixin LIU", "Adrian Weller", "Han Shi", "Weiyang Liu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71112" } }, { "uid": "icml-2026-oral-97", "conference": "ICML", "year": 2026, "track": "Oral", "title": "DISCO: Mitigating Bias in Deep Learning with Conditional Distance Correlation", "authors": [ "Emre Kavak", "Tom Nuno Wolf", "Christian Wachinger" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71124" } }, { "uid": "icml-2026-oral-98", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Geometric Flow Grounding: A Unified Manifold Decoupling Framework for Dynamics Discovery and Verification", "authors": [ "Chang Yu", "Yuxuan Luo", "Yixuan Du", "Yuqing Zhou", "Siyuan Li", "Jingbo Zhou", "jiawei jiang", "Zhen Lei", "Stan Z Li" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71128" } }, { "uid": "icml-2026-oral-99", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Pretrained Vision-Language-Action Models are Surprisingly Resistant to Forgetting in Continual Learning", "authors": [ "Huihan Liu", "Changyeon Kim", "Bo Liu", "Minghuan Liu", "Yuke Zhu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71117" } }, { "uid": "icml-2026-oral-100", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Solving Time-Dependent Differential Equations with Physical Dynamical Systems", "authors": [ "Chuan Liu", "Yijie Chen", "Ruibing Song", "Wenhao Huang", "Chunshu Wu", "Deqian Kong", "Ying Nian Wu", "Kaiyuan Yang", "Ang Li", "Tony Geng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71129" } }, { "uid": "icml-2026-oral-101", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Modeling Hierarchical Thinking in Large Reasoning Models", "authors": [ "G M Shahariar", "Erfan Shayegani", "Ali Nazari", "Nael Abu-Ghazaleh" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71121" } }, { "uid": "icml-2026-oral-102", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Rational Transductors", "authors": [ "Mehryar Mohri" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71133" } }, { "uid": "icml-2026-oral-103", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Markov Chain Monte Carlo without Evaluating the Target: an Auxiliary Variable Approach", "authors": [ "Wei Yuan", "Guanyang Wang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71137" } }, { "uid": "icml-2026-oral-104", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Disentangling Latent Risk Pathways via Bayesian Hypergraph Inference", "authors": [ "Shengxian Ding", "Haonan Gao", "Pangpang Liu", "Xinyuan Tian", "Yize Zhao" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71125" } }, { "uid": "icml-2026-oral-105", "conference": "ICML", "year": 2026, "track": "Oral", "title": "ReQAT: Achieving Full-Precision Reasoning Accuracy with 4-bit Floating-Point Quantization-Aware Training", "authors": [ "Janghwan Lee", "Sihwa Lee", "Jinseok Kim", "Yongjik Kim", "Jieun Lim", "Jinwook Oh", "Jungwook Choi" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71113" } }, { "uid": "icml-2026-oral-106", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Path-dependent Discrete Amortized Inference", "authors": [ "Tiago Silva", "Esmeralda S. Whitammer", "Salem Lahlou" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71138" } }, { "uid": "icml-2026-oral-107", "conference": "ICML", "year": 2026, "track": "Oral", "title": "To Grok Grokking: Provable Grokking in Ridge Regression", "authors": [ "Mingyue Xu", "Gal Vardi", "Itay Safran" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71134" } }, { "uid": "icml-2026-oral-108", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Training-Free Bayesian Filtering with Generative Emulators", "authors": [ "Thomas Savary", "François Rozet", "Gilles Louppe" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71130" } }, { "uid": "icml-2026-oral-109", "conference": "ICML", "year": 2026, "track": "Oral", "title": "XR-1: Towards Versatile Vision-Language-Action Models via Learning Unified Vision-Motion Representations", "authors": [ "Shichao Fan", "Kun Wu", "Zhengping Che", "Xinhua Wang", "Di Wu", "Fei Liao", "Ning Liu", "Yixue Zhang", "Zhen Zhao", "Zhiyuan Xu", "Meng Li", "Qingjie Liu", "Shanghang Zhang", "Min Wan", "Jian Tang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71118" } }, { "uid": "icml-2026-oral-110", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Skip a Layer or Loop It? Learning Program-of-Layers in LLMs", "authors": [ "Ziyue Li", "Yang Li", "Tianyi Zhou" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71114" } }, { "uid": "icml-2026-oral-111", "conference": "ICML", "year": 2026, "track": "Oral", "title": "On the Identifiability of Poisson Branching Structural Causal Model Under Latent Confounding", "authors": [ "Jie Qiao", "Zihuai Zeng", "Ruichu Cai", "Zhengming Chen", "Zhifeng Hao" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71126" } }, { "uid": "icml-2026-oral-112", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Reward-free Alignment for Conflicting Objectives", "authors": [ "Peter Chen", "Xiaopeng Li", "Xi Chen", "Tianyi Lin" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71122" } }, { "uid": "icml-2026-oral-113", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Mind Your Margin and Boundary: Are Your Distilled Datasets Truly Robust?", "authors": [ "Muquan Li", "Yingyi Ma", "Yihong Huang", "Hang Gou", "KE QIN", "Ming Li", "Yuan-Fang Li", "Tao He" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71143" } }, { "uid": "icml-2026-oral-114", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: There are futures that benchmark-driven AI cannot see", "authors": [ "Sobhan Lotfi", "Ava Iranmanesh", "Lachin Naghashyar", "Ali Shirali", "Fateme Haredasht", "Sanmi Koyejo", "Phil Torr", "Yong Suk Lee", "Fazl Barez", "Joel Lehman", "Peter Norvig", "Arvind Narayanan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71151" } }, { "uid": "icml-2026-oral-115", "conference": "ICML", "year": 2026, "track": "Oral", "title": "FlatLand: Personalized Graph Federated Learning via Tailored Lorentz Space", "authors": [ "Jiahong Liu", "Ram Samarth B B", "Xinyu Fu", "Menglin Yang", "Weixi Zhang", "ZHITAO YING", "Irwin King" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71159" } }, { "uid": "icml-2026-oral-116", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Necessary Conditions for Compositional Generalization of Embedding Models", "authors": [ "Arnas Uselis", "Andrea Dittadi", "Seong Joon Oh" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71155" } }, { "uid": "icml-2026-oral-117", "conference": "ICML", "year": 2026, "track": "Oral", "title": "3ViewSense: Spatial and Mental Perspective Reasoning from Orthographic Views in Vision-Language Models", "authors": [ "Shaoxiong Zhan", "Yanlin Lai", "Zheng Liu", "Zijian Lin", "Lin Hai", "Xiaodong Cai", "Shen Li", "Wen Huang", "Hai-Tao Zheng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71163" } }, { "uid": "icml-2026-oral-118", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Midtraining Bridges Pretraining and Posttraining Distributions", "authors": [ "Emmy Liu", "Graham Neubig", "Chenyan Xiong" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71139" } }, { "uid": "icml-2026-oral-119", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Distributional Inverse Reinforcement Learning", "authors": [ "Feiyang Wu", "Ye Zhao", "Anqi Wu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71147" } }, { "uid": "icml-2026-oral-120", "conference": "ICML", "year": 2026, "track": "Oral", "title": "ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models", "authors": [ "Long (Tony) Lian", "Sida Wang", "Felix Juefei-Xu", "Tsu-Jui Fu", "Xiuyu Li", "Adam Yala", "Trevor Darrell", "Alane Suhr", "Yuandong Tian", "Xi Lin" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71140" } }, { "uid": "icml-2026-oral-121", "conference": "ICML", "year": 2026, "track": "Oral", "title": "On the Difficulty of Learning a Meta-network for Training Data Selection", "authors": [ "Zilin Du", "Junqi Zhao", "Albert Boyang Li" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71144" } }, { "uid": "icml-2026-oral-122", "conference": "ICML", "year": 2026, "track": "Oral", "title": "On Computation and Reinforcement Learning", "authors": [ "Raj Ghugare", "Michał Bortkiewicz", "Alicja Ziarko", "Benjamin Eysenbach" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71148" } }, { "uid": "icml-2026-oral-123", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Non-Euclidean Gradient Descent Operates at the Edge of Stability", "authors": [ "Rustem Islamov", "Michael Crawshaw", "Jeremy Cohen", "Robert Gower" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71156" } }, { "uid": "icml-2026-oral-124", "conference": "ICML", "year": 2026, "track": "Oral", "title": "CausalGame: Benchmarking Causal Thinking of LLM Agents in Games", "authors": [ "Zhenhao Chen", "Yongqiang Chen", "Chenxi Liu", "Junchi Yu", "Xiangchen Song", "Zijian Li", "Jialin Li", "Phil Torr", "Bo Han", "Kun Zhang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71152" } }, { "uid": "icml-2026-oral-125", "conference": "ICML", "year": 2026, "track": "Oral", "title": "MV-FGAD: Towards Efficient and Effective Federated Graph Anomaly Detection via Multi-view Learning", "authors": [ "Junyi Yan", "KE LIANG", "Hao Yu", "Meng Liu", "Hao Tan", "Tianrui Liu", "Jun-Jie Huang", "Xinwang Liu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71160" } }, { "uid": "icml-2026-oral-126", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning", "authors": [ "Jiaqi Liu", "Kaiwen Xiong", "Peng Xia", "Yiyang Zhou", "Haonian Ji", "Lu Feng", "Siwei Han", "Mingyu Ding", "Huaxiu Yao" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71164" } }, { "uid": "icml-2026-oral-127", "conference": "ICML", "year": 2026, "track": "Oral", "title": "PRISM: Gauge-Invariant Tangent-Space Differentially Private LoRA", "authors": [ "Shihao Wang", "Xueru Zhang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71157" } }, { "uid": "icml-2026-oral-128", "conference": "ICML", "year": 2026, "track": "Oral", "title": "TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior", "authors": [ "Gül Sena Altıntaş", "Malikeh Ehghaghi", "Brian Lester", "Fengyuan Liu", "Wanru Zhao", "Marco Ciccone", "Colin Raffel" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71141" } }, { "uid": "icml-2026-oral-129", "conference": "ICML", "year": 2026, "track": "Oral", "title": "The Signal is in the Steps: Local Scoring for Reasoning Data Selection", "authors": [ "Hoang Anh Just", "Myeongseob Ko", "Ruoxi Jia" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71145" } }, { "uid": "icml-2026-oral-130", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Second-Order Smooth Planning with Optimal-Transport Bellman Smoothing", "authors": [ "Tuan Dam" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71149" } }, { "uid": "icml-2026-oral-131", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Characterizing, Evaluating, and Optimizing Complex Reasoning", "authors": [ "Haoran Zhang", "Yafu Li", "Zhi Wang", "Zhilin Wang", "Shunkai Zhang", "Xiaoye Qu", "Yu Cheng" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71153" } }, { "uid": "icml-2026-oral-132", "conference": "ICML", "year": 2026, "track": "Oral", "title": "PhenoBrain: Phenotype-Conditioned Long-Range Communication for Multi-Modal Brain Network Analysis", "authors": [ "Lingyuan Meng", "KE LIANG", "Hao Li", "Meng Liu", "Weijia Shi", "Miaomiao Li", "Yang Gao", "Xinwang Liu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71161" } }, { "uid": "icml-2026-oral-133", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Holi-Spatial: Evolving Video Streams into Holistic 3D Spatial Intelligence", "authors": [ "Yuanyuan Gao", "Hao Li", "Yifei Liu", "Xinhao Ji", "Yuning Gong", "Yuanjun Liao", "Fangfu Liu", "Manyuan Zhang", "Yuchen Yang", "Dan Xu", "Xue Yang", "Huaxi Huang", "Hongjie Zhang", "Ziwei Liu", "Xiao Sun", "Dingwen Zhang", "Zhihang Zhong" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71165" } }, { "uid": "icml-2026-oral-134", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Robust Contextual Optimization with Missing Covariates", "authors": [ "Qingyuan Xu", "Ruiwei Jiang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71158" } }, { "uid": "icml-2026-oral-135", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Towards Hierarchy–Uniformity Equilibrium: Recovering Semantic Depth in Hypergraph Contrastive Learning", "authors": [ "Ruiting Zhao", "Ming Li", "Lixin Cui", "Lu Bai", "Feilong Cao", "Ke Lv", "Pietro Lió" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71162" } }, { "uid": "icml-2026-oral-136", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Transforming Weather Data from Pixel to Latent Space", "authors": [ "Sijie Zhao", "Feng Liu", "Xueliang Zhang", "Hao Chen", "Tao Han", "JUNCHAO GONG", "Ran Tao", "Pengfeng Xiao", "Xinyu Gu", "LEI BAI" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71146" } }, { "uid": "icml-2026-oral-137", "conference": "ICML", "year": 2026, "track": "Oral", "title": "SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Models", "authors": [ "jing wu", "Jianhua Wu", "Jiayi Guan", "Jiahong Chen", "Jinghui Lu", "Hangjun Ye", "Bingzhao Gao", "Long Chen" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71166" } }, { "uid": "icml-2026-oral-138", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Stabilizing the Q-Gradient Field for Policy Smoothness in Actor-Critic Methods", "authors": [ "Jeong Woon Lee", "Kyoleen Kwak", "Daeho Kim", "Hyoseok Hwang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71150" } }, { "uid": "icml-2026-oral-139", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Rare Event Analysis of Large Language Models", "authors": [ "Jake McAllister Dorman", "Edward Gillman", "Dominic C Rose", "Jamie Mair", "Juan Garrahan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71154" } }, { "uid": "icml-2026-oral-140", "conference": "ICML", "year": 2026, "track": "Oral", "title": "WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inference", "authors": [ "Aiwei Liu", "Minghua He", "Shaoxun Zeng", "Sijun Zhang", "Linhao Zhang", "Chuhan Wu", "Wei Jia", "Yuan Liu", "Zhou Xiao", "Jie Zhou" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71142" } }, { "uid": "icml-2026-oral-141", "conference": "ICML", "year": 2026, "track": "Oral", "title": "A Random Matrix Perspective on the Consistency of Diffusion Models", "authors": [ "Binxu Wang", "Jacob A Zavatone-Veth", "Cengiz Pehlevan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71191" } }, { "uid": "icml-2026-oral-142", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Expressivity-Efficiency Tradeoffs for Hybrid Sequence Models", "authors": [ "John Cooper", "Mingchen Ma", "Ilias Diakonikolas", "Frederic Sala" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71167" } }, { "uid": "icml-2026-oral-143", "conference": "ICML", "year": 2026, "track": "Oral", "title": "$\\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment", "authors": [ "Victor Barres", "Honghua Dong", "Soham Ray", "Xujie Si", "Karthik Narasimhan" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71171" } }, { "uid": "icml-2026-oral-144", "conference": "ICML", "year": 2026, "track": "Oral", "title": "DOUBT: Decoupled Object-level Understanding and Bridging via vMF-based Trustworthiness for Hallucination Detection in MLLMs", "authors": [ "Kaiqi Chen", "Yang Qin", "Changhao He", "Xi Peng", "Peng Hu" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71175" } }, { "uid": "icml-2026-oral-145", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Faster Activation Functions at the Edge for Post-Training Speedups", "authors": [ "Anton Lydike", "Jun Bi", "Jackson Woodruff" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71179" } }, { "uid": "icml-2026-oral-146", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: AI Should Facilitate Democratic Deliberation at Scale", "authors": [ "José Ramón Enríquez", "Jiaxin Pei", "Alex Pentland" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71183" } }, { "uid": "icml-2026-oral-147", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Position: AI/ML Deepfake Research is Misaligned with AI Generated Non-Consensual Intimate Imagery (AIG-NCII)", "authors": [ "Qiwei Li", "Wells Lucas Santo", "Sarita Schoenebeck", "Eric Gilbert" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71187" } }, { "uid": "icml-2026-oral-148", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Equilibrium Pricing in Oligopolistic Data Markets", "authors": [ "Bhaskar Ray Chaudhury", "Jugal Garg", "Eklavya Sharma", "Jiaxin Song" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71184" } }, { "uid": "icml-2026-oral-149", "conference": "ICML", "year": 2026, "track": "Oral", "title": "FlashSinkhorn: IO-Aware Entropic Optimal Transport on GPU", "authors": [ "Felix X.-F. Ye", "Xingjie Li", "An Yu", "Ming-Ching Chang", "LINSONG CHU", "Davis Wertheimer" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71180" } }, { "uid": "icml-2026-oral-150", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Characterizing Agents in Production", "authors": [ "Melissa Pan", "Negar Arabzadeh", "Riccardo Cogo", "Yuxuan Zhu", "Alexander Xiong", "Lakshya A Agrawal", "Huanzhi Mao", "Emma Shen", "Sid Pallerla", "Liana Patel", "Shu Liu", "Tianneng Shi", "Xiaoyuan Liu", "Jared Davis", "Emmanuele Lacavalla", "Alessandro Basile", "Shuyi Yang", "Paul Castro", "Daniel Kang", "Koushik Sen", "Dawn Song", "Joseph E Gonzalez", "Ion Stoica", "Matei Zaharia", "Marquita Ellis" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71172" } }, { "uid": "icml-2026-oral-151", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Learning to Theorize the World from Observation", "authors": [ "Doojin Baek", "Gyubin Lee", "Junyeob Baek", "Hosung Lee", "Sungjin Ahn" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71176" } }, { "uid": "icml-2026-oral-152", "conference": "ICML", "year": 2026, "track": "Oral", "title": "How much can language models memorize?", "authors": [ "John Morris", "Chawin Sitawarin", "Narine Kokhlikyan", "Chuan Guo", "Edward Suh", "Alexander Rush", "Kamalika Chaudhuri", "Saeed Mahloujifar" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71168" } }, { "uid": "icml-2026-oral-153", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Equivalence of Context and Parameter Updates in Modern Transformer Blocks", "authors": [ "Adrian Goldwaser", "Michael Munn", "Xavi Gonzalvo", "Benoit Dherin" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71192" } }, { "uid": "icml-2026-oral-154", "conference": "ICML", "year": 2026, "track": "Oral", "title": "GoodDiffusion: Proactive Copyright Protection for Diffusion Generative Models via Learnable Sample-specific Signatures", "authors": [ "Shixi Qin", "zhiyong yang", "Shilong Bao", "Zitai Wang", "Qianqian Xu", "Qingming Huang" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71188" } }, { "uid": "icml-2026-oral-155", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Focus and Dilution: The Multi-stage Learning Process of Attention", "authors": [ "Zheng-An Chen", "Pengxiao Lin", "Zhi-Qin John Xu", "Tao Luo" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71193" } }, { "uid": "icml-2026-oral-156", "conference": "ICML", "year": 2026, "track": "Oral", "title": "FlashSketch: Sketch-Kernel Co-Design for Fast Sparse Sketching on GPUs", "authors": [ "Rajat Vadiraj Dwaraknath", "Sungyoon Kim", "Mert Pilanci" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71181" } }, { "uid": "icml-2026-oral-157", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Nash Equilibria in Games with Playerwise Concave Coupling Constraints: Existence and Computation", "authors": [ "Philip Jordan", "Maryam Kamgarpour" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71185" } }, { "uid": "icml-2026-oral-158", "conference": "ICML", "year": 2026, "track": "Oral", "title": "CVE-Factory: Scaling Expert-Level Agentic Tasks for Code Security Vulnerability", "authors": [ "Xianzhen Luo", "Jingyuan Zhang", "Shiqi Zhou", "JinYang Huang", "Chuan Xiao", "Qingfu Zhu", "Zhiyuan Ma", "YUE XING", "Yang Yue", "WencongZeng", "Wanxiang Che" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71173" } }, { "uid": "icml-2026-oral-159", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Orthogonal Concept Erasure for Diffusion Models", "authors": [ "Yuhao Sun", "Lingyun Yu", "Hao-Xiang Xu", "Fengyuan Miao", "Zhuoer Xu", "Hongtao Xie" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71189" } }, { "uid": "icml-2026-oral-160", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Prescriptive Scaling Reveals the Evolution of Language Model Capabilities", "authors": [ "Hanlin Zhang", "Jikai Jin", "Vasilis Syrgkanis", "Sham Kakade" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71169" } }, { "uid": "icml-2026-oral-161", "conference": "ICML", "year": 2026, "track": "Oral", "title": "On the Limits of LLM Adaptability: Impact of LLM Pre-Training on Annotation Task Performance", "authors": [ "Etienne Casanova", "Rafal Kocielnik", "R. Michael Alvarez" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71177" } }, { "uid": "icml-2026-oral-162", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Privacy-Aware Video Anomaly Detection: Guided Orthogonal Projection and a Comprehensive Evaluation Framework", "authors": [ "Wenxiang Diao", "Lei Wang", "Andrew Busch", "Jun Zhou", "Yongsheng Gao" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71190" } }, { "uid": "icml-2026-oral-163", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Procedural Pretraining: Warming Up Language Models with Abstract Data", "authors": [ "Liangze Jiang", "Zachary Shinnick", "Anton Hengel", "Hemanth Saratchandran", "Damien Teney" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71170" } }, { "uid": "icml-2026-oral-164", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Towards Long-Horizon Interpretability: Efficient and Faithful Multi-Token Attribution for Reasoning LLMs", "authors": [ "Wenbo Pan", "Zhichao Liu", "Xianlong Wang", "Yu Haining", "Xiaohua Jia" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71178" } }, { "uid": "icml-2026-oral-165", "conference": "ICML", "year": 2026, "track": "Oral", "title": "Which Algorithms Can Graph Neural Networks Learn?", "authors": [ "Solveig Wittig", "Antonis Vasileiou", "Robert R. Nerem", "Timo Stoll", "Floris Geerts", "Yusu Wang", "Christopher Morris" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71194" } }, { "uid": "icml-2026-oral-166", "conference": "ICML", "year": 2026, "track": "Oral", "title": "What Preferences Can—and Cannot—Predict in Multi-Agent Online Learning", "authors": [ "Omar Abbadi", "Rida Laraki", "Panayotis Mertikopoulos" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71186" } }, { "uid": "icml-2026-oral-167", "conference": "ICML", "year": 2026, "track": "Oral", "title": "OMAC: A Holistic Optimization Framework for LLM-Based Multi-Agent Collaboration", "authors": [ "Shijun Li", "Hilaf Hasson", "Joydeep Ghosh" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71174" } }, { "uid": "icml-2026-oral-168", "conference": "ICML", "year": 2026, "track": "Oral", "title": "SoftJAX & SoftTorch: Empowering Automatic Differentiation Libraries with Informative Gradients", "authors": [ "Anselm Paulus", "Andreas René Geist", "Vit Musil", "Sebastian Hoffmann", "Georg Martius" ], "affiliation": "", "summary": "", "links": { "openreview": "", "arxiv": "", "detail": "https://icml.cc/virtual/2026/oral/71182" } }, { "uid": "icml-2026-spotlight-1", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "Revenue Efficiency of Correlated Equilibria in First Price Auctions", "authors": [ "Anders Bo Ipsen" ], "affiliation": "Aarhus U", "summary": "Analyzes the revenue efficiency of correlated equilibria in first-price auctions.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-2", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "DecodeShare: Tracing the Shared Pathways of LLM Decode-Time Decisions", "authors": [ "Bruce Shao" ], "affiliation": "Wake Forest U", "summary": "An analysis method that traces and interprets the internal pathways shared by LLM decode-time decisions.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-3", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "Online Conformal Prediction via Universal Portfolio Algorithms", "authors": [ "Tuo Liu" ], "affiliation": "UPenn", "summary": "Constructs online conformal prediction using universal portfolio algorithms.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-4", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "Optimal Decision-Making Based on Prediction Sets", "authors": [ "Yuting Wang" ], "affiliation": "UPenn", "summary": "Derives optimal decision rules from prediction sets.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-5", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "Theory-Level Autoformalization: From Isolated Statements to Unified Formal Knowledge Bases", "authors": [ "Min et al." ], "affiliation": "UPenn", "summary": "Goes beyond statement-level autoformalization to build a unified, theory-level formal knowledge base.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-6", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "When to Trust the Cheap Check: Weak and Strong Verification for Reasoning", "authors": [ "Shayan Kiyani" ], "affiliation": "UPenn", "summary": "A strategy combining cheap weak checks with expensive strong checks for verifying reasoning.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-7", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "Fair Classification with Efficient and Post-hoc Controllable Fairness-Accuracy Trade-off", "authors": [ "Maaya Sakata" ], "affiliation": "U Tsukuba / RIKEN AIP", "summary": "A fair-classification method whose fairness-accuracy trade-off is efficiently and post-hoc controllable.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-8", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "SVRG and Beyond via Posterior Correction", "authors": [ "Nico Daheim" ], "affiliation": "TU Darmstadt / hessian.AI", "summary": "Reinterprets and extends SVRG and its variants from a posterior-correction perspective.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "icml-2026-spotlight-9", "conference": "ICML", "year": 2026, "track": "Spotlight", "title": "Position: No Retroactive Cure for Infringement during Training", "authors": [ "Satoru Utsunomiya" ], "affiliation": "U Tokyo", "summary": "Position paper: IP infringement during the training stage cannot be cured retroactively.", "links": { "openreview": "", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-1", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Generalized Linear Mode Connectivity for Transformers", "authors": [ "Alexander Theus", "Alessandro Cabodi", "Sotiris Anagnostidis", "Antonio Orvieto", "Sidak Pal Singh", "Valentina Boeva" ], "affiliation": "ETHZ - ETH Zurich", "summary": "We propose a unified framework for model merging that leverages multiple symmetry classes to enable low- and zero-loss interpolation between independently trained Transformer models, including Vision Transformers and GPT-2.", "links": { "openreview": "https://openreview.net/forum?id=KurYdcCbjv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-2", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Deep Compositional Phase Diffusion for Long Motion Sequence Generation", "authors": [ "Ho Yin Au", "Jie Chen", "Junkun Jiang", "Jingyu Xiang" ], "affiliation": "Hong Kong Baptist University", "summary": "The proposed Compositional Phase Diffusion framework consistently generates semantically aligned multi-clip motion with smooth transitions by using latent-phase diffusion modules (SPDM and TPDM) to preserve phase continuity and enable inbetweening.", "links": { "openreview": "https://openreview.net/forum?id=jzPQRbGkAq", "arxiv": "https://arxiv.org/abs/2510.14427", "detail": "" } }, { "uid": "neurips-2025-oral-3", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "GnnXemplar: Exemplars to Explanations - Natural Language Rules for Global GNN Interpretability", "authors": [ "Burouj Armgaan", "Eshan Jain", "Harsh Pandey", "Mahesh Chandran", "Sayan Ranu" ], "affiliation": "Indian Institute of Technology Delhi", "summary": "We generate global text-based explanations using representative nodes (exemplars) in the embedding space. The exemplars are selected via coverage maximization, and their signatures are explained using natural language rules from a self-refining LLM.", "links": { "openreview": "https://openreview.net/forum?id=eafIjoZAHm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-4", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "RAG4GFM: Bridging Knowledge Gaps in Graph Foundation Models through Graph Retrieval Augmented Generation", "authors": [ "Xingliang Wang", "Zemin Liu", "Junxiao Han", "Shuiguang Deng" ], "affiliation": "Zhejiang University", "summary": "Graph Foundation Models (GFMs) have demonstrated remarkable potential across graph learning tasks but face significant challenges in knowledge updating and reasoning faithfulness. To address these issues, we introduce the Retrieval-Augmented Generation (RAG) paradigm for GFMs, which leverages graph knowledge retrieval.", "links": { "openreview": "https://openreview.net/forum?id=tirl2l9oKg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-5", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Agnostic Active Learning Is Always Better Than Passive Learning", "authors": [ "Steve Hanneke" ], "affiliation": "Purdue University", "summary": "We prove that for every concept class, the optimal query complexity of agnostic active learning is strictly smaller than the sample complexity of agnostic passive learning.", "links": { "openreview": "https://openreview.net/forum?id=XPe55Uffd7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-6", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Learning Linear Attention in Polynomial Time", "authors": [ "Morris Yau", "Ekin Akyürek", "Jiayuan Mao", "Joshua B. Tenenbaum", "Stefanie Jegelka", "Jacob Andreas" ], "affiliation": "Massachusetts Institute of Technology", "summary": "We develop algorithms that are guaranteed to PAC learn transformers.", "links": { "openreview": "https://openreview.net/forum?id=QN0E0KX2LM", "arxiv": "https://arxiv.org/abs/2410.10101", "detail": "" } }, { "uid": "neurips-2025-oral-7", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Optimal Mistake Bounds for Transductive Online Learning", "authors": [ "Zachary Chase", "Steve Hanneke", "Shay Moran", "Jonathan Shafer" ], "affiliation": "University of California, San Diego", "summary": "We resolve a 30-year-old open problem concerning the power of unlabeled data in online learning by tightly quantifying the gap between transductive and standard online learning. We prove that for every concept class $\\mathcal{H}$ with Littlestone dimension $d$, the transductive mistake bound is at least $\\Omega(\\sqrt{d})$.", "links": { "openreview": "https://openreview.net/forum?id=EoebmBe9fG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-8", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "State Entropy Regularization for Robust Reinforcement Learning", "authors": [ "Yonatan Ashlag", "Uri Koren", "Mirco Mutti", "Esther Derman", "Pierre-Luc Bacon", "Shie Mannor" ], "affiliation": "Technion - Israel Institute of Technology, Technion", "summary": "State entropy regularization has empirically shown better exploration and sample complexity in reinforcement learning (RL). However, its theoretical guarantees have not been studied.", "links": { "openreview": "https://openreview.net/forum?id=rtG7n93Ru8", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-9", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticity", "authors": [ "Quentin Bertrand", "Anne Gagneux", "Mathurin Massias", "Rémi Emonet" ], "affiliation": "INRIA", "summary": "We leverage the closed-form formulation of flow matching to understand its generalization", "links": { "openreview": "https://openreview.net/forum?id=kVz9uvqUna", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-10", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Why Diffusion Models Don’t Memorize: The Role of Implicit Dynamical Regularization in Training", "authors": [ "Tony Bonnaire", "Raphaël Urfin", "Giulio Biroli", "Marc Mezard" ], "affiliation": "Université Paris-Saclay", "summary": "Implicit dynamical regularization during training gives diffusion models a generalization window that widens with the training set size, so stopping within this window prevents memorization.", "links": { "openreview": "https://openreview.net/forum?id=BSZqpqgqM0", "arxiv": "https://arxiv.org/abs/2505.17638", "detail": "" } }, { "uid": "neurips-2025-oral-11", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Adjoint Schrödinger Bridge Sampler", "authors": [ "Guan-Horng Liu", "Jaemoo Choi", "Yongxin Chen", "Benjamin Kurt Miller", "Ricky T. Q. Chen" ], "affiliation": "FAIR, Meta AI", "summary": "Computational methods for learning to sample from the Boltzmann distribution—where the target distribution is known only up to an unnormalized energy function—have advanced significantly recently. Due to the lack of explicit target samples, however, prior diffusion-based methods, known as _diffusion samplers_, often require importance-weighted esti", "links": { "openreview": "https://openreview.net/forum?id=rMhQBlhh4c", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-12", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Breaking the Performance Ceiling in Reinforcement Learning requires Inference Strategies", "authors": [ "Felix Chalumeau", "Daniel Rajaonarivonivelomanantsoa", "Ruan John de Kock", "Juan Claude Formanek", "Sasha Abramowitz", "Omayma Mahjoub", "Wiem Khlifi", "Simon Verster Du Toit", "Louay Ben Nessir", "Refiloe Shabe", "Arnol Manuel Fokam", "Siddarth Singh", "Ulrich Armel Mbou Sob", "Arnu Pretorius" ], "affiliation": "InstaDeep", "summary": "Using search strategies at inference-time can provide massive performance boost on numerous complex reinforcement learning tasks, within only a couple seconds of execution time.", "links": { "openreview": "https://openreview.net/forum?id=RxkCwOKVKa", "arxiv": "https://arxiv.org/abs/2505.21236", "detail": "" } }, { "uid": "neurips-2025-oral-13", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "High-Dimensional Calibration from Swap Regret", "authors": [ "Maxwell Fishelson", "Noah Golowich", "Mehryar Mohri", "Jon Schneider" ], "affiliation": "Massachusetts Institute of Technology", "summary": "An algorithm for calibrating forecasts of high-dimensional outcomes.", "links": { "openreview": "https://openreview.net/forum?id=UVDihUz0iT", "arxiv": "https://arxiv.org/abs/2505.21460", "detail": "" } }, { "uid": "neurips-2025-oral-14", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "In Search of Adam’s Secret Sauce", "authors": [ "Antonio Orvieto", "Robert M. Gower" ], "affiliation": "ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems, Tübingen AI Center, Tübingen, Germany", "summary": "Adam with equal betas works well, and its form can be simplified for further insights", "links": { "openreview": "https://openreview.net/forum?id=CH72XyZs4y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-15", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "An Optimized Franz-Parisi Criterion and its Equivalence with SQ Lower Bounds", "authors": [ "Siyu Chen", "Theodor Misiakiewicz", "Ilias Zadik", "Peiyuan Zhang" ], "affiliation": "Yale University", "summary": "We propose a refined Franz-Parisi criterion, and show that it is equivalent to Statistical Query lower bounds under a mild, verifiable assumption satisfied by a broad class of statistical models.", "links": { "openreview": "https://openreview.net/forum?id=U8BwT6Rmw4", "arxiv": "https://arxiv.org/abs/2506.06259", "detail": "" } }, { "uid": "neurips-2025-oral-16", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "MaxSup: Overcoming Representation Collapse in Label Smoothing", "authors": [ "Yuxuan Zhou", "Heng Li", "Zhi-Qi Cheng", "Xudong Yan", "Yifei Dong", "Mario Fritz", "Margret Keuper" ], "affiliation": "Baidu", "summary": "Label Smoothing (LS) is widely adopted to reduce overconfidence in neural network predictions and improve generalization. Despite these benefits, recent studies reveal two critical issues with LS.", "links": { "openreview": "https://openreview.net/forum?id=efOq8wHH9o", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-17", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Memory Mosaics at scale", "authors": [ "Jianyu Zhang", "Leon Bottou" ], "affiliation": "New York University", "summary": "Memory Mosaics, networks of associative memories, have demonstrated appealing compositional and in-context learning capabilities on medium-scale networks (GPT-2 scale) and synthetic small datasets. This work shows that these favorable properties remain when we scale memory mosaics to large language model sizes (llama-8B scale) and real-world datase", "links": { "openreview": "https://openreview.net/forum?id=IfD2MKTmWv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-18", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "The emergence of sparse attention: impact of data distribution and benefits of repetition", "authors": [ "Nicolas Zucchet", "Francesco D'Angelo", "Andrew Kyle Lampinen", "Stephanie C.Y. Chan" ], "affiliation": "ETHZ - ETH Zurich", "summary": "We show that learning sparse attention is prone to emerging behaviors during training, and study (theoretically and empirically) how data and model design influence emergence speed.", "links": { "openreview": "https://openreview.net/forum?id=jMhRbV47pS", "arxiv": "https://arxiv.org/abs/2505.17863", "detail": "" } }, { "uid": "neurips-2025-oral-19", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "ControlFusion: A Controllable Image Fusion Network with Language-Vision Degradation Prompts", "authors": [ "Linfeng Tang", "Yeda Wang", "Zhanchuan Cai", "Junjun Jiang", "Jiayi Ma" ], "affiliation": "Wuhan University", "summary": "Current image fusion methods struggle with real-world composite degradations and lack the flexibility to accommodate user-specific needs. To address this, we propose ControlFusion, a controllable fusion network guided by language-vision prompts that adaptively mitigates composite degradations.", "links": { "openreview": "https://openreview.net/forum?id=aLhA7AYLLR", "arxiv": "https://arxiv.org/abs/2503.23356", "detail": "" } }, { "uid": "neurips-2025-oral-20", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Identifiability of Deep Polynomial Neural Networks", "authors": [ "Konstantin Usevich", "Ricardo Augusto Borsoi", "Clara Dérand", "Marianne Clausel" ], "affiliation": "CNRS", "summary": "We provide necessary and sufficient conditions for the identifiability deep polinomial neural networks.", "links": { "openreview": "https://openreview.net/forum?id=MrUsZfQ9pC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-21", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Understanding and Mitigating Numerical Sources of Nondeterminism in LLM Inference", "authors": [ "Jiayi Yuan", "Hao Li", "Xinheng Ding", "Wenya Xie", "Yu-Jhe Li", "Wentian Zhao", "Kun Wan", "Jing Shi", "Xia Hu", "Zirui Liu" ], "affiliation": "Rice University", "summary": "This paper demonstrates that low precision causes non-reproducible LLM inference across different setups, proposing a hybrid-precision method, LayerCast, that computes in FP32 to achieve determinism while saving memory.", "links": { "openreview": "https://openreview.net/forum?id=Q3qAsZAEZw", "arxiv": "https://arxiv.org/abs/2506.09501", "detail": "" } }, { "uid": "neurips-2025-oral-22", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "PRIMT: Preference-based Reinforcement Learning with Multimodal Feedback and Trajectory Synthesis from Foundation Models", "authors": [ "Ruiqi Wang", "Dezhong Zhao", "Ziqin Yuan", "Tianyu Shao", "Guohua Chen", "Dominic Kao", "Sungeun Hong", "Byung-Cheol Min" ], "affiliation": "Purdue University", "summary": "Preference-based reinforcement learning (PbRL) has emerged as a promising paradigm for teaching robots complex behaviors without reward engineering. However, its effectiveness is often limited by two critical challenges: the reliance on extensive human input and the inherent difficulties in resolving query ambiguity and credit assignment during rew", "links": { "openreview": "https://openreview.net/forum?id=4xvE6Iy77Y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-23", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders", "authors": [ "David Chanin", "James Wilken-Smith", "Tomáš Dulka", "Hardik Bhatnagar", "Satvik Golechha", "Joseph Isaac Bloom" ], "affiliation": "MATS", "summary": "Sparse Autoencoders (SAEs) aim to decompose the activation space of large language models (LLMs) into human-interpretable latent directions or features. As we increase the number of features in the SAE, hierarchical features tend to split into finer features (“math” may split into “algebra”, “geometry”, etc.), a phenomenon referred to as feature sp", "links": { "openreview": "https://openreview.net/forum?id=R73ybUciQF", "arxiv": "https://arxiv.org/abs/2409.14507", "detail": "" } }, { "uid": "neurips-2025-oral-24", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "EvoLM: In Search of Lost Training Dynamics for Language Model Reasoning", "authors": [ "Zhenting Qi", "Fan Nie", "Alexandre Alahi", "James Zou", "Himabindu Lakkaraju", "Yilun Du", "Eric P. Xing", "Sham M. Kakade", "Hanlin Zhang" ], "affiliation": "Harvard University", "summary": "Modern language model (LM) training has been divided into multiple stages, making it difficult for downstream developers to evaluate the impact of design choices made at each stage. We present EvoLM, a model suite that enables systematic and transparent analysis of LMs' training dynamics across pre-training, continued pre-training, supervised fine-", "links": { "openreview": "https://openreview.net/forum?id=B6bE2GC71a", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-25", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Analog In-memory Training on General Non-ideal Resistive Elements: The Impact of Response Functions", "authors": [ "Zhaoxian Wu", "Quan Xiao", "Tayfun Gokmen", "Omobayode Fagbohungbe", "Tianyi Chen" ], "affiliation": "Cornell University", "summary": "Leveraging a residual learning framework to support the model training on non-ideal analog in-memory computing hardware", "links": { "openreview": "https://openreview.net/forum?id=WhEPg4mUs6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-26", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Discovering Opinion Intervals from Conflicts in Signed Graphs", "authors": [ "Peter Blohm", "Florian Chen", "Aristides Gionis", "Stefan Neumann" ], "affiliation": "Technische Universität Wien", "summary": "We present a novel problem that allows to infer a small and interpretable set of prevalent opinion ranges in signed graphs, that explain the users' interactions.", "links": { "openreview": "https://openreview.net/forum?id=zJdutIT6vT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-27", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "A Clean Slate for Offline Reinforcement Learning", "authors": [ "Matthew Thomas Jackson", "Uljad Berdica", "Jarek Luca Liesen", "Shimon Whiteson", "Jakob Nicolaus Foerster" ], "affiliation": "Google DeepMind", "summary": "We propose a principled taxonomy, evaluation procedure, and unified algorithm space for offline RL.", "links": { "openreview": "https://openreview.net/forum?id=8P3QNSckMp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-28", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Spectral Perturbation Bounds for Low-Rank Approximation with Applications to Privacy", "authors": [ "Phuc Tran", "Van Vu", "Nisheeth K. Vishnoi" ], "affiliation": "VinUniversity", "summary": "We derive sharp spectral-norm bounds for noisy low-rank approximation, improving prior results by up to $\\sqrt{n}$. Applied to DP-PCA, our method resolves an open problem and matches empirical error via a novel contour bootstrapping technique.", "links": { "openreview": "https://openreview.net/forum?id=F0JzotXYgC", "arxiv": "https://arxiv.org/abs/2510.25670", "detail": "" } }, { "uid": "neurips-2025-oral-29", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Improved Regret Bounds for Gaussian Process Upper Confidence Bound in Bayesian Optimization", "authors": [ "Shogo Iwazaki" ], "affiliation": "MI-6 Ltd.", "summary": "This paper addresses the Bayesian optimization problem (also referred to as the Bayesian setting of the Gaussian process bandit), where the learner seeks to minimize the regret under a function drawn from a known Gaussian process (GP). Under a Mat\\'ern kernel with some extent of smoothness, we show that the Gaussian process upper confidence bound (", "links": { "openreview": "https://openreview.net/forum?id=gxfusMqPIs", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-30", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Auto-Compressing Networks", "authors": [ "Vaggelis Dorovatas", "Georgios Paraskevopoulos", "Alexandros Potamianos" ], "affiliation": "National Technical University of Athens", "summary": "Deep neural networks with short residual connections have demonstrated remarkable success across domains, but increasing depth often introduces computational redundancy without corresponding improvements in representation quality. We introduce Auto-Compressing Networks (ACNs), an architectural variant where additive long feedforward connections fro", "links": { "openreview": "https://openreview.net/forum?id=eIDa6pd9iQ", "arxiv": "https://arxiv.org/abs/2506.09714", "detail": "" } }, { "uid": "neurips-2025-oral-31", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "MokA: Multimodal Low-Rank Adaptation for MLLMs", "authors": [ "Yake Wei", "Yu Miao", "Dongzhan Zhou", "Di Hu" ], "affiliation": "Renmin University of China", "summary": "In this paper, we reveal that most current efficient multimodal fine-tuning methods are hindered by a key limitation: they are directly borrowed from LLMs, often neglecting the intrinsic differences of multimodal scenarios and even affecting the full utilization of all modalities. Inspired by our empirical observation, we argue that unimodal adapta", "links": { "openreview": "https://openreview.net/forum?id=oJ84bedrtM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-32", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Advancing Expert Specialization for Better MoE", "authors": [ "Hongcan Guo", "Haolang Lu", "Guoshun Nan", "Bolun Chu", "Jialin Zhuang", "Yuan Yang", "Wenhao Che", "Xinye Cao", "Sicong Leng", "Qimei Cui", "Xudong Jiang" ], "affiliation": "ByteDance Inc.", "summary": "Our proposed orthogonality and variance losses improve performance in downstream fine-tuning of Mixture-of-Experts models by enhancing expert specificity, addressing expert homogenization caused by load balancing, while maintaining load balance.", "links": { "openreview": "https://openreview.net/forum?id=iydmH9boLb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-33", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "From Condensation to Rank Collapse: A Two-Stage Analysis of Transformer Training Dynamics", "authors": [ "Zheng-An Chen", "Tao Luo" ], "affiliation": "Shanghai Jiaotong University", "summary": "Although transformer-based models have shown exceptional empirical performance, the fundamental principles governing their training dynamics are inadequately characterized beyond configuration-specific studies. Inspired by empirical evidence showing improved reasoning capabilities under small initialization scales in language models, we employ the", "links": { "openreview": "https://openreview.net/forum?id=gm5mkiTGOy", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-34", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Large Language Diffusion Models", "authors": [ "Shen Nie", "Fengqi Zhu", "Zebin You", "Xiaolu Zhang", "Jingyang Ou", "Jun Hu", "JUN ZHOU", "Yankai Lin", "Ji-Rong Wen", "Chongxuan Li" ], "affiliation": "Renmin University of China", "summary": "We present LLaDA, a diffusion language model trained from scratch that is competitive to LLaMA 3 in performance.", "links": { "openreview": "https://openreview.net/forum?id=KnqiC0znVF", "arxiv": "https://arxiv.org/abs/2502.09992", "detail": "" } }, { "uid": "neurips-2025-oral-35", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Boosting Knowledge Utilization in Multimodal Large Language Models via Adaptive Logits Fusion and Attention Reallocation", "authors": [ "Wenbin An", "Jiahao Nie", "Feng Tian", "Haonan Lin", "mingxiang cai", "Yaqiang Wu", "QianYing Wang", "Xiaoqin Zhang", "Shijian Lu" ], "affiliation": "Xi'an Jiaotong University", "summary": "Despite their recent progress, Multimodal Large Language Models (MLLMs) often struggle in knowledge-intensive tasks due to the limited and outdated parametric knowledge acquired during training. Multimodal Retrieval Augmented Generation addresses this issue by retrieving contextual knowledge from external databases, thereby enhancing MLLMs with exp", "links": { "openreview": "https://openreview.net/forum?id=qYkhCah8OZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-36", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Interactive Cross-modal Learning for Text-3D Scene Retrieval", "authors": [ "Yanglin Feng", "Yongxiang Li", "Yuan Sun", "Yang Qin", "Dezhong Peng", "Peng Hu" ], "affiliation": "Sichuan University", "summary": "We propose an Interactive Text-to-3D Scene Retrieval Method to handle inherent query limitations.", "links": { "openreview": "https://openreview.net/forum?id=fohuurA03P", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-37", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Rethinking Joint Maximum Mean Discrepancy for Visual Domain Adaptation", "authors": [ "Wei Wang", "Haifeng Xia", "Chao Huang", "Zhengming Ding", "Cong Wang", "Haojie Li", "Xiaochun Cao" ], "affiliation": "SUN YAT-SEN UNIVERSITY", "summary": "In domain adaption (DA), joint maximum mean discrepancy (JMMD), as a famous distribution-distance metric, aims to measure joint probability distribution difference between the source domain and target domain, while it is still not fully explored and especially hard to be applied into a subspace-learning framework as its empirical estimation involve", "links": { "openreview": "https://openreview.net/forum?id=XoN10bZtR9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-38", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Pan-LUT: Efficient Pan-sharpening via Learnable Look-Up Tables", "authors": [ "Zhongnan Cai", "Yingying Wang", "Hui Zheng", "Panwang Pan", "ZiXu Lin", "Ge Meng", "Chenxin Li", "Chunming He", "Jiaxin Xie", "Yunlong Lin", "Junbin Lu", "Yue Huang", "Xinghao Ding" ], "affiliation": "Xiamen University", "summary": "Recently, deep learning-based pan-sharpening algorithms have achieved notable advancements over traditional methods. However, deep learning-based methods incur substantial computational overhead during inference, especially with large images.", "links": { "openreview": "https://openreview.net/forum?id=OzdAnGHEPx", "arxiv": "https://arxiv.org/abs/2503.23793", "detail": "" } }, { "uid": "neurips-2025-oral-39", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Dynamical Decoupling of Generalization and Overfitting in Large Two-Layer Networks", "authors": [ "Andrea Montanari", "Pierfrancesco Urbani" ], "affiliation": "Stanford University", "summary": "Large neural networks first learn low dimensional feature representation then overfit the data and revert to a kernel regime.", "links": { "openreview": "https://openreview.net/forum?id=ImpizBSKcu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-40", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "1000 Layer Networks for Self-Supervised RL: Scaling Depth Can Enable New Goal-Reaching Capabilities", "authors": [ "Kevin Wang", "Ishaan Javali", "Michał Bortkiewicz", "Tomasz Trzcinski", "Benjamin Eysenbach" ], "affiliation": "Princeton University", "summary": "While most RL methods use shallow MLPs (~2–5 layers), we show that scaling up to 1000-layers for contrastive RL (CRL) can significantly boost performance, ranging from doubling performance to 50x on a diverse suite of robotic tasks.", "links": { "openreview": "https://openreview.net/forum?id=s0JVsx3bx1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-41", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Depth-Bounds for Neural Networks via the Braid Arrangement", "authors": [ "Moritz Leo Grillo", "Christoph Hertrich", "Georg Loho" ], "affiliation": "Max-Planck Institute", "summary": "We contribute towards resolving the open question of how many hidden layers are required in ReLU networks for exactly representing all continuous and piecewise linear functions on $\\mathbb{R}^d$. While the question has been resolved in special cases, the best known lower bound in general is still 2.", "links": { "openreview": "https://openreview.net/forum?id=XO9fhSZkBh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-42", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Tighter CMI-Based Generalization Bounds via Stochastic Projection and Quantization", "authors": [ "Milad Sefidgaran", "Kimia Nadjahi", "Abdellatif Zaidi" ], "affiliation": "Huawei Paris Research Center", "summary": "In this paper, we leverage stochastic projection and lossy compression to establish new conditional mutual information (CMI) bounds on the generalization error of statistical learning algorithms. It is shown that these bounds are generally tighter than the existing ones.", "links": { "openreview": "https://openreview.net/forum?id=VYLdKb5dzO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-43", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "A Snapshot of Influence: A Local Data Attribution Framework for Online Reinforcement Learning", "authors": [ "Yuzheng Hu", "Fan Wu", "Haotian Ye", "David Forsyth", "James Zou", "Nan Jiang", "Jiaqi W. Ma", "Han Zhao" ], "affiliation": "University of Illinois at Urbana-Champaign", "summary": "We propose the first framework of data attribution for online RL.", "links": { "openreview": "https://openreview.net/forum?id=sYK4yPDuT1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-44", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "High-dimensional neuronal activity from low-dimensional latent dynamics: a solvable model", "authors": [ "Valentin Schmutz", "Ali Haydaroğlu", "Shuqi Wang", "Yixiao Feng", "Matteo Carandini", "Kenneth D. Harris" ], "affiliation": "University College London, University of London", "summary": "We show that high-dimensional neural activity can arise from low-dimensional latent dynamics, both in RNNs and in the brain.", "links": { "openreview": "https://openreview.net/forum?id=cGks3s79hW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-45", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Adaptive Surrogate Gradients for Sequential Reinforcement Learning in Spiking Neural Networks", "authors": [ "Korneel Van den Berghe", "Stein Stroobants", "Vijay Janapa Reddi", "Guido De Croon" ], "affiliation": "Delft University of Technology", "summary": "We improve training of spiking neural networks for energy-efficient robotic control by analyzing surrogate gradient slopes and introducing a privileged policy-guided method, achieving a 2.1× performance boost and strong real-world results.", "links": { "openreview": "https://openreview.net/forum?id=oGmROC4e4W", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-46", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Class-wise Balancing Data Replay for Federated Class-Incremental Learning", "authors": [ "Zhuang Qi", "Ying-Peng Tang", "Lei Meng", "Han Yu", "Xiaoxiao Li", "Xiangxu Meng" ], "affiliation": "Shandong University", "summary": "Federated Class Incremental Learning (FCIL) aims to collaboratively process continuously increasing incoming tasks across multiple clients. Among various approaches, data replay has become a promising solution, which can alleviate forgetting by reintroducing representative samples from previous tasks.", "links": { "openreview": "https://openreview.net/forum?id=aUAG1WS7J2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-47", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Task-Optimized Convolutional Recurrent Networks Align with Tactile Processing in the Rodent Brain", "authors": [ "Trinity Chung", "Yuchen Shen", "Nathan Kong", "Aran Nayebi" ], "affiliation": "", "summary": "Task-optimized convolutional recurrent neural networks trained on realistic tactile inputs align with rodent somatosensory data, suggesting brain tactile processing uses temporally-precise representations shaped by categorization-driven optimization.", "links": { "openreview": "https://openreview.net/forum?id=m7MD0sa8Re", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-48", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "ElasticMM: Efficient Multimodal LLMs Serving with Elastic Multimodal Parallelism", "authors": [ "Zedong Liu", "Shenggan Cheng", "Guangming Tan", "Yang You", "Dingwen Tao" ], "affiliation": "University of Electronic Science and Technology of China", "summary": "Multimodal large language models (MLLMs) extend LLMs to handle images, videos, and audio by incorporating feature extractors and projection modules. However, these additional components—combined with complex inference pipelines and heterogeneous workloads—introduce significant inference overhead.", "links": { "openreview": "https://openreview.net/forum?id=Zd6VyjmN1S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-49", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Dynamical Low-Rank Compression of Neural Networks with Robustness under Adversarial Attacks", "authors": [ "Steffen Schotthöfer", "H. Lexie Yang", "Stefan Schnake" ], "affiliation": "Oak Ridge National Laboratory", "summary": "Deployment of neural networks on resource-constrained devices demands models that are both compact and robust to adversarial inputs. However, compression and adversarial robustness often conflict.", "links": { "openreview": "https://openreview.net/forum?id=7AwFJzgIUW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-50", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "QoQ-Med: Building Multimodal Clinical Foundation Models with Domain-Aware GRPO Training", "authors": [ "Wei Dai", "Peilin Chen", "Chanakya Ekbote", "Paul Pu Liang" ], "affiliation": "Massachusetts Institute of Technology", "summary": "We built a generalist clinical foundation model across both time series and vision modalities with a novel RL training algorithm named DRPO.", "links": { "openreview": "https://openreview.net/forum?id=ZwCVFBFUFb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-51", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free", "authors": [ "Zihan Qiu", "Zekun Wang", "Bo Zheng", "Zeyu Huang", "Kaiyue Wen", "Songlin Yang", "Rui Men", "Le Yu", "Fei Huang", "Suozhi Huang", "Dayiheng Liu", "Jingren Zhou", "Junyang Lin" ], "affiliation": "Alibaba Group", "summary": "We find applying a query-dependent head-specific sigmoid gate after the Scaled Dot-Product Attention (SDPA) consistently improves performance, improves scaling properties and mitigates the `massive activation' and `attention sink'.", "links": { "openreview": "https://openreview.net/forum?id=1b7whO4SfY", "arxiv": "https://arxiv.org/abs/2505.06708", "detail": "" } }, { "uid": "neurips-2025-oral-52", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Learning long range dependencies through time reversal symmetry breaking", "authors": [ "Guillaume Pourcel", "Maxence Ernoult" ], "affiliation": "INRIA", "summary": "We propose a backward-mode AD proxy using only forward passes applying to Hamiltonian recurrent units and stacks thereof (namely, SSMs) with theoretical guarantees and experimental evidence", "links": { "openreview": "https://openreview.net/forum?id=w1ihNiIBOc", "arxiv": "https://arxiv.org/abs/2506.05259", "detail": "" } }, { "uid": "neurips-2025-oral-53", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "FuXi-Ocean: A Global Ocean Forecasting System with Sub-Daily Resolution", "authors": [ "Qiusheng Huang", "Yuan Niu", "Xiaohui Zhong", "AnboyuGuo", "Lei Chen", "dianjun zhang", "Xuefeng Zhang", "Hao Li" ], "affiliation": "Fudan University", "summary": "The first deep-learning model for sub-daily ocean forecast", "links": { "openreview": "https://openreview.net/forum?id=WJujF9An5L", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-54", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Superposition Yields Robust Neural Scaling", "authors": [ "Yizhou Liu", "Ziming Liu", "Jeff Gore" ], "affiliation": "Massachusetts Institute of Technology", "summary": "Neural scaling law in LLMs is explained through representation interference due to superposition", "links": { "openreview": "https://openreview.net/forum?id=knPz7gtjPW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-55", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression", "authors": [ "Tom Burgert", "Oliver Stoll", "Paolo Rota", "Begüm Demir" ], "affiliation": "Technische Universität Berlin", "summary": "We revisit the texture bias hypothesis in CNNs by proposing a domain-agnostic suppression protocol, finding that contrary to prior claims, CNNs primarily rely on local shape instead of texture features.", "links": { "openreview": "https://openreview.net/forum?id=i5WnXNjwbR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-56", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "On Linear Mode Connectivity of Mixture-of-Experts Architectures", "authors": [ "Viet-Hoang Tran", "Van-Hoan Trinh", "Khanh Vinh Bui", "Tan Minh Nguyen" ], "affiliation": "National University of Singapore", "summary": "We investigate Linear Mode Connectivity (LMC) in Mixture-of-Experts (MoE) architectures by analyzing their underlying permutation symmetries and proposing expert-matching algorithms that align independently trained MoEs to reveal LMC.", "links": { "openreview": "https://openreview.net/forum?id=RF3miSqdXa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-57", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "OpenHOI: Open-World Hand-Object Interaction Synthesis with Multimodal Large Language Model", "authors": [ "Zhenhao Zhang", "Ye Shi", "Lingxiao Yang", "Suting Ni", "Qi Ye", "Jingya Wang" ], "affiliation": "", "summary": "Introduce the first Open-World Hand-Object Interaction (HOI) Synthesis framework that can generate Long-horizon HOI sequences of Unseen Objects from Open-vocabulary instruction with 3D Multimodal Large Language Model.", "links": { "openreview": "https://openreview.net/forum?id=0biUwyjKkm", "arxiv": "https://arxiv.org/abs/2505.18947", "detail": "" } }, { "uid": "neurips-2025-oral-58", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Think", "authors": [ "Ge Wu", "Shen Zhang", "Ruijing Shi", "Shanghua Gao", "Zhenyuan Chen", "Lei Wang", "Zhaowei Chen", "Hongcheng Gao", "Yao Tang", "jian Yang", "Ming-Ming Cheng", "Xiang Li" ], "affiliation": "Nankai University", "summary": "REPA and its variants effectively mitigate training challenges in diffusion models by incorporating external visual representations from pretrained models, through alignment between the noisy hidden projections of denoising networks and foundational clean image representations. We argue that the external alignment, which is absent during the entire", "links": { "openreview": "https://openreview.net/forum?id=koEALFNBj1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-59", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Dynam3D: Dynamic Layered 3D Tokens Empower VLM for Vision-and-Language Navigation", "authors": [ "Zihan Wang", "Seungjun Lee", "Gim Hee Lee" ], "affiliation": "National University of Singapore", "summary": "A multi-level patch-instance-zone 3D representation model with a hierarchical dynamic online update strategy for embodied navigation.", "links": { "openreview": "https://openreview.net/forum?id=s6k9l5yX8e", "arxiv": "https://arxiv.org/abs/2505.11383", "detail": "" } }, { "uid": "neurips-2025-oral-60", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Learning (Approximately) Equivariant Networks via Constrained Optimization", "authors": [ "Andrei Manolache", "Luiz F. O. Chamon", "Mathias Niepert" ], "affiliation": "Universität Stuttgart", "summary": "We introduce Adaptive Constrained Equivariance, a homotopy-inspired constrained optimization apprach for training equivariant neural networks.", "links": { "openreview": "https://openreview.net/forum?id=NM4emKloy6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-61", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "SAGE: A Unified Framework for Generalizable Object State Recognition with State-Action Graph Embedding", "authors": [ "Yuan Zang", "Zitian Tang", "Junho Cho", "Jaewook Yoo", "Chen Sun" ], "affiliation": "Brown University", "summary": "We propose a unified and generalizable framework for object state recognition in video data.", "links": { "openreview": "https://openreview.net/forum?id=jRXgRC6fu7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-62", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?", "authors": [ "Yang Yue", "Zhiqi Chen", "Rui Lu", "Andrew Zhao", "Zhaokai Wang", "Yang Yue", "Shiji Song", "Gao Huang" ], "affiliation": "", "summary": "We systematically examine the current state of RLVR and surprisingly find that it does not elicit fundamentally new reasoning patterns—revealing a gap between the potential of RL and the actual impact of current RLVR methods.", "links": { "openreview": "https://openreview.net/forum?id=4OsgYD7em5", "arxiv": "https://arxiv.org/abs/2504.13837", "detail": "" } }, { "uid": "neurips-2025-oral-63", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Learning to Learn with Contrastive Meta-Objective", "authors": [ "Shiguang Wu", "Yaqing Wang", "Yatao Bian", "Quanming Yao" ], "affiliation": "Tsinghua University, Tsinghua University", "summary": "Meta-learning enables learning systems to adapt quickly to new tasks, similar to humans. Different meta-learning approaches all work under/with the mini-batch episodic training framework.", "links": { "openreview": "https://openreview.net/forum?id=s6YHno8Ke3", "arxiv": "https://arxiv.org/abs/2410.05975", "detail": "" } }, { "uid": "neurips-2025-oral-64", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "KVzip: Query-Agnostic KV Cache Compression with Context Reconstruction", "authors": [ "Jang-Hyun Kim", "Jinuk Kim", "Sangwoo Kwon", "Jae W. Lee", "Sangdoo Yun", "Hyun Oh Song" ], "affiliation": "Apple", "summary": "We propose a novel query-agnostic KV cache eviction method for multi-query scenario.", "links": { "openreview": "https://openreview.net/forum?id=JFygzwx8SJ", "arxiv": "https://arxiv.org/abs/2505.23416", "detail": "" } }, { "uid": "neurips-2025-oral-65", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "HyperET: Efficient Training in Hyperbolic Space for Multi-modal Large Language Models", "authors": [ "Zelin Peng", "Zhengqin Xu", "Qingyang Liu", "Xiaokang Yang", "Wei Shen" ], "affiliation": "Shanghai Jiaotong University", "summary": "Multi-modal large language models (MLLMs) have emerged as a transformative approach for aligning visual and textual understanding. They typically require extremely high computational resources (e.g., thousands of GPUs) for training to achieve cross-modal alignment at multi-granularity levels.", "links": { "openreview": "https://openreview.net/forum?id=NM8Apk61NA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-66", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "SAVVY: Spatial Awareness via Audio-Visual LLMs through Seeing and Hearing", "authors": [ "Mingfei Chen", "Zijun Cui", "Xiulong Liu", "Jinlin Xiang", "Caleb Zheng", "Jingyuan Li", "Eli Shlizerman" ], "affiliation": "University of Washington", "summary": "A novel 3D audio-visual QA benchmark and training-free spatial reasoning pipeline for Audio-Visual LLMs", "links": { "openreview": "https://openreview.net/forum?id=zwCb9cKHpd", "arxiv": "https://arxiv.org/abs/2506.05414", "detail": "" } }, { "uid": "neurips-2025-oral-67", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "A multiscale analysis of mean-field transformers in the moderate interaction regime", "authors": [ "Giuseppe Bruno", "Federico Pasqualotto", "Andrea Agazzi" ], "affiliation": "Universität Bern", "summary": "In this paper, we study the evolution of tokens through the depth of encoder-only transformer models at inference time by modeling them as a system of particles interacting in a mean-field way and studying the corresponding dynamics. More specifically, we consider this problem in the moderate interaction regime, where the number $N$ of tokens is la", "links": { "openreview": "https://openreview.net/forum?id=WCRPgBpbcA", "arxiv": "https://arxiv.org/abs/2509.25040", "detail": "" } }, { "uid": "neurips-2025-oral-68", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Exploring Diffusion Transformer Designs via Grafting", "authors": [ "Keshigeyan Chandrasegaran", "Michael Poli", "Daniel Y Fu", "Dongjun Kim", "Lea M. Hadzic", "Manling Li", "Agrim Gupta", "Stefano Massaroli", "Azalia Mirhoseini", "Juan Carlos Niebles", "Stefano Ermon", "Li Fei-Fei" ], "affiliation": "Stanford University", "summary": "We propose grafting, a simple approach to materialize new architectures by editing pretrained diffusion transformers. It enables architectural exploration under small compute budgets.", "links": { "openreview": "https://openreview.net/forum?id=CaSQgef484", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-69", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Generalized Gradient Norm Clipping & Non-Euclidean $(L_0,L_1)$-Smoothness", "authors": [ "Thomas Pethick", "Wanyun Xie", "Mete Erdogan", "Kimon Antonakopoulos", "Tony Silveti-Falls", "Volkan Cevher" ], "affiliation": "Swiss Federal Institute of Technology Lausanne", "summary": "This work introduces a hybrid non-Euclidean optimization method which generalizes gradient norm clipping by combining steepest descent and conditional gradient approaches. The method achieves the best of both worlds by establishing a descent property under a generalized notion of ($L_0$,$L_1$)-smoothness.", "links": { "openreview": "https://openreview.net/forum?id=rMdf8jhLR7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-70", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Rethinking Multimodal Learning from the Perspective of Mitigating Classification Ability Disproportion", "authors": [ "Qing-Yuan Jiang", "Longfei Huang", "Yang Yang" ], "affiliation": "Nanjing University of Science and Technology", "summary": "Multimodal learning (MML) is significantly constrained by modality imbalance, leading to suboptimal performance in practice. While existing approaches primarily focus on balancing the learning of different modalities to address this issue, they fundamentally overlook the inherent disproportion in model classification ability, which serves as the pr", "links": { "openreview": "https://openreview.net/forum?id=Q6IyUpBmrG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-71", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "TransferTraj: A Vehicle Trajectory Learning Model for Region and Task Transferability", "authors": [ "Tonglong Wei", "Yan Lin", "Zeyu Zhou", "Haomin Wen", "Jilin Hu", "Shengnan Guo", "Youfang Lin", "Gao Cong", "Huaiyu Wan" ], "affiliation": "Beijing Jiaotong University", "summary": "Propose a vehicle trajectory learning model capable of transferring across regions and tasks without retraining.", "links": { "openreview": "https://openreview.net/forum?id=XQ87Vo9GIz", "arxiv": "https://arxiv.org/abs/2505.12672", "detail": "" } }, { "uid": "neurips-2025-oral-72", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "PhySense: Sensor Placement Optimization for Accurate Physics Sensing", "authors": [ "Yuezhou Ma", "Haixu Wu", "Hang Zhou", "Huikun Weng", "Jianmin Wang", "Mingsheng Long" ], "affiliation": "Tsinghua University, Tsinghua University", "summary": "We propose a synergistic two-stage framework that learns to jointly reconstruct physical fields and to optimize sensor placements with theoretical guarantees, both aiming for accurate physics sensing.", "links": { "openreview": "https://openreview.net/forum?id=zIzZxDsNNP", "arxiv": "https://arxiv.org/abs/2505.18190", "detail": "" } }, { "uid": "neurips-2025-oral-73", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "InfinityStar: Unified Spacetime AutoRegressive Modeling for Visual Generation", "authors": [ "Jinlai Liu", "Jian Han", "Bin Yan", "Wuhui", "Fengda Zhu", "Xing Wang", "Yi Jiang", "BINGYUE PENG", "Zehuan Yuan" ], "affiliation": "ByteDance Inc.", "summary": "InfinityStar is the first discrete autoregressive video generator capable of producing industrial-level 720p videos.", "links": { "openreview": "https://openreview.net/forum?id=JcEqp4aPmb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-74", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Does Stochastic Gradient really succeed for bandits?", "authors": [ "Dorian Baudry", "Emmeran Johnson", "Simon Vary", "Ciara Pike-Burke", "Patrick Rebeschini" ], "affiliation": "INRIA", "summary": "We propose a novel regret analysis of a simple policy gradient algorithm for bandits, characterizing regret regimes depending on its learning rate.", "links": { "openreview": "https://openreview.net/forum?id=gL4muAFwsh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-75", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Perception Encoder: The best visual embeddings are not at the output of the network", "authors": [ "Daniel Bolya", "Po-Yao Huang", "Peize Sun", "Jang Hyun Cho", "Andrea Madotto", "Chen Wei", "Tengyu Ma", "Jiale Zhi", "Jathushan Rajasegaran", "Hanoona Abdul Rasheed", "Junke Wang", "Marco Monteiro", "Hu Xu", "Shiyu Dong", "Nikhila Ravi", "Shang-Wen Li", "Piotr Dollar", "Christoph Feichtenhofer" ], "affiliation": "Meta", "summary": "We develop a CLIP model that is SotA on both image and video zero-shot recognition. Using its strong, general features we further create SotA encoders for language and spatial tasks.", "links": { "openreview": "https://openreview.net/forum?id=INqBOmwIpG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-oral-76", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "PlayerOne: Egocentric World Simulator", "authors": [ "Yuanpeng Tu", "Hao Luo", "Xi Chen", "Xiang Bai", "Fan Wang", "Hengshuang Zhao" ], "affiliation": "The University of Hong Kong", "summary": "We introduce PlayerOne, the first egocentric realistic world simulator, facilitating immersive and unrestricted exploration within vividly dynamic environments. Given an egocentric scene image from the user, PlayerOne can accurately construct the corresponding world and generate egocentric videos that are strictly aligned with the real-scene human", "links": { "openreview": "https://openreview.net/forum?id=Gq4Gay8rDB", "arxiv": "https://arxiv.org/abs/2506.09995", "detail": "" } }, { "uid": "neurips-2025-oral-77", "conference": "NeurIPS", "year": 2025, "track": "Oral", "title": "Mean Flows for One-step Generative Modeling", "authors": [ "Zhengyang Geng", "Mingyang Deng", "Xingjian Bai", "J Zico Kolter", "Kaiming He" ], "affiliation": "Carnegie Mellon University", "summary": "We propose a principled and effective framework for one-step generative modeling. We introduce the notion of average velocity to characterize flow fields, in contrast to instantaneous velocity modeled by Flow Matching methods.", "links": { "openreview": "https://openreview.net/forum?id=uWj4s7rMnR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-1", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MoCha: Towards Movie-Grade Talking Character Generation", "authors": [ "Cong Wei", "Bo Sun", "Haoyu Ma", "Ji Hou", "Felix Juefei-Xu", "Zecheng He", "Xiaoliang Dai", "Luxin Zhang", "Kunpeng Li", "Tingbo Hou", "Animesh Sinha", "Peter Vajda", "Wenhu Chen" ], "affiliation": "", "summary": "Recent advancements in video generation have achieved impressive motion realism, yet they often overlook character-driven storytelling, a crucial task for automated film, animation generation", "links": { "openreview": "https://openreview.net/forum?id=sMWNkIyM41", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-2", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "WISA: World simulator assistant for physics-aware text-to-video generation", "authors": [ "Jing Wang", "Ao Ma", "Ke Cao", "Jun Zheng", "Jiasong Feng", "Zhanjie Zhang", "Wanyuan Pang", "Xiaodan Liang" ], "affiliation": "", "summary": "Recent advances in text-to-video (T2V) generation, exemplified by models such as Sora and Kling, have demonstrated strong potential for constructing world simulators", "links": { "openreview": "https://openreview.net/forum?id=l6TuKz5zvT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-3", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ModHiFi: Identifying High Fidelity predictive components for Model Modification", "authors": [ "Dhruva Kashyap", "Chaitanya Murti", "Pranav K Nayak", "Tanay Narshana", "Chiranjib Bhattacharyya" ], "affiliation": "", "summary": "Open weight models, which are ubiquitous, rarely provide access to their training data or loss function", "links": { "openreview": "https://openreview.net/forum?id=lClK4uBxSG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-4", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Structure of Relation Decoding Linear Operators in Large Language Models", "authors": [ "Miranda Anna Christ", "Adrián Csiszárik", "Gergely Becsó", "Dániel Varga" ], "affiliation": "", "summary": "This paper investigates the structure of linear operators introduced in Hernandez et al", "links": { "openreview": "https://openreview.net/forum?id=XsBzmJzJ2l", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-5", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "KLASS: KL-Guided Fast Inference in Masked Diffusion Models", "authors": [ "Seo Hyun Kim", "Sunwoo Hong", "Hojung Jung", "Youngrok Park", "Se-Young Yun" ], "affiliation": "", "summary": "Masked diffusion models have demonstrated competitive results on various tasks including language generation", "links": { "openreview": "https://openreview.net/forum?id=gOG9Zoyn4R", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-6", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "HM3: Hierarchical Multi-Objective Model Merging for Pretrained Models", "authors": [ "Yu Zhou", "Xingyu Wu", "Jibin Wu", "Liang Feng", "KC Tan" ], "affiliation": "", "summary": "Model merging is a technique that combines multiple large pretrained models into a single model, enhancing performance and broadening task adaptability without original data or additional training", "links": { "openreview": "https://openreview.net/forum?id=JeP0lpusYw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-7", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Structured Sparse Transition Matrices to Enable State Tracking in State-Space Models", "authors": [ "Aleksandar Terzic", "Nicolas Menet", "Michael Hersche", "Thomas Hofmann", "Abbas Rahimi" ], "affiliation": "", "summary": "Modern state-space models (SSMs) often utilize structured transition matrices which enable efficient computation but pose restrictions on the model’s expressivity, as measured in terms of the ability", "links": { "openreview": "https://openreview.net/forum?id=RDbuSCWhad", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-8", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning", "authors": [ "Chaofan Lin", "Jiaming Tang", "Shuo Yang", "Hanshuo Wang", "Tian Tang", "Boyu Tian", "Ion Stoica", "Song Han", "Mingyu Gao" ], "affiliation": "", "summary": "Leveraging attention sparsity to accelerate long-context large language models (LLMs) has been of great importance recently", "links": { "openreview": "https://openreview.net/forum?id=Ve693NkzcU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-9", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation", "authors": [ "UZAIR AKBAR", "Niki Kilbertus", "Hao Shen", "Krikamol Muandet", "Bo Dai" ], "affiliation": "", "summary": "The technique of data augmentation (DA) is often used in machine learning for regularization purposes to better generalize under i", "links": { "openreview": "https://openreview.net/forum?id=C1LVIInfZO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-10", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Deciphering the Extremes: A Novel Approach for Pathological Long-tailed Recognition in Scientific Discovery", "authors": [ "Zhe Zhao", "HaiBin Wen", "Xianfu Liu", "Rui Mao", "Pengkun Wang", "Liheng Yu", "Linjiang Chen", "Bo An", "Qingfu Zhang", "Yang Wang" ], "affiliation": "", "summary": "Scientific discovery across diverse fields increasingly grapples with datasets exhibiting pathological long-tailed distributions: a few common phenomena overshadow a multitude of rare yet scientifical", "links": { "openreview": "https://openreview.net/forum?id=E16vULI6AF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-11", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OpenCUA: Open Foundations for Computer-Use Agents", "authors": [ "Xinyuan Wang", "Bowen Wang", "Dunjie Lu", "Junlin Yang", "Tianbao Xie", "Junli Wang", "Jiaqi Deng", "Xiaole Guo", "Yiheng Xu", "Chen Henry Wu", "Zhennan Shen", "Zhuokai Li", "Ryan Li", "Xiaochuan Li", "Junda Chen", "Zheng Boyuan", "LI PEIHANG", "Fangyu Lei", "Ruisheng Cao", "Yeqiao Fu", "Dongchan Shin", "Martin Shin", "Hu Jiarui", "Yuyan Wang", "Jixuan Chen", "Yuxiao Ye", "Danyang Zhang", "Yipu Wang", "Heng Wang", "Diyi Yang", "Victor Zhong", "Y.Charles", "Zhilin Yang", "Tao Yu" ], "affiliation": "", "summary": "Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks", "links": { "openreview": "https://openreview.net/forum?id=6iRZvJiC9Q", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-12", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Near-Optimal Experiment Design in Linear non-Gaussian Cyclic Models", "authors": [ "Ehsan Sharifian", "Saber Salehkaleybar", "Negar Kiyavash" ], "affiliation": "", "summary": "We study the problem of causal structure learning from a combination of observational and interventional data generated by a linear non-Gaussian structural equation model that might contain cycles", "links": { "openreview": "https://openreview.net/forum?id=opAU0pYlcP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-13", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Escaping saddle points without Lipschitz smoothness: the power of nonlinear preconditioning", "authors": [ "Alexander Bodard", "Panagiotis Patrinos" ], "affiliation": "", "summary": "We study generalized smoothness in nonconvex optimization, focusing on $(L_0, L_1)$-smoothness and anisotropic smoothness", "links": { "openreview": "https://openreview.net/forum?id=7qrhHzZpTA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-14", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Activation Control for Efficiently Eliciting Long Chain-of-thought Ability of Language Models", "authors": [ "Zekai Zhao", "Qi Liu", "Kun Zhou", "Zihan Liu", "Yifei Shao", "Zhiting Hu", "Biwei Huang" ], "affiliation": "", "summary": "Despite the remarkable reasoning performance, eliciting the long chain-of-thought(CoT) ability in large language models(LLMs) typically requires costly reinforcement learning or supervised fine-tuning", "links": { "openreview": "https://openreview.net/forum?id=XNo4yS9n1k", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-15", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Direct Fisher Score Estimation for Likelihood Maximization", "authors": [ "Sherman Khoo", "Yakun Wang", "Song Liu", "Mark Beaumont" ], "affiliation": "", "summary": "We study the problem of likelihood maximization when the likelihood function is intractable but model simulations are readily available", "links": { "openreview": "https://openreview.net/forum?id=2h8bFmEQwh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-16", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Path-Enhanced Contrastive Learning for Recommendation", "authors": [ "Haoran Sun", "Fei Xiong", "Yuanzhe Hu", "Liang Wang" ], "affiliation": "", "summary": "Collaborative filtering (CF) methods are now facing the challenge of data sparsity in recommender systems", "links": { "openreview": "https://openreview.net/forum?id=xKmlBQhgI4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-17", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "T-REGS: Minimum Spanning Tree Regularization for Self-Supervised Learning", "authors": [ "Julie Mordacq", "David Loiseaux", "Vicky Kalogeiton", "Steve Oudot" ], "affiliation": "", "summary": "Self-supervised learning (SSL) has emerged as a powerful paradigm for learning representations without labeled data, often by enforcing invariance to input transformations such as rotations or blurrin", "links": { "openreview": "https://openreview.net/forum?id=jvObbvshjE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-18", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Generating Informative Samples for Risk-Averse Fine-Tuning of Downstream Tasks", "authors": [ "Heasung Kim", "Taekyun Lee", "Hyeji Kim", "Gustavo De Veciana" ], "affiliation": "", "summary": "Risk-averse modeling is critical in safety-sensitive and high-stakes applications", "links": { "openreview": "https://openreview.net/forum?id=kfB5Ciz2XZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-19", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforced Self-Play", "authors": [ "Ran Xu", "Yuchen Zhuang", "Zihan Dong", "Ruiyu Wang", "Yue Yu", "Joyce C. Ho", "Linjun Zhang", "Haoyu Wang", "Wenqi Shi", "Carl Yang" ], "affiliation": "", "summary": "Search-augmented LLMs often struggle with complex reasoning tasks due to ineffective multi-hop retrieval and limited reasoning ability", "links": { "openreview": "https://openreview.net/forum?id=jSgCM0uZn3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-20", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method", "authors": [ "Qingwen Zhang", "Xiaomeng Zhu", "Yushan Zhang", "Yixi Cai", "Olov Andersson", "Patric Jensfelt" ], "affiliation": "", "summary": "Previous dominant methods for scene flow estimation focus mainly on input from two consecutive frames, neglecting valuable information in the temporal domain", "links": { "openreview": "https://openreview.net/forum?id=T9qNDtvAJX", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-21", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Web-Shepherd: Advancing PRMs for Reinforcing Web Agents", "authors": [ "Hyungjoo Chae", "Sunghwan Kim", "Junhee Cho", "Seungone Kim", "Seungjun Moon", "Gyeom Hwangbo", "Dongha Lim", "Minjin Kim", "Yeonjun Hwang", "Minju Gwak", "Dongwook Choi", "Minseok Kang", "Gwanhoon Im", "ByeongUng Cho", "Hyojun Kim", "Jun Hee Han", "Taeyoon Kwon", "Minju Kim", "Beong-woo Kwak", "Dongjin Kang", "Jinyoung Yeo" ], "affiliation": "", "summary": "Web navigation is a unique domain that can automate many repetitive real-life tasks and is challenging as it requires long-horizon sequential decision making beyond typical multimodal large language m", "links": { "openreview": "https://openreview.net/forum?id=G2kMroO9UV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-22", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "How do Transformers Learn Implicit Reasoning?", "authors": [ "Jiaran Ye", "Zijun Yao", "Zhidian Huang", "Liangming Pan", "Jinxin Liu", "Yushi Bai", "Amy Xin", "Liu Weichuan", "Xiaoyin Che", "Lei Hou", "Juanzi Li" ], "affiliation": "", "summary": "Recent work suggests that large language models (LLMs) can perform multi-hop reasoning implicitly---producing correct answers without explicitly verbalizing intermediate steps---but the underlying mec", "links": { "openreview": "https://openreview.net/forum?id=19ygs48nOa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-23", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the sample complexity of semi-supervised multi-objective learning", "authors": [ "Tobias Wegel", "Geelon So", "Junhyung Park", "Fanny Yang" ], "affiliation": "", "summary": "In multi-objective learning (MOL), several possibly competing prediction tasks must be solved jointly by a single model", "links": { "openreview": "https://openreview.net/forum?id=IrgQe6YjKm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-24", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Diversity-Aware Policy Optimization for Large Language Model Reasoning", "authors": [ "Jian Yao", "Ran Cheng", "Xingyu Wu", "Jibin Wu", "KC Tan" ], "affiliation": "", "summary": "The reasoning capabilities of large language models (LLMs) have advanced rapidly, particularly following the release of DeepSeek-R1, which has inspired a surge of research into data quality and reinfo", "links": { "openreview": "https://openreview.net/forum?id=5eZ0iykpDU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-25", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fixed-Point RNNs: Interpolating from Diagonal to Dense", "authors": [ "Sajad Movahedi", "Felix Sarnthein", "Nicola Muca Cirone", "Antonio Orvieto" ], "affiliation": "", "summary": "Linear recurrent neural networks (RNNs) and state-space models (SSMs) such as Mamba have become promising alternatives to softmax-attention as sequence mixing layers in Transformer architectures", "links": { "openreview": "https://openreview.net/forum?id=KT8y9pFgJE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-26", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Bridging Theory and Practice in Link Representation with Graph Neural Networks", "authors": [ "Veronica Lachi", "Francesco Ferrini", "Antonio Longa", "Bruno Lepri", "Andrea Passerini", "Manfred Jaeger" ], "affiliation": "", "summary": "Graph Neural Networks (GNNs) are widely used to compute representations of node pairs for downstream tasks such as link prediction", "links": { "openreview": "https://openreview.net/forum?id=WYnvP3DePZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-27", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ErrorTrace: A Black-Box Traceability Mechanism Based on Model Family Error Space", "authors": [ "Chuanchao Zang", "Xiangtao Meng", "Wenyu Chen", "Tianshuo Cong", "Zha Yaxing", "Dong Qi", "Zheng Li", "Shanqing Guo" ], "affiliation": "", "summary": "The open-source release of large language models (LLMs) enables malicious users to create unauthorized derivative models at low cost, posing significant threats to intellectual property (IP) and marke", "links": { "openreview": "https://openreview.net/forum?id=3P3PL7aCXM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-28", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Protein Design with Dynamic Protein Vocabulary", "authors": [ "Nuowei Liu", "Jiahao Kuang", "Yanting Liu", "Tao Ji", "Changzhi Sun", "Man Lan", "Yuanbin Wu" ], "affiliation": "", "summary": "Protein design is a fundamental challenge in biotechnology, aiming to design novel sequences with specific functions within the vast space of possible proteins", "links": { "openreview": "https://openreview.net/forum?id=MpJkAzwUtl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-29", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Reliable Code-as-Policies: A Neuro-Symbolic Framework for Embodied Task Planning", "authors": [ "Sanghyun Ahn", "Wonje Choi", "Junyong Lee", "Jinwoo Park", "Honguk Woo" ], "affiliation": "", "summary": "Recent advances in large language models (LLMs) have enabled the automatic generation of executable code for task planning and control in embodied agents such as robots, demonstrating the potential of", "links": { "openreview": "https://openreview.net/forum?id=VaC4sa96EI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-30", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AI-Researcher: Autonomous Scientific Innovation", "authors": [ "Jiabin Tang", "Lianghao Xia", "Zhonghang Li", "Chao Huang" ], "affiliation": "", "summary": "The powerful reasoning capabilities of Large Language Models (LLMs) in mathematics and coding, combined with their ability to automate complex tasks through agentic frameworks, present unprecedented o", "links": { "openreview": "https://openreview.net/forum?id=kQWyOYUAC4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-31", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Abstain Mask Retain Core: Time Series Prediction by Adaptive Masking Loss with Representation Consistency", "authors": [ "Renzhao Liang", "Sizhe Xu", "Chenggang Xie", "Jingru Chen", "Feiyang Ren", "Shu Yang", "Takahiro Yabe" ], "affiliation": "", "summary": "Time series forecasting plays a pivotal role in critical domains such as energy management and financial markets", "links": { "openreview": "https://openreview.net/forum?id=KrglRiOKYT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-32", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation", "authors": [ "Kefei Zhu", "Fengshuo Bai", "YuanHao Xiang", "Yishuai Cai", "Xinglin Chen", "Ruochong Li", "Xingtao Wang", "Hao Dong", "Yaodong Yang", "Xiaopeng Fan", "Yuanpei Chen" ], "affiliation": "", "summary": "Dexterous manipulation is critical for advancing robot capabilities in real-world applications, yet diverse and high-quality datasets remain scarce", "links": { "openreview": "https://openreview.net/forum?id=a49F7EAm6l", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-33", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learning Robust Vision-Language Models from Natural Latent Spaces", "authors": [ "Zhangyun Wang", "Ni Ding", "Aniket Mahanti" ], "affiliation": "", "summary": "Pre-trained vision-language models (VLMs) exhibit significant vulnerability to imperceptible adversarial perturbations", "links": { "openreview": "https://openreview.net/forum?id=7G9YKty2UZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-34", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Accelerating Diffusion LLMs via Adaptive Parallel Decoding", "authors": [ "Daniel Mingyi Israel", "Guy Van den Broeck", "Aditya Grover" ], "affiliation": "", "summary": "The generation speed of LLMs are bottlenecked by autoregressive decoding, where tokens are predicted sequentially one by one", "links": { "openreview": "https://openreview.net/forum?id=xwqTt26NJf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-35", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Inner Speech as Behavior Guides: Steerable Imitation of Diverse Behaviors for Human-AI coordination", "authors": [ "Rakshit Trivedi", "Kartik Sharma", "David C. Parkes" ], "affiliation": "", "summary": "Effective human-AI coordination requires artificial agents capable of exhibiting and responding to human-like behaviors while adapting to changing contexts", "links": { "openreview": "https://openreview.net/forum?id=AwLRF1lZvI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-36", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Graph-Based Attention for Differentiable MaxSAT Solving", "authors": [ "Sota Moriyama", "Katsumi Inoue" ], "affiliation": "", "summary": "The use of deep learning to solve fundamental AI problems such as Boolean Satisfiability (SAT) has been explored recently to develop robust and scalable reasoning systems", "links": { "openreview": "https://openreview.net/forum?id=g9XLUU3TaG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-37", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Pass@K Policy Optimization: Solving Harder Reinforcement Learning Problems", "authors": [ "Christian Walder", "Deep Tejas Karkhanis" ], "affiliation": "", "summary": "Reinforcement Learning algorithms commonly sample multiple ($n>1$) solution attempts for each problem and reward them independently", "links": { "openreview": "https://openreview.net/forum?id=W6WC6047X2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-38", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LogicTree: Improving Complex Reasoning of LLMs via Instantiated Multi-step Synthetic Logical Data", "authors": [ "Zehao Wang", "Lin Yang", "Jie Wang", "Kehan Wang", "Hanzhu Chen", "Bin Wang", "Jianye HAO", "Defu Lian", "Bin Li", "Enhong Chen" ], "affiliation": "", "summary": "Despite their remarkable performance on various tasks, Large Language Models (LLMs) still struggle with logical reasoning, particularly in complex and multi-step reasoning processes", "links": { "openreview": "https://openreview.net/forum?id=z4AMrCOetn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-39", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Multitask Learning with Stochastic Interpolants", "authors": [ "Hugo Negrel", "Florentin Coeurdoux", "Michael Samuel Albergo", "Eric Vanden-Eijnden" ], "affiliation": "", "summary": "We propose a framework for learning maps between probability distributions that broadly generalizes the time dynamics of flow and diffusion models", "links": { "openreview": "https://openreview.net/forum?id=9k9ZsDs9Vc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-40", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FUDOKI: Discrete Flow-based Unified Understanding and Generation via Kinetic-Optimal Velocities", "authors": [ "Jin Wang", "Yao Lai", "Aoxue Li", "Shifeng Zhang", "Jiacheng Sun", "Ning Kang", "Chengyue Wu", "Zhenguo Li", "Ping Luo" ], "affiliation": "", "summary": "The rapid progress of large language models (LLMs) has catalyzed the emergence of multimodal large language models (MLLMs) that unify visual understanding and image generation within a single framewor", "links": { "openreview": "https://openreview.net/forum?id=RSVdHXZN6D", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-41", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fast MRI for All: Bridging Access Gaps by Training without Raw Data", "authors": [ "Yasar Utku Alcalar", "Merve Gulle", "Mehmet Akcakaya" ], "affiliation": "", "summary": "Physics-driven deep learning (PD-DL) approaches have become popular for improved reconstruction of fast magnetic resonance imaging (MRI) scans", "links": { "openreview": "https://openreview.net/forum?id=ugBmWX3H1R", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-42", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Knowledge Insulating Vision-Language-Action Models: Train Fast, Run Fast, Generalize Better", "authors": [ "Danny Driess", "Jost Tobias Springenberg", "brian ichter", "LILI YU", "Adrian Li-Bell", "Karl Pertsch", "Allen Z. Ren", "Homer Walke", "Quan Vuong", "Lucy Xiaoyang Shi", "Sergey Levine" ], "affiliation": "", "summary": "Vision-language-action (VLA) models provide a powerful approach to training control policies for physical systems, such as robots, by combining end-to-end learning with transfer of semantic knowledge", "links": { "openreview": "https://openreview.net/forum?id=cb0xbZ3APM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-43", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Complete Structure Guided Point Cloud Completion via Cluster- and Instance-Level Contrastive Learning", "authors": [ "Yang Chen", "Yirun Zhou", "WEIZHONG ZHANG", "Cheng Jin" ], "affiliation": "", "summary": "Point cloud completion, aiming to reconstruct missing part from incomplete point clouds, is a pivotal task in 3D computer vision", "links": { "openreview": "https://openreview.net/forum?id=4f6mEr1DQs", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-44", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ARECHO: Autoregressive Evaluation via Chain-Based Hypothesis Optimization for Speech Multi-Metric Estimation", "authors": [ "Jiatong Shi", "Yifan Cheng", "Bo-Hao Su", "Hye-jin Shim", "Jinchuan Tian", "Samuele Cornell", "Yiwen Zhao", "Siddhant Arora", "Shinji Watanabe" ], "affiliation": "", "summary": "Speech signal analysis poses significant challenges, particularly in tasks such as speech quality evaluation and profiling, where the goal is to predict multiple perceptual and objective metrics", "links": { "openreview": "https://openreview.net/forum?id=P2yIMJP5b1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-45", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Projective Equivariant Networks via Second-order Fundamental Differential Invariants", "authors": [ "Yikang Li", "Yeqing Qiu", "Yuxuan Chen", "Lingshen He", "Lexiang Hu", "Zhouchen Lin" ], "affiliation": "", "summary": "Equivariant networks enhance model efficiency and generalization by embedding symmetry priors into their architectures", "links": { "openreview": "https://openreview.net/forum?id=crczm2smVo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-46", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ZeroS: Zero‑Sum Linear Attention for Efficient Transformers", "authors": [ "Jiecheng Lu", "Xu Han", "Yan Sun", "Viresh Pati", "Yubin Kim", "Siddhartha Somani", "Shihao Yang" ], "affiliation": "", "summary": "Linear attention methods offer Transformers $O(N)$ complexity but typically underperform standard softmax attention", "links": { "openreview": "https://openreview.net/forum?id=Ms6IXbfzzX", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-47", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Blameless Users in a Clean Room: Defining Copyright Protection for Generative Models", "authors": [ "Aloni Cohen" ], "affiliation": "", "summary": "Are there any conditions under which a generative model’s outputs are guaranteed not to infringe the copyrights of its training data? This is the question of \"provable copyright protection\" first pose", "links": { "openreview": "https://openreview.net/forum?id=V8SndhCN0z", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-48", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates", "authors": [ "Seunghyeok Shin", "Dabin Kim", "Hongki Lim" ], "affiliation": "", "summary": "Reconstructing high-quality point clouds from images remains challenging in computer vision", "links": { "openreview": "https://openreview.net/forum?id=IJLqUjtrls", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-49", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "From Shortcut to Induction Head: How Data Diversity Shapes Algorithm Selection in Transformers", "authors": [ "Ryotaro Kawata", "Yujin Song", "Alberto Bietti", "Naoki Nishikawa", "Taiji Suzuki", "Samuel Vaiter", "Denny Wu" ], "affiliation": "", "summary": "Transformers can implement both generalizable algorithms (e", "links": { "openreview": "https://openreview.net/forum?id=n0QvMU2kON", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-50", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learning Interestingness in Automated Mathematical Theory Formation", "authors": [ "George Tsoukalas", "Rahul Saha", "Amitayush Thakur", "Sabrina Reguyal", "Swarat Chaudhuri" ], "affiliation": "", "summary": "We take two key steps in automating the open-ended discovery of new mathematical theories, a grand challenge in artificial intelligence", "links": { "openreview": "https://openreview.net/forum?id=RespmwOoCH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-51", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DNA-DetectLLM: Unveiling AI-Generated Text via a DNA-Inspired Mutation-Repair Paradigm", "authors": [ "Xiaowei Zhu", "Yubing Ren", "Fang Fang", "Qingfeng Tan", "Shi Wang", "Yanan Cao" ], "affiliation": "", "summary": "The rapid advancement of large language models (LLMs) has blurred the line between AI-generated and human-written text", "links": { "openreview": "https://openreview.net/forum?id=yQoHUijSHx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-52", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Signal and Noise: A Framework for Reducing Uncertainty in Language Model Evaluation", "authors": [ "David Heineman", "Valentin Hofmann", "Ian Magnusson", "Yuling Gu", "Noah A. Smith", "Hannaneh Hajishirzi", "Kyle Lo", "Jesse Dodge" ], "affiliation": "", "summary": "Developing large language models is expensive and often involves making decisions with small experiments, typically by evaluating on large, multi-task evaluation suites", "links": { "openreview": "https://openreview.net/forum?id=sAFottNlra", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-53", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Tradeoffs between Mistakes and ERM Oracle Calls in Online and Transductive Online Learning", "authors": [ "Idan Attias", "Steve Hanneke", "Arvind Ramaswami" ], "affiliation": "", "summary": "We study online and transductive online learning in settings where the learner can interact with the concept class only via Empirical Risk Minimization (ERM) or weak consistency oracles on arbitrary s", "links": { "openreview": "https://openreview.net/forum?id=u2GzxdWLFW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-54", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The World Is Bigger! A Computationally-Embedded Perspective on the Big World Hypothesis", "authors": [ "Alex Lewandowski", "Aditya A. Ramesh", "Edan Meyer", "Dale Schuurmans", "Marlos C. Machado" ], "affiliation": "", "summary": "Continual learning is often motivated by the idea, known as the big world hypothesis, that ``the world is bigger'' than the agent", "links": { "openreview": "https://openreview.net/forum?id=gJclyLFSdU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-55", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Multimodal Disease Progression Modeling via Spatiotemporal Disentanglement and Multiscale Alignment", "authors": [ "Chen Liu", "Wenfang Yao", "Kejing Yin", "William K. Cheung", "Jing Qin" ], "affiliation": "", "summary": "Longitudinal multimodal data, including electronic health records (EHR) and sequential chest X-rays (CXRs), is critical for modeling disease progression, yet remains underutilized due to two key chall", "links": { "openreview": "https://openreview.net/forum?id=2afhRWVb6p", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-56", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Hardness of Conditional Independence Testing In Practice", "authors": [ "Zheng He", "Roman Pogodin", "Yazhe Li", "Namrata Deka", "Arthur Gretton", "Danica J. Sutherland" ], "affiliation": "", "summary": "Tests of conditional independence (CI) underpin a number of important problems in machine learning and statistics, from causal discovery to evaluation of predictor fairness and out-of-distribution rob", "links": { "openreview": "https://openreview.net/forum?id=Tn1M71PDfF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-57", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Emergence and Evolution of Interpretable Concepts in Diffusion Models", "authors": [ "Berk Tinaz", "Zalan Fabian", "Mahdi Soltanolkotabi" ], "affiliation": "", "summary": "Diffusion models have become the go-to method for text-to-image generation, producing high-quality images from pure noise", "links": { "openreview": "https://openreview.net/forum?id=C7LxkebvVW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-58", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PCA++: How Uniformity Induces Robustness to Background Noise in Contrastive Learning", "authors": [ "Mingqi Wu", "Qiang Sun", "Archer Y. Yang" ], "affiliation": "", "summary": "High-dimensional data often conceal low-dimensional signals beneath structured background noise, limiting standard PCA", "links": { "openreview": "https://openreview.net/forum?id=at87L8EuzR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-59", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Optimal Nuisance Function Tuning for Estimating a Doubly Robust Functional under Proportional Asymptotics", "authors": [ "Sean McGrath", "Debarghya Mukherjee", "Rajarshi Mukherjee", "Zixiao Wang" ], "affiliation": "", "summary": "In this paper, we explore the asymptotically optimal tuning parameter choice in ridge regression for estimating nuisance functions of a statistical functional that has recently gained prominence in co", "links": { "openreview": "https://openreview.net/forum?id=htN3xhUpjD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-60", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GeoLLaVA-8K: Scaling Remote-Sensing Multimodal Large Language Models to 8K Resolution", "authors": [ "Fengxiang Wang", "Mingshuo Chen", "Yueying Li", "Di Wang", "Haotian Wang", "Zonghao Guo", "Zefan Wang", "Shan Boqi", "Long Lan", "Yulin Wang", "Hongzhen Wang", "Wenjing Yang", "Bo Du", "Jing Zhang" ], "affiliation": "", "summary": "Ultra-high-resolution (UHR) remote sensing (RS) imagery offers valuable data for Earth observation but pose challenges for existing multimodal foundation models due to two key bottlenecks: (1) limited", "links": { "openreview": "https://openreview.net/forum?id=LTgUInLTbP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-61", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Environment Inference for Learning Generalizable Dynamical System", "authors": [ "Shixuan Liu", "Yue He", "Haotian Wang", "Wenjing Yang", "Yunfei Wang", "Peng Cui", "Zhong Liu" ], "affiliation": "", "summary": "Data-driven methods offer efficient and robust solutions for analyzing complex dynamical systems but rely on the assumption of I", "links": { "openreview": "https://openreview.net/forum?id=2M5dTDdGxl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-62", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Abstract Rendering: Certified Rendering Under 3D Semantic Uncertainty", "authors": [ "Chenxi Ji", "Yangge Li", "Xiangru Zhong", "Huan Zhang", "Sayan Mitra" ], "affiliation": "", "summary": "Rendering produces 2D images from 3D scene representations, yet how continuous variations in camera pose and scenes influence these images—and, consequently, downstream visual models—remains underexpl", "links": { "openreview": "https://openreview.net/forum?id=EXIKFM1Q9R", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-63", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Inference-Time Reward Hacking in Large Language Models", "authors": [ "Hadi Khalaf", "Claudio Mayrink Verdun", "Alex Oesterling", "Himabindu Lakkaraju", "Flavio Calmon" ], "affiliation": "", "summary": "A common paradigm to improve the performance of large language models is optimizing for a reward model", "links": { "openreview": "https://openreview.net/forum?id=hSX7Dd8dxy", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-64", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "High-Performance Arithmetic Circuit Optimization via Differentiable Architecture Search", "authors": [ "Xilin Xia", "Jie Wang", "Wanbo Zhang", "Zhihai Wang", "Mingxuan Yuan", "Jianye HAO", "Feng Wu" ], "affiliation": "", "summary": "Arithmetic circuit optimization remains a fundamental challenge in modern integrated circuit design", "links": { "openreview": "https://openreview.net/forum?id=dtsYpcJr1R", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-65", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ROOT: Rethinking Offline Optimization as Distributional Translation via Probabilistic Bridge", "authors": [ "Manh Cuong Dao", "The Hung Tran", "Phi Le Nguyen", "Thao Nguyen Truong", "Trong Nghia Hoang" ], "affiliation": "", "summary": "This paper studies the black-box optimization task which aims to find the maxima of a black-box function using a static set of its observed input-output pairs", "links": { "openreview": "https://openreview.net/forum?id=QFuuxfmqb5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-66", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Go With the Flow: Fast Diffusion for Gaussian Mixture Models", "authors": [ "George Rapakoulias", "Ali Reza Pedram", "Fengjiao Liu", "Lingjiong Zhu", "Panagiotis Tsiotras" ], "affiliation": "", "summary": "Schrodinger Bridges (SBs) are diffusion processes that steer, in finite time, a given initial distribution to another final one while minimizing a suitable cost functional", "links": { "openreview": "https://openreview.net/forum?id=bmznY5wYXH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-67", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SHF: Symmetrical Hierarchical Forest with Pretrained Vision Transformer Encoder for High-Resolution Medical Segmentation", "authors": [ "Enzhi Zhang", "Peng Chen", "Rui Zhong", "Du Wu", "Jun Igarashi", "Isaac Lyngaas", "Xiao Wang", "Masaharu Munetomo", "Mohamed Wahib" ], "affiliation": "", "summary": "This paper presents a novel approach to addressing the long-sequence problem in high-resolution medical images for Vision Transformers (ViTs)", "links": { "openreview": "https://openreview.net/forum?id=3IXdXBpuLn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-68", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ProxySPEX: Inference-Efficient Interpretability via Sparse Feature Interactions in LLMs", "authors": [ "Landon Butler", "Abhineet Agarwal", "Justin Singh Kang", "Yigit Efe Erginbas", "Bin Yu", "Kannan Ramchandran" ], "affiliation": "", "summary": "Large Language Models (LLMs) have achieved remarkable performance by capturing complex interactions between input features", "links": { "openreview": "https://openreview.net/forum?id=KI8qan2EA7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-69", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fair Cooperation in Mixed-Motive Games via Conflict-Aware Gradient Adjustment", "authors": [ "Woojun Kim", "Katia P. Sycara" ], "affiliation": "", "summary": "Multi-agent reinforcement learning in mixed-motive settings presents a fundamental challenge: agents must balance individual interests with collective goals, which are neither fully aligned nor strict", "links": { "openreview": "https://openreview.net/forum?id=yPsJ1PKiAi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-70", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts", "authors": [ "Shiting Xiao", "Rishabh Kabra", "Yuhang Li", "Donghyun Lee", "Joao Carreira", "Priyadarshini Panda" ], "affiliation": "", "summary": "The ability to segment objects based on open-ended language prompts remains a critical challenge, requiring models to ground textual semantics into precise spatial masks while handling diverse and uns", "links": { "openreview": "https://openreview.net/forum?id=fLx3vQPmDu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-71", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Broken Tokens? Your Language Model can Secretly Handle Non-Canonical Tokenizations", "authors": [ "Brian Siyuan Zheng", "Alisa Liu", "Orevaoghene Ahia", "Jonathan Hayase", "Yejin Choi", "Noah A. Smith" ], "affiliation": "", "summary": "Modern tokenizers employ deterministic algorithms to map text into a single ``canonical\" token sequence, yet the same string can be encoded as many non-canonical tokenizations using the language model", "links": { "openreview": "https://openreview.net/forum?id=WrYWolqKh3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-72", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Hybrid-Balance GFlowNet for Solving Vehicle Routing Problems", "authors": [ "Ni Zhang", "Zhiguang Cao" ], "affiliation": "", "summary": "Existing GFlowNet-based methods for vehicle routing problems (VRPs) typically employ Trajectory Balance (TB) to achieve global optimization but often neglect important aspects of local optimization", "links": { "openreview": "https://openreview.net/forum?id=GxGrGswvND", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-73", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Feedback-Aware MCTS for Goal-Oriented Information Seeking", "authors": [ "Harshita Chopra", "Chirag Shah" ], "affiliation": "", "summary": "Effective decision-making and problem-solving in conversational systems require the ability to identify and acquire missing information through targeted questioning", "links": { "openreview": "https://openreview.net/forum?id=ustF8MMZDJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-74", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Self-Supervised Learning of Motion Concepts by Optimizing Counterfactuals", "authors": [ "Stefan Stojanov", "David Wendt", "Seungwoo Kim", "Rahul Mysore Venkatesh", "Kevin Feigelis", "Klemen Kotar", "Khai Loong Aw", "Jiajun Wu", "Daniel LK Yamins" ], "affiliation": "", "summary": "Estimating motion primitives from video (e", "links": { "openreview": "https://openreview.net/forum?id=fGuTN7huo5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-75", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Uncertain Knowledge Graph Completion via Semi-Supervised Confidence Distribution Learning", "authors": [ "Tianxing Wu", "Shutong Zhu", "Jingting Wang", "Ning Xu", "Guilin Qi", "Haofen Wang" ], "affiliation": "", "summary": "Uncertain knowledge graphs (UKGs) associate each triple with a confidence score to provide more precise knowledge representations", "links": { "openreview": "https://openreview.net/forum?id=Mat9FTfiYD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-76", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OCTDiff: Bridged Diffusion Model for Portable OCT Super-Resolution and Enhancement", "authors": [ "Ye Tian", "Angela McCarthy", "Gabriel Gomide", "Nancy Liddle", "Jedrzej Golebka", "Royce Chen", "Jeff Liebmann", "Kaveri A. Thakoor" ], "affiliation": "", "summary": "Medical imaging super-resolution is critical for improving diagnostic utility and reducing costs, particularly for low-cost modalities such as portable Optical Coherence Tomography (OCT)", "links": { "openreview": "https://openreview.net/forum?id=VN5bMTfSZS", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-77", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Bits Leaked per Query: Information-Theoretic Bounds for Adversarial Attacks on LLMs", "authors": [ "Masahiro Kaneko", "Timothy Baldwin" ], "affiliation": "", "summary": "Adversarial attacks by malicious users that threaten the safety of large language models (LLMs) can be viewed as attempts to infer a target property $T$ that is unknown when an instruction is issued,", "links": { "openreview": "https://openreview.net/forum?id=DwXX8c7xst", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-78", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Characterizing control between interacting subsystems with deep Jacobian estimation", "authors": [ "Adam Joseph Eisen", "Mitchell Ostrow", "Sarthak Chandra", "Leo Kozachkov", "Earl K Miller", "Ila R Fiete" ], "affiliation": "", "summary": "Biological function arises through the dynamical interactions of multiple subsystems, including those between brain areas, within gene regulatory networks, and more", "links": { "openreview": "https://openreview.net/forum?id=I822ZIRtms", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-79", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Sharper Convergence Rates for Nonconvex Optimisation via Reduction Mappings", "authors": [ "Evan Markou", "Thalaiyasingam Ajanthan", "Stephen Gould" ], "affiliation": "", "summary": "Many high-dimensional optimisation problems exhibit rich geometric structures in their set of minimisers, often forming smooth manifolds due to over-parametrisation or symmetries", "links": { "openreview": "https://openreview.net/forum?id=ZvqbNFWQkh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-80", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Selective Omniprediction and Fair Abstention", "authors": [ "Sílvia Casacuberta", "Varun Kanade" ], "affiliation": "", "summary": "We propose new learning algorithms for building selective classifiers, which are predictors that are allowed to abstain on some fraction of the domain", "links": { "openreview": "https://openreview.net/forum?id=BoYGLpNXZd", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-81", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Algorithms and SQ Lower Bounds for Robustly Learning Real-valued Multi-Index Models", "authors": [ "Ilias Diakonikolas", "Giannis Iakovidis", "Daniel Kane", "Lisheng Ren" ], "affiliation": "", "summary": "We study the complexity of learning real-valued Multi-Index Models (MIMs) under the Gaussian distribution", "links": { "openreview": "https://openreview.net/forum?id=cD4whTwm6G", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-82", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Principled Approach to Randomized Selection under Uncertainty: Applications to Peer Review and Grant Funding", "authors": [ "Alexander Koujianos Goldberg", "Giulia Fanti", "Nihar B Shah" ], "affiliation": "", "summary": "Many decision-making processes involve evaluating and selecting items, including scientific peer review, job hiring, school admissions, and investment decisions", "links": { "openreview": "https://openreview.net/forum?id=0qRXETZZwv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-83", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Error Forcing in Recurrent Neural Networks", "authors": [ "A Erdem Sağtekin", "Colin Bredenberg", "Cristina Savin" ], "affiliation": "", "summary": "How should feedback influence recurrent neural network (RNN) learning? One way to address the known limitations of backpropagation through time is to directly adjust neural activities during the learn", "links": { "openreview": "https://openreview.net/forum?id=xVI8g50Qfk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-84", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Aligning Text-to-Image Diffusion Models to Human Preference by Classification", "authors": [ "Longquan Dai", "Xiaolu Wei", "He Wang", "Shaomeng Wang", "Jinhui Tang" ], "affiliation": "", "summary": "Text-to-image diffusion models are typically trained on large-scale web data, often resulting in outputs that misalign with human preferences", "links": { "openreview": "https://openreview.net/forum?id=YvypxK4kut", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-85", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Sketched Gaussian Mechanism for Private Federated Learning", "authors": [ "Qiaobo Li", "Zhijie Chen", "Arindam Banerjee" ], "affiliation": "", "summary": "Communication cost and privacy are two major considerations in federated learning (FL)", "links": { "openreview": "https://openreview.net/forum?id=AUs0rScwK0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-86", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Joint‑Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self‑Supervised Learning", "authors": [ "Hugues Van Assel", "Mark Ibrahim", "Tommaso Biancalani", "Aviv Regev", "Randall Balestriero" ], "affiliation": "", "summary": "Reconstruction and joint-embedding have emerged as two leading paradigms in Self‑Supervised Learning (SSL)", "links": { "openreview": "https://openreview.net/forum?id=UOaLsgn5wb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-87", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Accelerating data-driven algorithm selection for combinatorial partitioning problems", "authors": [ "Vaggos Chatziafratis", "Ishani Karmarkar", "Yingxi Li", "Ellen Vitercik" ], "affiliation": "", "summary": "Data-driven algorithm selection is a powerful approach for choosing effective heuristics for computational problems", "links": { "openreview": "https://openreview.net/forum?id=sXpyn3lAb5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-88", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Purifying Shampoo: Investigating Shampoo's Heuristics by Decomposing its Preconditioner", "authors": [ "Runa Eschenhagen", "Aaron Defazio", "Tsung-Hsien Lee", "Richard E. Turner", "Hao-Jun Michael Shi" ], "affiliation": "", "summary": "The recent success of Shampoo in the AlgoPerf contest has sparked renewed interest in Kronecker-factorization-based optimization algorithms for training neural networks", "links": { "openreview": "https://openreview.net/forum?id=kePsKwxvaV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-89", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Replicable Distribution Testing", "authors": [ "Ilias Diakonikolas", "Jingyi Gao", "Daniel Kane", "Sihan Liu", "Christopher Ye" ], "affiliation": "", "summary": "We initiate a systematic investigation of distribution testing in the framework of algorithmic replicability", "links": { "openreview": "https://openreview.net/forum?id=qYDBgSeAlU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-90", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FAPEX: Fractional Amplitude-Phase Expressor for Robust Cross-Subject Seizure Prediction", "authors": [ "Ruizhe Zheng", "Lingyan Mao", "DINGDING HAN", "Tian Luo", "Yi Wang", "Jing Ding", "Yuguo Yu" ], "affiliation": "", "summary": "Precise, generalizable subject-agnostic seizure prediction (SASP) remains a fundamental challenge due to the intrinsic complexity and significant spectral variability of electrophysiologial signals ac", "links": { "openreview": "https://openreview.net/forum?id=A9jXG3FUMT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-91", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems", "authors": [ "Hao Liang", "Shuqing Shi", "Yudi Zhang", "Biwei Huang", "Yali Du" ], "affiliation": "", "summary": "Large‑scale networked systems, such as traffic, power, and wireless grids, challenge reinforcement‑learning agents with both scale and environment shifts", "links": { "openreview": "https://openreview.net/forum?id=dfcQFL89OM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-92", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Conformal Mixed-Integer Constraint Learning with Feasibility Guarantees", "authors": [ "Daniel Ovalle", "Lorenz T. Biegler", "Ignacio E Grossmann", "Carl D Laird", "Mateo Dulce Rubio" ], "affiliation": "", "summary": "We propose Conformal Mixed-Integer Constraint Learning (C-MICL), a novel framework that provides probabilistic feasibility guarantees for data-driven constraints in optimization problems", "links": { "openreview": "https://openreview.net/forum?id=ZvUZvT8tgg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-93", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Why Do Some Language Models Fake Alignment While Others Don't?", "authors": [ "Abhay Sheshadri", "John Hughes", "Julian Michael", "Alex Troy Mallen", "Arun Jose", "Fabien Roger" ], "affiliation": "", "summary": "*Alignment Faking in Large Language Models* presented a demonstration of Claude 3 Opus and Claude 3", "links": { "openreview": "https://openreview.net/forum?id=1Imp4KZyjA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-94", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "⁠When Data Can't Meet: Estimating Correlation Across Privacy Barriers", "authors": [ "Abhinav Chakraborty", "Arnab Auddy", "T. Tony Cai" ], "affiliation": "", "summary": "We consider the problem of estimating the correlation of two random variables $X$ and $Y$, where the pairs $(X,Y)$ are not observed together, but are instead separated co-ordinate-wise at two servers:", "links": { "openreview": "https://openreview.net/forum?id=36uy2GgAy6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-95", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TRIDENT: Tri-Modal Molecular Representation Learning with Taxonomic Annotations and Local Correspondence", "authors": [ "Feng Jiang", "Mangal Prakash", "Hehuan Ma", "Jianyuan Deng", "Yuzhi Guo", "Amina Mollaysa", "Tommaso Mansi", "Rui Liao", "Junzhou Huang" ], "affiliation": "", "summary": "Molecular property prediction aims to learn representations that map chemical structures to functional properties", "links": { "openreview": "https://openreview.net/forum?id=M6l3pyvUfr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-96", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Deep Value Benchmark: Measuring Whether Models Generalize Deep Values or Shallow Preferences", "authors": [ "Joshua Ashkinaze", "Hua Shen", "Sai Avula", "Eric Gilbert", "Ceren Budak" ], "affiliation": "", "summary": "We introduce the Deep Value Benchmark (DVB), an evaluation framework that directly tests whether large language models (LLMs) learn fundamental human values or merely surface-level preferences", "links": { "openreview": "https://openreview.net/forum?id=bzxlOyjWbU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-97", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Return of ChebNet: Understanding and Improving an Overlooked GNN on Long Range Tasks", "authors": [ "Ali Hariri", "Alvaro Arroyo", "Alessio Gravina", "Moshe Eliasof", "Carola-Bibiane Schönlieb", "Davide Bacciu", "Xiaowen Dong", "Kamyar Azizzadenesheli", "Pierre Vandergheynst" ], "affiliation": "", "summary": "ChebNet, one of the earliest spectral GNNs, has largely been overshadowed by Message Passing Neural Networks (MPNNs), which gained popularity for their simplicity and effectiveness in capturing local", "links": { "openreview": "https://openreview.net/forum?id=oLyfML1Qze", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-98", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Preconditioned Langevin Dynamics with Score-based Generative Models for Infinite-Dimensional Linear Bayesian Inverse Problems", "authors": [ "Lorenzo Baldassari", "Josselin Garnier", "Knut Solna", "Maarten V. de Hoop" ], "affiliation": "", "summary": "Designing algorithms for solving high-dimensional Bayesian inverse problems directly in infinite‑dimensional function spaces – where such problems are naturally formulated – is crucial to ensure stabi", "links": { "openreview": "https://openreview.net/forum?id=rVyBrD8h2b", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-99", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transferring Linear Features Across Language Models With Model Stitching", "authors": [ "Alan Chen", "Jack Merullo", "Alessandro Stolfo", "Ellie Pavlick" ], "affiliation": "", "summary": "In this work, we demonstrate that affine mappings between residual streams of language models is a cheap way to effectively transfer represented features between models", "links": { "openreview": "https://openreview.net/forum?id=Qvvy0X63Fv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-100", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Causal Differentiating Concepts: Interpreting LM Behavior via Causal Representation Learning", "authors": [ "Navita Goyal", "Hal Daumé III", "Alexandre Drouin", "Dhanya Sridhar" ], "affiliation": "", "summary": "Language model activations entangle concepts that mediate their behavior, making it difficult to interpret these factors, which has implications for generalizability and robustness", "links": { "openreview": "https://openreview.net/forum?id=Zf6Oj5x9sE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-101", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Ambient Proteins - Training Diffusion Models on Noisy Structures", "authors": [ "Giannis Daras", "Jeffrey Ouyang-Zhang", "Krithika Ravishankar", "Constantinos Costis Daskalakis", "Adam Klivans", "Daniel Jesus Diaz" ], "affiliation": "", "summary": "We present Ambient Protein Diffusion, a framework for training protein diffusion models that generates structures with unprecedented diversity and quality", "links": { "openreview": "https://openreview.net/forum?id=Xll01vw606", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-102", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On Universality Classes of Equivariant Networks", "authors": [ "Marco Pacini", "Gabriele Santin", "Bruno Lepri", "Shubhendu Trivedi" ], "affiliation": "", "summary": "Equivariant neural networks provide a principled framework for incorporating symmetry into learning architectures and have been extensively analyzed through the lens of their *separation power*, that", "links": { "openreview": "https://openreview.net/forum?id=V4YAS7NLXi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-103", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Guarantees for Alternating Least Squares in Overparameterized Tensor Decompositions", "authors": [ "Dionysis Arvanitakis", "Vaidehi Srinivas", "Aravindan Vijayaraghavan" ], "affiliation": "", "summary": "Tensor decomposition is a canonical non-convex optimization problem that is computationally challenging, and yet important due to applications in factor analysis and parameter estimation of latent var", "links": { "openreview": "https://openreview.net/forum?id=9FDErIfoVE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-104", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Checklists Are Better Than Reward Models For Aligning Language Models", "authors": [ "Vijay Viswanathan", "Yanchao Sun", "Xiang Kong", "Meng Cao", "Graham Neubig", "Tongshuang Wu" ], "affiliation": "", "summary": "Language models must be adapted to understand and follow user instructions", "links": { "openreview": "https://openreview.net/forum?id=RPRqKhjrr6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-105", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Blackbox Model Provenance via Palimpsestic Membership Inference", "authors": [ "Rohith Kuditipudi", "Jing Huang", "Sally Zhu", "Diyi Yang", "Christopher Potts", "Percy Liang" ], "affiliation": "", "summary": "Suppose Alice trains an open-weight language model and Bob uses a blackbox derivative of Alice’s model to produce text", "links": { "openreview": "https://openreview.net/forum?id=VRhVS59yhP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-106", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scaling can lead to compositional generalization", "authors": [ "Florian Redhardt", "Yassir Akram", "Simon Schug" ], "affiliation": "", "summary": "Can neural networks systematically capture discrete, compositional task structure despite their continuous, distributed nature? The impressive capabilities of large-scale neural networks suggest that", "links": { "openreview": "https://openreview.net/forum?id=hZt0daVIZi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-107", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning", "authors": [ "Jaehun Jung", "Seungju Han", "Ximing Lu", "Skyler Hallinan", "David Acuna", "Shrimai Prabhumoye", "Mostofa Patwary", "Mohammad Shoeybi", "Bryan Catanzaro", "Yejin Choi" ], "affiliation": "", "summary": "Data diversity is crucial for training a strong language model", "links": { "openreview": "https://openreview.net/forum?id=R0dC7Xzwbk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-108", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AnaCP: Toward Upper-Bound Continual Learning via Analytic Contrastive Projection", "authors": [ "Saleh Momeni", "Changnan Xiao", "Bing Liu" ], "affiliation": "", "summary": "This paper studies the problem of class-incremental learning (CIL), a core setting within continual learning where a model learns a sequence of tasks, each containing a distinct set of classes", "links": { "openreview": "https://openreview.net/forum?id=qQbvLU34F1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-109", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning", "authors": [ "Riccardo De Santi", "Marin Vlastelica", "Ya-Ping Hsieh", "Zebang Shen", "Niao He", "Andreas Krause" ], "affiliation": "", "summary": "Adapting large-scale foundational flow and diffusion generative models to optimize task-specific objectives while preserving prior information is crucial for real-world applications such as molecular", "links": { "openreview": "https://openreview.net/forum?id=JzCjNJlSxI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-110", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "What Expressivity Theory Misses: Message Passing Complexity for GNNs", "authors": [ "Niklas Kemper", "Tom Wollschläger", "Stephan Günnemann" ], "affiliation": "", "summary": "Expressivity theory, characterizing which graphs a GNN can distinguish, has become the predominant framework for analyzing GNNs, with new models striving for higher expressivity", "links": { "openreview": "https://openreview.net/forum?id=ATYKgiDqt5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-111", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Ambient Diffusion Omni: Training Good Models with Bad Data", "authors": [ "Giannis Daras", "Adrian Rodriguez-Munoz", "Adam Klivans", "Antonio Torralba", "Constantinos Costis Daskalakis" ], "affiliation": "", "summary": "We show how to use low-quality, synthetic, and out-of-distribution images to improve the quality of a diffusion model", "links": { "openreview": "https://openreview.net/forum?id=MVYz4GmcUH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-112", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UMA: A Family of Universal Models for Atoms", "authors": [ "Brandon M Wood", "Misko Dzamba", "Xiang Fu", "Meng Gao", "Muhammed Shuaibi", "Luis Barroso-Luque", "Kareem Abdelmaqsoud", "Vahe Gharakhanyan", "John R. Kitchin", "Daniel S. Levine", "Kyle Michel", "Anuroop Sriram", "Taco Cohen", "Abhishek Das", "Sushree Jagriti Sahoo", "Ammar Rizvi", "Zachary Ward Ulissi", "C. Lawrence Zitnick" ], "affiliation": "", "summary": "The ability to quickly and accurately compute properties from atomic simulations is critical for advancing a large number of applications in chemistry and materials science including drug discovery, e", "links": { "openreview": "https://openreview.net/forum?id=SvopaNxYWt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-113", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Best Instruction-Tuning Data are Those That Fit", "authors": [ "Dylan Zhang", "Qirun Dai", "Hao Peng" ], "affiliation": "", "summary": "High-quality supervised finetuning (SFT) data are essential for unlocking pretrained LLMs’ capabilities", "links": { "openreview": "https://openreview.net/forum?id=4jFSekBaDT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-114", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Forecasting in Offline Reinforcement Learning for Non-stationary Environments", "authors": [ "Suzan Ece Ada", "Georg Martius", "Emre Ugur", "Erhan Oztop" ], "affiliation": "", "summary": "Offline Reinforcement Learning (RL) provides a promising avenue for training policies from pre-collected datasets when gathering additional interaction data is infeasible", "links": { "openreview": "https://openreview.net/forum?id=24UJqxw1kv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-115", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Purifying Approximate Differential Privacy with Randomized Post-processing", "authors": [ "Yingyu Lin", "Erchi Wang", "Yian Ma", "Yu-Xiang Wang" ], "affiliation": "", "summary": "We propose a framework to convert $(\\varepsilon, \\delta)$-approximate Differential Privacy (DP) mechanisms into $(\\varepsilon', 0)$-pure DP mechanisms under certain conditions, a process we call ``pur", "links": { "openreview": "https://openreview.net/forum?id=hHn8xGTRKO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-116", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Trust Region Constrained Measure Transport in Path Space for Stochastic Optimal Control and Inference", "authors": [ "Denis Blessing", "Julius Berner", "Lorenz Richter", "Carles Domingo-Enrich", "Yuanqi Du", "Arash Vahdat", "Gerhard Neumann" ], "affiliation": "", "summary": "Solving stochastic optimal control problems with quadratic control costs can be viewed as approximating a target path space measure, e", "links": { "openreview": "https://openreview.net/forum?id=6RlbOEcOS4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-117", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "How Well Can Differential Privacy Be Audited in One Run?", "authors": [ "Amit Keinan", "Moshe Shenfeld", "Katrina Ligett" ], "affiliation": "", "summary": "Recent methods for auditing the privacy of machine learning algorithms have improved computational efficiency by simultaneously intervening on multiple training examples in a single training run", "links": { "openreview": "https://openreview.net/forum?id=004uTlSufe", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-118", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Counteractive RL: Rethinking Core Principles for Efficient and Scalable Deep Reinforcement Learning", "authors": [ "Ezgi Korkmaz" ], "affiliation": "", "summary": "Following the pivotal success of learning strategies to win at tasks, solely by interacting with an environment without any supervision, agents have gained the ability to make sequential decisions in", "links": { "openreview": "https://openreview.net/forum?id=qaHrpITIvB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-119", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Is Noise Conditioning Necessary? A Unified Theory of Unconditional Graph Diffusion Models", "authors": [ "Jipeng Li", "Yanning Shen" ], "affiliation": "", "summary": "Explicit noise-level conditioning is widely regarded as essential for the effective operation of Graph Diffusion Models (GDMs)", "links": { "openreview": "https://openreview.net/forum?id=bmoGGgSLzB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-120", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "An Evidence-Based Post-Hoc Adjustment Framework for Anomaly Detection Under Data Contamination", "authors": [ "Sukanya Patra", "Souhaib Ben Taieb" ], "affiliation": "", "summary": "Unsupervised anomaly detection (AD) methods typically assume clean training data, yet real-world datasets often contain undetected or mislabeled anomalies, leading to significant performance degradati", "links": { "openreview": "https://openreview.net/forum?id=NgLFQTBPRR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-121", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SHAP values via sparse Fourier representation", "authors": [ "Ali Gorji", "Andisheh Amrollahi", "Andreas Krause" ], "affiliation": "", "summary": "SHAP (SHapley Additive exPlanations) values are a widely used method for local feature attribution in interpretable and explainable AI", "links": { "openreview": "https://openreview.net/forum?id=wab4BEAUt6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-122", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Quantum Doubly Stochastic Transformers", "authors": [ "Jannis Born", "Filip Skogh", "Kahn Rhrissorrakrai", "Filippo Utro", "Nico Wagner", "Aleksandros Sobczyk" ], "affiliation": "", "summary": "At the core of the Transformer, the softmax normalizes the attention matrix to be right stochastic", "links": { "openreview": "https://openreview.net/forum?id=0EILv1HcmG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-123", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Reverse Engineering Human Preferences with Reinforcement Learning", "authors": [ "Lisa Alazraki", "Yi-Chern Tan", "Jon Ander Campos", "Maximilian Mozes", "Marek Rei", "Max Bartolo" ], "affiliation": "", "summary": "The capabilities of Large Language Models (LLMs) are routinely evaluated by other LLMs trained to predict human preferences", "links": { "openreview": "https://openreview.net/forum?id=heY0zzGvYm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-124", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Efficient Fairness-Performance Pareto Front Computation", "authors": [ "Mark Kozdoba", "Binyamin Perets", "Shie Mannor" ], "affiliation": "", "summary": "There is a well known intrinsic trade-off between the fairness of a representation and the performance of classifiers derived from the representation", "links": { "openreview": "https://openreview.net/forum?id=65oFEAP42P", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-125", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Curl Descent : Non-Gradient Learning Dynamics with Sign-Diverse Plasticity", "authors": [ "Hugo Ninou", "Jonathan Kadmon", "N Alex Cayco Gajic" ], "affiliation": "", "summary": "Gradient-based algorithms are a cornerstone of artificial neural network training, yet it remains unclear whether biological neural networks use similar gradient-based strategies during learning", "links": { "openreview": "https://openreview.net/forum?id=SB1CsuJ11a", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-126", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ElliCE: Efficient and Provably Robust Algorithmic Recourse via the Rashomon Sets", "authors": [ "Bohdan Turbal", "Iryna Voitsitska", "Lesia Semenova" ], "affiliation": "", "summary": "Machine learning models now influence decisions that directly affect people’s lives, making it important to understand not only their predictions, but also how individuals could act to obtain better r", "links": { "openreview": "https://openreview.net/forum?id=GNDgQie8W4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-127", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Head Pursuit: Probing Attention Specialization in Multimodal Transformers", "authors": [ "Lorenzo Basile", "Valentino Maiorca", "Diego Doimo", "Francesco Locatello", "Alberto Cazzaniga" ], "affiliation": "", "summary": "Language and vision-language models have shown impressive performance across a wide range of tasks, but their internal mechanisms remain only partly understood", "links": { "openreview": "https://openreview.net/forum?id=WQ9rnkaUWm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-128", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards a Golden Classifier-Free Guidance Path via Foresight Fixed Point Iterations", "authors": [ "Kaibo Wang", "Jianda Mao", "Tong Wu", "Yang Xiang" ], "affiliation": "", "summary": "Classifier-Free Guidance (CFG) is an essential component of text-to-image diffusion models, and understanding and advancing its operational mechanisms remains a central focus of research", "links": { "openreview": "https://openreview.net/forum?id=yf8O4xEB4T", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-129", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Credal Prediction based on Relative Likelihood", "authors": [ "Timo Löhr", "Paul Hofman", "Felix Mohr", "Eyke Hüllermeier" ], "affiliation": "", "summary": "Predictions in the form of sets of probability distributions, so-called credal sets, provide a suitable means to represent a learner's epistemic uncertainty", "links": { "openreview": "https://openreview.net/forum?id=rKM3oqruN3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-130", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Power of Iterative Filtering for Supervised Learning with (Heavy) Contamination", "authors": [ "Adam Klivans", "Konstantinos Stavropoulos", "Kevin Tian", "Arsen Vasilyan" ], "affiliation": "", "summary": "Inspired by recent work on learning with distribution shift, we give a general outlier removal algorithm called *iterative polynomial filtering* and show a number of striking applications for supervis", "links": { "openreview": "https://openreview.net/forum?id=E7knuYAvpt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-131", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Light-Weight Diffusion Multiplier and Uncertainty Quantification for Fourier Neural Operators", "authors": [ "Albert Matveev", "Sanmitra Ghosh", "Aamal Hussain", "James-Michael Leahy", "Michalis Michaelides" ], "affiliation": "", "summary": "Operator learning is a powerful paradigm for solving partial differential equations, with Fourier Neural Operators serving as a widely adopted foundation", "links": { "openreview": "https://openreview.net/forum?id=YXSKYFZweV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-132", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Discrete Spatial Diffusion: Intensity-Preserving Diffusion Modeling", "authors": [ "Javier E. Santos", "Agnese Marcato", "Roman Colman", "Nicholas Lubbers", "Yen Ting Lin" ], "affiliation": "", "summary": "Generative diffusion models have achieved remarkable success in producing high-quality images", "links": { "openreview": "https://openreview.net/forum?id=kz3w2A2y0e", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-133", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CTRL-ALT-DECEIT Sabotage Evaluations for Automated AI R&D", "authors": [ "Francis Rhys Ward", "Teun van der Weij", "Hanna Gábor", "Sam Martin", "Raja Mehta Moreno", "Harel Lidar", "Louis Makower", "Thomas Jodrell", "Lauren Robson" ], "affiliation": "", "summary": "AI systems are increasingly able to autonomously conduct realistic software engineering tasks, and may soon be deployed to automate machine learning (ML) R\\&D itself", "links": { "openreview": "https://openreview.net/forum?id=XBMjXb6f4w", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-134", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "An Analytical Theory of Spectral Bias in the Learning Dynamics of Diffusion Models", "authors": [ "Binxu Wang", "Cengiz Pehlevan" ], "affiliation": "", "summary": "We develop an analytical framework for understanding how the learned distribution evolves during diffusion model training", "links": { "openreview": "https://openreview.net/forum?id=SDhOClkyqC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-135", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DeepHalo: A Neural Choice Model with Controllable Context Effects", "authors": [ "Shuhan Zhang", "Zhi Wang", "Rui Gao", "Shuang Li" ], "affiliation": "", "summary": "Modeling human decision-making is central to applications such as recommendation, preference learning, and human-AI alignment", "links": { "openreview": "https://openreview.net/forum?id=x9XepNPGJ5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-136", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Multi-Agent Learning under Uncertainty: Recurrence vs. Concentration", "authors": [ "Kyriakos Lotidis", "Panayotis Mertikopoulos", "Nicholas Bambos", "Jose Blanchet" ], "affiliation": "", "summary": "In this paper, we examine the convergence landscape of multi-agent learning under uncertainty", "links": { "openreview": "https://openreview.net/forum?id=CrwzbjO3aU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-137", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Quantum speedup of non-linear Monte Carlo problems", "authors": [ "Jose Blanchet", "Yassine Hamoudi", "Mario Szegedy", "Guanyang Wang" ], "affiliation": "", "summary": "The mean of a random variable can be understood as a *linear* functional on the space of probability distributions", "links": { "openreview": "https://openreview.net/forum?id=IkfBLlYuHA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-138", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fine-grained List-wise Alignment for Generative Medication Recommendation", "authors": [ "Chenxiao Fan", "Chongming Gao", "Wentao Shi", "Yaxin Gong", "Zhao Zihao", "Fuli Feng" ], "affiliation": "", "summary": "Accurate and safe medication recommendations are critical for effective clinical decision-making, especially in multimorbidity cases", "links": { "openreview": "https://openreview.net/forum?id=Quo3XadYcZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-139", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach", "authors": [ "Jonas Geiping", "Sean Michael McLeish", "Neel Jain", "John Kirchenbauer", "Siddharth Singh", "Brian R. Bartoldson", "Bhavya Kailkhura", "Abhinav Bhatele", "Tom Goldstein" ], "affiliation": "", "summary": "We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space", "links": { "openreview": "https://openreview.net/forum?id=S3GhJooWIC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-140", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Diffusion-Based Hierarchical Graph Neural Networks for Simulating Nonlinear Solid Mechanics", "authors": [ "Tobias Würth", "Niklas Freymuth", "Gerhard Neumann", "Luise Kärger" ], "affiliation": "", "summary": "Graph-based learned simulators have emerged as a promising approach for simulating physical systems on unstructured meshes, offering speed and generalization across diverse geometries", "links": { "openreview": "https://openreview.net/forum?id=MfBw0dlBfi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-141", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Breaking the Batch Barrier (B3) of Contrastive Learning via Smart Batch Mining", "authors": [ "Raghuveer Thirukovalluru", "Rui Meng", "Ye Liu", "Karthikeyan K", "Mingyi Su", "Ping Nie", "Semih Yavuz", "Yingbo Zhou", "Wenhu Chen", "Bhuwan Dhingra" ], "affiliation": "", "summary": "Contrastive learning (CL) is a prevalent technique for training embedding models, which pulls semantically similar examples (positives) closer in the representation space while pushing dissimilar ones", "links": { "openreview": "https://openreview.net/forum?id=M8zmlixh9y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-142", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Dynamic Algorithm for Explainable $k$-medians Clustering under $\\ell_p$ Norm", "authors": [ "Konstantin Makarychev", "Ilias Papanikolaou", "Liren Shan" ], "affiliation": "", "summary": "We study the problem of explainable $k$-medians clustering introduced by Dasgupta, Frost, Moshkovitz, and Rashtchian (2020)", "links": { "openreview": "https://openreview.net/forum?id=T1GXVrXJR4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-143", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Utility Engineering: Analyzing and Controlling Emergent Value Systems in AIs", "authors": [ "Mantas Mazeika", "Xuwang Yin", "Rishub Tamirisa", "Jaehyuk Lim", "Bruce W. Lee", "Richard Ren", "Long Phan", "Norman Mu", "Oliver Zhang", "Dan Hendrycks" ], "affiliation": "", "summary": "As AIs rapidly advance and become more agentic, the risk they pose is governed not only by their capabilities but increasingly by their propensities, including goals and values", "links": { "openreview": "https://openreview.net/forum?id=x9vcgXmRD0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-144", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Is the acquisition worth the cost? Surrogate losses for Consistent Two-stage Classifiers", "authors": [ "Florence Regol", "Joseph Cotnareanu", "Theodore Glavas", "Mark Coates" ], "affiliation": "", "summary": "Recent years have witnessed the emergence of a spectrum of foundation models, covering a broad range of capabilities and costs", "links": { "openreview": "https://openreview.net/forum?id=X0Etmtge6w", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-145", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training", "authors": [ "Jintao Zhang", "Jia wei", "Haoxu Wang", "Pengle Zhang", "Xiaoming Xu", "Haofeng Huang", "Kai Jiang", "Jianfei Chen", "Jun Zhu" ], "affiliation": "", "summary": "The efficiency of attention is important due to its quadratic time complexity", "links": { "openreview": "https://openreview.net/forum?id=JbJVWljk7r", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-146", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Eluder dimension: localise it!", "authors": [ "Alireza Bakhtiari", "Alex Ayoub", "Samuel McLaughlin Robertson", "David Janz", "Csaba Szepesvari" ], "affiliation": "", "summary": "We establish a lower bound on the eluder dimension in generalised linear model classes, showing that standard eluder dimension-based analysis cannot lead to first-order regret bounds", "links": { "openreview": "https://openreview.net/forum?id=e8R0ytPhLv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-147", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Communication-Efficient Language Model Training Scales Reliably and Robustly: Scaling Laws for DiLoCo", "authors": [ "Zachary Charles", "Gabriel Teston", "Lucio M. Dery", "J Keith Rush", "Nova Fallen", "Zachary Garrett", "Arthur Szlam", "Arthur Douillard" ], "affiliation": "", "summary": "As we scale to more massive machine learning models, the frequent synchronization demands inherent in data-parallel approaches create significant slowdowns, posing a critical challenge to further scal", "links": { "openreview": "https://openreview.net/forum?id=X4SCxcgb3O", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-148", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FlashMD: long-stride, universal prediction of molecular dynamics", "authors": [ "Filippo Bigi", "Sanggyu Chong", "Agustinus Kristiadi", "Michele Ceriotti" ], "affiliation": "", "summary": "Molecular dynamics (MD) provides insights into atomic-scale processes by integrating over time the equations that describe the motion of atoms under the action of interatomic forces", "links": { "openreview": "https://openreview.net/forum?id=ogZu06NgQs", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-149", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Not All Data are Good Labels: On the Self-supervised Labeling for Time Series Forecasting", "authors": [ "Yuxuan Yang", "Dalin Zhang", "Yuxuan Liang", "Hua Lu", "Gang Chen", "Huan Li" ], "affiliation": "", "summary": "Time Series Forecasting (TSF) is a crucial task in various domains, yet existing TSF models rely heavily on high-quality data and insufficiently exploit all available data", "links": { "openreview": "https://openreview.net/forum?id=kQokjfoGjk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-150", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "3D Equivariant Visuomotor Policy Learning via Spherical Projection", "authors": [ "Boce Hu", "Dian Wang", "David Klee", "Heng Tian", "Xupeng Zhu", "Haojie Huang", "Robert Platt", "Robin Walters" ], "affiliation": "", "summary": "Equivariant models have recently been shown to improve the data efficiency of diffusion policy by a significant margin", "links": { "openreview": "https://openreview.net/forum?id=kXJd4JxF34", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-151", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Predictable Scale (Part II) --- Farseer: A Refined Scaling Law in LLMs", "authors": [ "Houyi Li", "Wenzhen Zheng", "Qiufeng Wang", "Zhenyu Ding", "Haoying Wang", "Zili Wang", "Shijie Xuyang", "Ning Ding", "Shuigeng Zhou", "Xiangyu Zhang", "Daxin Jiang" ], "affiliation": "", "summary": "Training Large Language Models (LLMs) is prohibitively expensive, creating a critical scaling gap where insights from small-scale experiments often fail to transfer to resource-intensive production sy", "links": { "openreview": "https://openreview.net/forum?id=2Gnp8sdwVe", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-152", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Adaptive Prediction-Powered AutoEval with Reliability and Efficiency Guarantees", "authors": [ "Sangwoo Park", "Matteo Zecchin", "Osvaldo Simeone" ], "affiliation": "", "summary": "Selecting artificial intelligence (AI) models, such as large language models (LLMs), from multiple candidates requires accurate performance estimation", "links": { "openreview": "https://openreview.net/forum?id=nfhmjdZUbQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-153", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Ridge Boosting is Both Robust and Efficient", "authors": [ "David Bruns-Smith", "Zhongming Xie", "Avi Feller" ], "affiliation": "", "summary": "Estimators in statistics and machine learning must typically trade off between efficiency, having low variance for a fixed target, and distributional robustness, such as \\textit{multiaccuracy}, or hav", "links": { "openreview": "https://openreview.net/forum?id=mf0p4PO7ko", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-154", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Unifying View of Linear Function Approximation in Off-Policy RL Through Matrix Splitting and Preconditioning", "authors": [ "Zechen Wu", "Amy Greenwald", "Ronald Parr" ], "affiliation": "", "summary": "In off-policy policy evaluation (OPE) tasks within reinforcement learning, Temporal Difference Learning(TD) and Fitted Q-Iteration (FQI) have traditionally been viewed as differing in the number of up", "links": { "openreview": "https://openreview.net/forum?id=pH3daDPj4c", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-155", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Generalizable Reasoning through Compositional Energy Minimization", "authors": [ "Alexandru Oarga", "Yilun Du" ], "affiliation": "", "summary": "Generalization is a key challenge in machine learning, specifically in reasoning tasks, where models are expected to solve problems more complex than those encountered during training", "links": { "openreview": "https://openreview.net/forum?id=5k0AHYc4MJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-156", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Universal Sequence Preconditioning", "authors": [ "Annie Marsden", "Elad Hazan" ], "affiliation": "", "summary": "We study the problem of preconditioning in the setting of sequential prediction", "links": { "openreview": "https://openreview.net/forum?id=rwmVd8BKW5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-157", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Enigmata: Scaling Logical Reasoning in Large Language Models with Synthetic Verifiable Puzzles", "authors": [ "Jiangjie Chen", "Qianyu He", "Siyu Yuan", "Aili Chen", "Zhicheng Cai", "Weinan Dai", "Hongli Yu", "Jiaze Chen", "Xuefeng Li", "Qiying Yu", "Hao Zhou", "Mingxuan Wang" ], "affiliation": "", "summary": "Large Language Models (LLMs), such as OpenAI’s o1 and DeepSeek’s R1, excel at advanced reasoning tasks like math and coding via Reinforcement Learning with Verifiable Rewards (RLVR), but still struggl", "links": { "openreview": "https://openreview.net/forum?id=fmnxunacr4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-158", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Memory-Enhanced Neural Solvers for Routing Problems", "authors": [ "Felix Chalumeau", "Refiloe Shabe", "Noah De Nicola", "Arnu Pretorius", "Thomas D Barrett", "Nathan Grinsztajn" ], "affiliation": "", "summary": "Routing Problems are central to many real-world applications, yet remain challenging due to their (NP-)hard nature", "links": { "openreview": "https://openreview.net/forum?id=p7WHZy8TCG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-159", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Uni-MuMER: Unified Multi-Task Fine-Tuning of Vision-Language Model for Handwritten Mathematical Expression Recognition", "authors": [ "Yu Li", "Jin Jiang", "Jianhua Zhu", "Shuai Peng", "Baole Wei", "Yuxuan Zhou", "Liangcai Gao" ], "affiliation": "", "summary": "Handwritten Mathematical Expression Recognition (HMER) remains a persistent challenge in Optical Character Recognition (OCR) due to the inherent freedom of symbol layouts and variability in handwritin", "links": { "openreview": "https://openreview.net/forum?id=oHbVboLXz6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-160", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Two Heads are Better than One: Simulating Large Transformers with Small Ones", "authors": [ "Hantao Yu", "Josh Alman" ], "affiliation": "", "summary": "The quadratic complexity of self‑attention prevents transformers from scaling effectively to long input sequences", "links": { "openreview": "https://openreview.net/forum?id=Xeb2EYBKkr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-161", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Gaze Beyond the Frame: Forecasting Egocentric 3D Visual Span", "authors": [ "Heeseung Yun", "Joonil Na", "Jaeyeon Kim", "Calvin Murdock", "Gunhee Kim" ], "affiliation": "", "summary": "People continuously perceive and interact with their surroundings based on underlying intentions that drive their exploration and behaviors", "links": { "openreview": "https://openreview.net/forum?id=8rKSfL3GsK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-162", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Unifying Proportional Fairness in Centroid and Non-Centroid Clustering", "authors": [ "Benjamin Cookson", "Nisarg Shah", "Ziqi Yu" ], "affiliation": "", "summary": "Proportional fairness criteria inspired by democratic ideals of proportional representation have received growing attention in the clustering literature", "links": { "openreview": "https://openreview.net/forum?id=o7Z8TClGjp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-163", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ARIA: Training Language Agents with Intention-driven Reward Aggregation", "authors": [ "Ruihan Yang", "Yikai Zhang", "Aili Chen", "Xintao Wang", "Jiangjie Chen", "Siyu Yuan", "Deqing Yang", "Yanghua Xiao" ], "affiliation": "", "summary": "Large language models (LLMs) have enabled agents to perform complex reasoning and decision-making through free-form language interactions", "links": { "openreview": "https://openreview.net/forum?id=eumRwpgdMU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-164", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Reasoning Planning for Language Models", "authors": [ "Bao Nguyen", "Hieu Trung Nguyen", "Ruifeng She", "Xiaojin Fu", "Viet Anh Nguyen" ], "affiliation": "", "summary": "Selecting an appropriate reasoning method for a given query remains a key challenge in language model generation", "links": { "openreview": "https://openreview.net/forum?id=QFjssnKdBI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-165", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "When Less Language is More: Language-Reasoning Disentanglement Makes LLMs Better Multilingual Reasoners", "authors": [ "Weixiang Zhao", "Jiahe Guo", "Yang Deng", "Tongtong Wu", "Wenxuan Zhang", "Yulin Hu", "Xingyu Sui", "Yanyan Zhao", "Wanxiang Che", "Bing Qin", "Tat-Seng Chua", "Ting Liu" ], "affiliation": "", "summary": "Multilingual reasoning remains a significant challenge for large language models (LLMs), with performance disproportionately favoring high-resource languages", "links": { "openreview": "https://openreview.net/forum?id=fleQlZ2VTx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-166", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "EF-3DGS: Event-Aided Free-Trajectory 3D Gaussian Splatting", "authors": [ "Bohao Liao", "Wei Zhai", "Zengyu Wan", "Zhixin Cheng", "Wenfei Yang", "Yang Cao", "Tianzhu Zhang", "Zheng-Jun Zha" ], "affiliation": "", "summary": "Scene reconstruction from casually captured videos has wide real-world applications", "links": { "openreview": "https://openreview.net/forum?id=shFhW4zqd6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-167", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Beyond Scalar Rewards: An Axiomatic Framework for Lexicographic MDPs", "authors": [ "Mehran Shakerinava", "Siamak Ravanbakhsh", "Adam Oberman" ], "affiliation": "", "summary": "Recent work has formalized the reward hypothesis through the lens of expected utility theory, by interpreting reward as utility", "links": { "openreview": "https://openreview.net/forum?id=cUy3tYIOS5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-168", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Compress to Impress: Efficient LLM Adaptation Using a Single Gradient Step on 100 Samples", "authors": [ "Shiva Sreeram", "Alaa Maalouf", "Pratyusha Sharma", "Daniela Rus" ], "affiliation": "", "summary": "Recently, Sharma et al", "links": { "openreview": "https://openreview.net/forum?id=tXxsCbKdQv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-169", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Understanding Prompt Tuning and In-Context Learning via Meta-Learning", "authors": [ "Tim Genewein", "Li Kevin Wenliang", "Jordi Grau-Moya", "Anian Ruoss", "Laurent Orseau", "Marcus Hutter" ], "affiliation": "", "summary": "Prompting is one of the main ways to adapt a pretrained model to target tasks", "links": { "openreview": "https://openreview.net/forum?id=avRktRfQ8c", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-170", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Error Broadcast and Decorrelation as a Potential Artificial and Natural Learning Mechanism", "authors": [ "Mete Erdogan", "Cengiz Pehlevan", "Alper Tunga Erdogan" ], "affiliation": "", "summary": "We introduce *Error Broadcast and Decorrelation* (EBD), a novel learning framework for neural networks that addresses credit assignment by directly broadcasting output errors to individual layers, cir", "links": { "openreview": "https://openreview.net/forum?id=IZ1KYTU9ON", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-171", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Expressive Power of Mixture-of-Experts for Structured Complex Tasks", "authors": [ "Mingze Wang", "Weinan E" ], "affiliation": "", "summary": "Mixture-of-experts networks (MoEs) have demonstrated remarkable efficiency in modern deep learning", "links": { "openreview": "https://openreview.net/forum?id=zSrb8rtH9M", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-172", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TokenSwap: A Lightweight Method to Disrupt Memorized Sequences in LLMs", "authors": [ "Parjanya Prajakta Prashant", "Kaustubh Ponkshe", "Babak Salimi" ], "affiliation": "", "summary": "As language models scale, their performance improves dramatically across a wide range of tasks, but so does their tendency to memorize and regurgitate parts of their training data verbatim", "links": { "openreview": "https://openreview.net/forum?id=gNiT81iag0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-173", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Streaming Attention Approximation via Discrepancy Theory", "authors": [ "Ekaterina Kochetkova", "Kshiteej Sheth", "Insu Han", "Amir Zandieh", "Michael Kapralov" ], "affiliation": "", "summary": "Large language models (LLMs) have achieved impressive success, but their high memory requirements present challenges for long-context token generation", "links": { "openreview": "https://openreview.net/forum?id=p3HBEtNDRY", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-174", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SegMASt3R: Geometry Grounded Segment Matching", "authors": [ "Rohit Jayanti", "Swayam Agrawal", "Vansh Garg", "Siddharth Tourani", "Muhammad Haris Khan", "Sourav Garg", "Madhava Krishna" ], "affiliation": "", "summary": "Segment matching is an important intermediate task in computer vision that establishes correspondences between semantically or geometrically coherent regions across images", "links": { "openreview": "https://openreview.net/forum?id=DI2AAFnLrc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-175", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Among Us: A Sandbox for Measuring and Detecting Agentic Deception", "authors": [ "Satvik Golechha", "Adrià Garriga-Alonso" ], "affiliation": "", "summary": "Prior studies on deception in language-based AI agents typically assess whether the agent produces a false statement about a topic, or makes a binary choice prompted by a goal, rather than allowing op", "links": { "openreview": "https://openreview.net/forum?id=XP3v1THxsq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-176", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AutoToM: Scaling Model-based Mental Inference via Automated Agent Modeling", "authors": [ "Zhining Zhang", "Chuanyang Jin", "Mung Yao Jia", "Shunchi Zhang", "Tianmin Shu" ], "affiliation": "", "summary": "Theory of Mind (ToM), the ability to understand people's minds based on their behavior, is key to developing socially intelligent agents", "links": { "openreview": "https://openreview.net/forum?id=oeZZusZheP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-177", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Interpretable and Efficient Attention: Compressing All by Contracting a Few", "authors": [ "Qishuai Wen", "Zhiyuan Huang", "Chun-Guang Li" ], "affiliation": "", "summary": "Attention mechanisms have achieved significant empirical success in multiple fields, but their underlying optimization objectives remain unclear yet", "links": { "openreview": "https://openreview.net/forum?id=6SI1pvb5xl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-178", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "EvoBrain: Dynamic Multi-Channel EEG Graph Modeling for Time-Evolving Brain Networks", "authors": [ "Rikuto Kotoge", "Zheng Chen", "Tasuku Kimura", "Yasuko Matsubara", "Takufumi Yanagisawa", "Haruhiko Kishima", "Yasushi Sakurai" ], "affiliation": "", "summary": "Dynamic GNNs, which integrate temporal and spatial features in Electroencephalography (EEG) data, have shown great potential in automating seizure detection", "links": { "openreview": "https://openreview.net/forum?id=XmV7KRABBl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-179", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Product Distribution Learning with Imperfect Advice", "authors": [ "Arnab Bhattacharyya", "Davin Choo", "Philips George John", "Themis Gouleakis" ], "affiliation": "", "summary": "Given i", "links": { "openreview": "https://openreview.net/forum?id=idjZKbf78s", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-180", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RoboScape: Physics-informed Embodied World Model", "authors": [ "Yu Shang", "Xin Zhang", "Yinzhou Tang", "Lei Jin", "Chen Gao", "Wei Wu", "Yong Li" ], "affiliation": "", "summary": "World models have become indispensable tools for embodied intelligence, serving as powerful simulators capable of generating realistic robotic videos while addressing critical data scarcity challenges", "links": { "openreview": "https://openreview.net/forum?id=wbZCBBrq3W", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-181", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "InstructHOI: Context-Aware Instruction for Multi-Modal Reasoning in Human-Object Interaction Detection", "authors": [ "Jinguo Luo", "Weihong Ren", "Quanlong Zheng", "Yanhao Zhang", "Zhenlong Yuan", "Zhiyong Wang", "Haonan Lu", "Honghai LIU" ], "affiliation": "", "summary": "Recently, Large Foundation Models (LFMs), e", "links": { "openreview": "https://openreview.net/forum?id=WjYvHSjXrP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-182", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ARM: Adaptive Reasoning Model", "authors": [ "Siye Wu", "Jian Xie", "Yikai Zhang", "Aili Chen", "Kai Zhang", "Yu Su", "Yanghua Xiao" ], "affiliation": "", "summary": "While large reasoning models demonstrate strong performance on complex tasks, they lack the ability to adjust reasoning token usage based on task difficulty", "links": { "openreview": "https://openreview.net/forum?id=z9oeQrcNh9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-183", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Computational Efficiency under Covariate Shift in Kernel Ridge Regression", "authors": [ "Andrea Della Vecchia", "Arnaud Mavakala Watusadisi", "Ernesto De Vito", "Lorenzo Rosasco" ], "affiliation": "", "summary": "This paper addresses the covariate shift problem in the context of nonparametric regression within reproducing kernel Hilbert spaces (RKHSs)", "links": { "openreview": "https://openreview.net/forum?id=TU2MZHZLkP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-184", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Self-Assembling Graph Perceptrons", "authors": [ "Jialong Chen", "Tong Wang", "Bowen Deng", "Luonan Chen", "Zibin Zheng", "Chuan Chen" ], "affiliation": "", "summary": "Inspired by the workings of biological brains, humans have designed artificial neural networks (ANNs), sparking profound advancements across various fields", "links": { "openreview": "https://openreview.net/forum?id=5iHDGJFf49", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-185", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scaling and context steer LLMs along the same computational path as the human brain", "authors": [ "Joséphine Raugel", "Jérémy Rapin", "Stéphane d'Ascoli", "Valentin Wyart", "Jean-Remi King" ], "affiliation": "", "summary": "Recent studies suggest that the representations learned by large language models (LLMs) are partially aligned to those of the human brain", "links": { "openreview": "https://openreview.net/forum?id=4YKlo58RcQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-186", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "InfiFPO: Implicit Model Fusion via Preference Optimization in Large Language Models", "authors": [ "Yanggan Gu", "Yuanyi Wang", "Zhaoyi Yan", "Yiming Zhang", "Qi Zhou", "Fei Wu", "Hongxia Yang" ], "affiliation": "", "summary": "Model fusion combines multiple Large Language Models (LLMs) with different strengths into a more powerful, integrated model through lightweight training methods", "links": { "openreview": "https://openreview.net/forum?id=ZwBtDbuzjY", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-187", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SViMo: Synchronized Diffusion for Video and Motion Generation in Hand-object Interaction Scenarios", "authors": [ "Lingwei Dang", "Ruizhi Shao", "Hongwen Zhang", "Wei MIN", "Yebin Liu", "Qingyao Wu" ], "affiliation": "", "summary": "Hand-Object Interaction (HOI) generation has significant application potential", "links": { "openreview": "https://openreview.net/forum?id=huZzy5w2Js", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-188", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "HyPINO: Multi-Physics Neural Operators via HyperPINNs and the Method of Manufactured Solutions", "authors": [ "Rafael Bischof", "Michal Piovarci", "Michael Anton Kraus", "Siddhartha Mishra", "Bernd Bickel" ], "affiliation": "", "summary": "We present HyPINO, a multi-physics neural operator designed for zero-shot generalization across a broad class of PDEs without requiring task-specific fine-tuning", "links": { "openreview": "https://openreview.net/forum?id=W1Cu6JsRsd", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-189", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RidgeLoRA: Matrix Ridge Enhanced Low-Rank Adaptation of Large Language Models", "authors": [ "Junda Zhu", "Jun Ai", "Yujun Li", "Yichun Yin", "Yasheng Wang", "Lifeng Shang", "Qun Liu" ], "affiliation": "", "summary": "As one of the state-of-the-art parameter-efficient fine-tuning~(PEFT) methods, Low-Rank Adaptation (LoRA) enables model optimization with reduced computational cost through trainable low-rank matrix", "links": { "openreview": "https://openreview.net/forum?id=0RF80tUWuv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-190", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens", "authors": [ "Xixian Yong", "Xiao Zhou", "Yingying Zhang", "Jinlin Li", "Yefeng Zheng", "Xian Wu" ], "affiliation": "", "summary": "The recent rise of Large Reasoning Models (LRMs) has significantly improved multi-step reasoning performance, but often at the cost of generating excessively long reasoning chains", "links": { "openreview": "https://openreview.net/forum?id=DpOSndSOZz", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-191", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Hyperbolic Fine-Tuning for Large Language Models", "authors": [ "Menglin Yang", "Ram Samarth B B", "Aosong Feng", "Bo Xiong", "Jiahong Liu", "Irwin King", "Rex Ying" ], "affiliation": "", "summary": "Large language models (LLMs) have demonstrated remarkable performance across various tasks", "links": { "openreview": "https://openreview.net/forum?id=TkEdQv0bXB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-192", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Deep Continuous-Time State-Space Models for Marked Event Sequences", "authors": [ "Yuxin Chang", "Alex James Boyd", "Cao Xiao", "Taha Kass-Hout", "Parminder Bhatia", "Padhraic Smyth", "Andrew Warrington" ], "affiliation": "", "summary": "Marked temporal point processes (MTPPs) model sequences of events occurring at irregular time intervals, with wide-ranging applications in fields such as healthcare, finance and social networks", "links": { "openreview": "https://openreview.net/forum?id=74SvE2GZwW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-193", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "HBLLM: Wavelet-Enhanced High-Fidelity 1-Bit Quantization for LLMs", "authors": [ "Ningning CHEN", "Weicai Ye", "Ying Jiang" ], "affiliation": "", "summary": "We introduce HBLLM, a wavelet-enhanced high-fidelity $1$-bit post-training quantization method for Large Language Models (LLMs)", "links": { "openreview": "https://openreview.net/forum?id=3FsM6wWQL4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-194", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Which Algorithms Have Tight Generalization Bounds?", "authors": [ "Michael Gastpar", "Ido Nachum", "Jonathan Shafer", "Thomas Weinberger" ], "affiliation": "", "summary": "We study which machine learning algorithms have tight generalization bounds with respect to a given collection of population distributions", "links": { "openreview": "https://openreview.net/forum?id=qXAABCxYQ2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-195", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Axial Neural Networks for Dimension-Free Foundation Models", "authors": [ "Hyunsu Kim", "Jonggeon Park", "Joan Bruna", "Hongseok Yang", "Juho Lee" ], "affiliation": "", "summary": "The advent of foundation models in AI has significantly advanced general-purpose learning, enabling remarkable capabilities in zero-shot inference and in-context learning", "links": { "openreview": "https://openreview.net/forum?id=cLQlsOGqbM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-196", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Horizon Reduction Makes RL Scalable", "authors": [ "Seohong Park", "Kevin Frans", "Deepinder Mann", "Benjamin Eysenbach", "Aviral Kumar", "Sergey Levine" ], "affiliation": "", "summary": "In this work, we study the scalability of offline reinforcement learning (RL) algorithms", "links": { "openreview": "https://openreview.net/forum?id=hguaupzLCU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-197", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OpenBox: Annotate Any Bounding Boxes in 3D", "authors": [ "In-Jae Lee", "Mungyeom Kim", "Kwonyoung Ryu", "Pierre Musacchio", "Jaesik Park" ], "affiliation": "", "summary": "Unsupervised and open-vocabulary 3D object detection has recently gained attention, particularly in autonomous driving, where reducing annotation costs and recognizing unseen objects are critical for", "links": { "openreview": "https://openreview.net/forum?id=7ieZWCc7rB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-198", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SGCD: Stain-Guided CycleDiffusion for Unsupervised Domain Adaptation of Histopathology Image Classification", "authors": [ "Hsi-Ling Chen", "Chun-Shien Lu", "Pau-Choo Chung" ], "affiliation": "", "summary": "The effectiveness of domain translation in addressing image-based problems of Unsupervised Domain Adaptation (UDA) depends on the quality of the translated images and the preservation of crucial discr", "links": { "openreview": "https://openreview.net/forum?id=z2SGaPIhLT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-199", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FFN Fusion: Rethinking Sequential Computation in Large Language Models", "authors": [ "Akhiad Bercovich", "Mohammed Dabbah", "Omri Puny", "Ido Galil", "Amnon Geifman", "Yonatan Geifman", "Izhak Golan", "Ehud Dov Karpas", "Itay Levy", "Zach Moshe", "Najeeb Nabwani", "Tomer Ronen", "Itamar Schen", "Ido Shahaf", "Oren Tropp", "Ran Zilberstein", "Ran El-Yaniv" ], "affiliation": "", "summary": "We introduce \\textit{FFN Fusion}, an architectural optimization technique that reduces sequential computation in large language models by identifying and exploiting natural opportunities for paralleli", "links": { "openreview": "https://openreview.net/forum?id=XUmGMBRv4M", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-200", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Distilling LLM Agent into Small Models with Retrieval and Code Tools", "authors": [ "Minki Kang", "Jongwon Jeong", "Seanie Lee", "Jaewoong Cho", "Sung Ju Hwang" ], "affiliation": "", "summary": "Large language models (LLMs) excel at complex reasoning tasks but remain computationally expensive, limiting their practical deployment", "links": { "openreview": "https://openreview.net/forum?id=VkicTqszOn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-201", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Understanding Parametric and Contextual Knowledge Reconciliation within Large Language Models", "authors": [ "Jun Zhao", "Yongzhuo Yang", "Xiang Hu", "Jingqi Tong", "Yi Lu", "Wei Wu", "Tao Gui", "Qi Zhang", "Xuanjing Huang" ], "affiliation": "", "summary": "Retrieval-Augmented Generation (RAG) provides additional contextual knowledge to complement the parametric knowledge in Large Language Models (LLMs)", "links": { "openreview": "https://openreview.net/forum?id=76cFMRgEzQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-202", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Differentiable Decision Tree via \"ReLU+Argmin\" Reformulation", "authors": [ "Qiangqiang Mao", "Jiayang Ren", "Yixiu Wang", "Chenxuanyin Zou", "Jingjing Zheng", "Yankai Cao" ], "affiliation": "", "summary": "Decision tree, despite its unmatched interpretability and lightweight structure, faces two key issues that limit its broader applicability: non-differentiability and low testing accuracy", "links": { "openreview": "https://openreview.net/forum?id=F11iEhKoYp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-203", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning", "authors": [ "Borong Zhang", "Yuhao Zhang", "Jiaming Ji", "Yingshan Lei", "Josef Dai", "Yuanpei Chen", "Yaodong Yang" ], "affiliation": "", "summary": "Vision-language-action models (VLAs) show potential as generalist robot policies", "links": { "openreview": "https://openreview.net/forum?id=dt940loCBT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-204", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "3D Interaction Geometric Pre-training for Molecular Relational Learning", "authors": [ "Namkyeong Lee", "Yunhak Oh", "Heewoong Noh", "Gyoung S. Na", "Minkai Xu", "Hanchen", "Tianfan Fu", "Chanyoung Park" ], "affiliation": "", "summary": "Molecular Relational Learning (MRL) is a rapidly growing field that focuses on understanding the interaction dynamics between molecules, which is crucial for applications ranging from catalyst enginee", "links": { "openreview": "https://openreview.net/forum?id=PZaxCfLGLA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-205", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Improved Bounds for Swap Multicalibration and Swap Omniprediction", "authors": [ "Haipeng Luo", "Spandan Senapati", "Vatsal Sharan" ], "affiliation": "", "summary": "In this paper, we consider the related problems of multicalibration --- a multigroup fairness notion and omniprediction --- a simultaneous loss minimization paradigm, both in the distributional and on", "links": { "openreview": "https://openreview.net/forum?id=pXoiIDdynI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-206", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UniteFormer: Unifying Node and Edge Modalities in Transformers for Vehicle Routing Problems", "authors": [ "Dian Meng", "Zhiguang Cao", "Jie Gao", "Yaoxin Wu", "Yaqing Hou" ], "affiliation": "", "summary": "Neural solvers for the Vehicle Routing Problem (VRP) have typically relied on either node or edge inputs, limiting their flexibility and generalization in real-world scenarios", "links": { "openreview": "https://openreview.net/forum?id=BRklmFlCsD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-207", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Simultaneous Swap Regret Minimization via KL-Calibration", "authors": [ "Haipeng Luo", "Spandan Senapati", "Vatsal Sharan" ], "affiliation": "", "summary": "Calibration is a fundamental concept that aims at ensuring the reliability of probabilistic predictions by aligning them with real-world outcomes", "links": { "openreview": "https://openreview.net/forum?id=Ib4ZXPXpss", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-208", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Wide-Horizon Thinking and Simulation-Based Evaluation for Real-World LLM Planning with Multifaceted Constraints", "authors": [ "Dongjie Yang", "Chengqiang Lu", "Qimeng Wang", "Xinbei Ma", "Yan Gao", "Yao Hu", "hai zhao" ], "affiliation": "", "summary": "Unlike reasoning, which often entails a deep sequence of deductive steps, complex real-world planning is characterized by the need to synthesize a broad spectrum of parallel and potentially conflictin", "links": { "openreview": "https://openreview.net/forum?id=b50IW9yV2M", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-209", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Does Object Binding Naturally Emerge in Large Pretrained Vision Transformers?", "authors": [ "Yihao Li", "Saeed Salehi", "Lyle Ungar", "Konrad Kording" ], "affiliation": "", "summary": "Object binding, the brain’s ability to bind the many features that collectively represent an object into a coherent whole, is central to human cognition", "links": { "openreview": "https://openreview.net/forum?id=5BS6gBb4yP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-210", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Probing Neural Combinatorial Optimization Models", "authors": [ "Zhiqin Zhang", "Yining Ma", "Zhiguang Cao", "Hoong Chuin Lau" ], "affiliation": "", "summary": "Neural combinatorial optimization (NCO) has achieved remarkable performance, yet its learned model representations and decision rationale remain a black box", "links": { "openreview": "https://openreview.net/forum?id=ycnc9aLnQu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-211", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Principled Path to Fitted Distributional Evaluation", "authors": [ "Sungee Hong", "Jiayi Wang", "Zhengling Qi", "Raymond K. W. Wong" ], "affiliation": "", "summary": "In reinforcement learning, distributional off-policy evaluation (OPE) focuses on estimating the return distribution of a target policy using offline data collected under a different policy", "links": { "openreview": "https://openreview.net/forum?id=Gte3F0ONhr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-212", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Improving LLM General Preference Alignment via Optimistic Online Mirror Descent", "authors": [ "Yuheng Zhang", "Dian Yu", "Tao Ge", "Linfeng Song", "Zhichen Zeng", "Haitao Mi", "Nan Jiang", "Dong Yu" ], "affiliation": "", "summary": "Reinforcement learning from human feedback (RLHF) has demonstrated remarkable effectiveness in aligning large language models (LLMs) with human preferences", "links": { "openreview": "https://openreview.net/forum?id=kZstGANG8D", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-213", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Progressive Inference-Time Annealing of Diffusion Models for Sampling from Boltzmann Densities", "authors": [ "Tara Akhound-Sadegh", "Jungyoon Lee", "Joey Bose", "Valentin De Bortoli", "Arnaud Doucet", "Michael M. Bronstein", "Dominique Beaini", "Siamak Ravanbakhsh", "Kirill Neklyudov", "Alexander Tong" ], "affiliation": "", "summary": "Sampling efficiently from a target unnormalized probability density remains a core challenge, with relevance across countless high-impact scientific applications", "links": { "openreview": "https://openreview.net/forum?id=vf2GHcxzMV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-214", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Compositional Monte Carlo Tree Diffusion for Extendable Planning", "authors": [ "Jaesik Yoon", "Hyeonseo Cho", "Sungjin Ahn" ], "affiliation": "", "summary": "Monte Carlo Tree Diffusion (MCTD) integrates diffusion models with structured tree search to enable effective trajectory exploration through stepwise reasoning", "links": { "openreview": "https://openreview.net/forum?id=om2CpclG4y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-215", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fast Monte Carlo Tree Diffusion: 100× Speedup via Parallel and Sparse Planning", "authors": [ "Jaesik Yoon", "Hyeonseo Cho", "Yoshua Bengio", "Sungjin Ahn" ], "affiliation": "", "summary": "Diffusion models have recently emerged as a powerful approach for trajectory planning", "links": { "openreview": "https://openreview.net/forum?id=JRVZTACwb0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-216", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Zero-shot Denoising via Neural Compression: Theoretical and algorithmic framework", "authors": [ "Ali Zafari", "Xi Chen", "Shirin Jalali" ], "affiliation": "", "summary": "Zero-shot denoising aims to denoise observations without access to training samples or clean reference images", "links": { "openreview": "https://openreview.net/forum?id=DwZD97uHgm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-217", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Gradient-Variation Online Adaptivity for Accelerated Optimization with Hölder Smoothness", "authors": [ "Yuheng Zhao", "Yu-Hu Yan", "Kfir Yehuda Levy", "Peng Zhao" ], "affiliation": "", "summary": "Smoothness is known to be crucial for acceleration in offline optimization, and for gradient-variation regret minimization in online learning", "links": { "openreview": "https://openreview.net/forum?id=miYvfEKvEl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-218", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to Search", "authors": [ "Arnav Kumar Jain", "Vibhakar Mohta", "Subin Kim", "Atiksh Bhardwaj", "Juntao Ren", "Yunhai Feng", "Sanjiban Choudhury", "Gokul Swamy" ], "affiliation": "", "summary": "The fundamental limitation of the behavioral cloning (BC) approach to imitation learning is that it only teaches an agent what the expert did at states the expert visited", "links": { "openreview": "https://openreview.net/forum?id=qN5hmLkBtC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-219", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MAESTRO : Adaptive Sparse Attention and Robust Learning for Multimodal Dynamic Time Series", "authors": [ "Payal Mohapatra", "Yueyuan Sui", "Akash Pandey", "Stephen Xia", "Qi Zhu" ], "affiliation": "", "summary": "From clinical healthcare to daily living, continuous sensor monitoring across multiple modalities has shown great promise for real-world intelligent decision-making but also faces various challenges", "links": { "openreview": "https://openreview.net/forum?id=1K28gV5MeF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-220", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Ctrl-DNA: Controllable Cell-Type-Specific Regulatory DNA Design via Constrained RL", "authors": [ "Xingyu Chen", "Shihao Ma", "Runsheng Lin", "Jiecong Lin", "BO WANG" ], "affiliation": "", "summary": "Designing regulatory DNA sequences that achieve precise cell-type-specific gene expression is crucial for advancements in synthetic biology, gene therapy and precision medicine", "links": { "openreview": "https://openreview.net/forum?id=JeXkIy0JyM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-221", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Universal Near Optimality of Hedge in Combinatorial Settings", "authors": [ "Zhiyuan Fan", "Arnab Maiti", "Lillian J. Ratliff", "Kevin Jamieson", "Gabriele Farina" ], "affiliation": "", "summary": "In this paper, we study the classical Hedge algorithm in combinatorial settings", "links": { "openreview": "https://openreview.net/forum?id=V7m2oQ5OFW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-222", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transfer Faster, Price Smarter: Minimax Dynamic Pricing under Cross-Market Preference Shift", "authors": [ "Yi Zhang", "Elynn Chen", "Yujun Yan" ], "affiliation": "", "summary": "We study contextual dynamic pricing when a target market can leverage $K$ auxiliary markets—offline logs or concurrent streams—whose *mean utilities differ by a structured preference shift*", "links": { "openreview": "https://openreview.net/forum?id=g2f0UoasGs", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-223", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transformer brain encoders explain human high-level visual responses", "authors": [ "Hossein Adeli", "Minni Sun", "Nikolaus Kriegeskorte" ], "affiliation": "", "summary": "A major goal of neuroscience is to understand brain computations during visual processing in naturalistic settings", "links": { "openreview": "https://openreview.net/forum?id=Tt3XLyuDrE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-224", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuning", "authors": [ "Yeongbin Seo", "Dongha Lee", "Jaehyung Kim", "Jinyoung Yeo" ], "affiliation": "", "summary": "Autoregressive (AR) language models generate text one token at a time, which limits their inference speed", "links": { "openreview": "https://openreview.net/forum?id=HvIRFV0J90", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-225", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MoBA: Mixture of Block Attention for Long-Context LLMs", "authors": [ "Enzhe Lu", "Zhejun Jiang", "Jingyuan Liu", "Yulun Du", "Tao Jiang", "Chao Hong", "Shaowei Liu", "Weiran He", "Enming Yuan", "Yuzhi Wang", "Zhiqi Huang", "Huan Yuan", "Suting Xu", "Xinran Xu", "Guokun Lai", "Yanru Chen", "Huabin Zheng", "Junjie Yan", "Jianlin Su", "Yuxin Wu", "Yutao Zhang", "Zhilin Yang", "Xinyu Zhou", "Mingxing Zhang", "Jiezhong Qiu" ], "affiliation": "", "summary": "Scaling the effective context length is essential for advancing large language models (LLMs) toward artificial general intelligence (AGI)", "links": { "openreview": "https://openreview.net/forum?id=RlqYCpTu1P", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-226", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DERD-Net: Learning Depth from Event-based Ray Densities", "authors": [ "Diego de Oliveira Hitzges", "Suman Ghosh", "Guillermo Gallego" ], "affiliation": "", "summary": "Event cameras offer a promising avenue for multi-view stereo depth estimation and Simultaneous Localization And Mapping (SLAM) due to their ability to detect blur-free 3D edges at high-speed and over", "links": { "openreview": "https://openreview.net/forum?id=0KnZasL9nA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-227", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Beyond Expectations: Quantile-Guided Alignment for Risk-Calibrated Language Models", "authors": [ "Xinran Wang", "Jin Du", "Azal Ahmad Khan", "Qi Le", "Enmao Diao", "Jiawei Zhou", "Jie Ding", "Ali Anwar" ], "affiliation": "", "summary": "Large language models can generate rare but catastrophic outputs, such as harmful conversations or insecure code", "links": { "openreview": "https://openreview.net/forum?id=R7HJj1YvJH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-228", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "VPO: Reasoning Preferences Optimization Based on $\\mathcal{V}$-Usable Information", "authors": [ "Zecheng Wang", "Chunshan Li", "Yupeng Zhang", "Han Liu", "Bingning Wang", "Dianhui Chu", "Dianbo Sui" ], "affiliation": "", "summary": "Direct Preference Optimization (DPO) is a widely used preference optimization algorithm in large language model (LLM) alignment, which reparameterizes the reward function in reinforcement learning wit", "links": { "openreview": "https://openreview.net/forum?id=LCZmI3iM8X", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-229", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MoCha: Towards Movie-Grade Talking Character Generation", "authors": [ "Cong Wei", "Bo Sun", "Haoyu Ma", "Ji Hou", "Felix Juefei-Xu", "Zecheng He", "Xiaoliang Dai", "Luxin Zhang", "Kunpeng Li", "Tingbo Hou", "Animesh Sinha", "Peter Vajda", "Wenhu Chen" ], "affiliation": "", "summary": "Recent advancements in video generation have achieved impressive motion realism, yet they often overlook character-driven storytelling, a crucial task for automated film, animation generation", "links": { "openreview": "https://openreview.net/forum?id=S9E1nfYPwl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-230", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Stable Minima of ReLU Neural Networks Suffer from the Curse of Dimensionality: The Neural Shattering Phenomenon", "authors": [ "Tongtong Liang", "Dan Qiao", "Yu-Xiang Wang", "Rahul Parhi" ], "affiliation": "", "summary": "We study the implicit bias of flatness / low (loss) curvature and its effects on generalization in two-layer overparameterized ReLU networks with multivariate inputs---a problem well motivated by the", "links": { "openreview": "https://openreview.net/forum?id=HhCl2BIHfk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-231", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fisher meets Feynman: score-based variational inference with a product of experts", "authors": [ "Diana Cai", "Robert M. Gower", "David Blei", "Lawrence K. Saul" ], "affiliation": "", "summary": "We introduce a highly expressive yet distinctly tractable family for black-box variational inference (BBVI)", "links": { "openreview": "https://openreview.net/forum?id=yG8vmj3EAU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-232", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy", "authors": [ "Xiang Li", "Chendi Wang", "Buxin Su", "Qi Long", "Weijie J Su" ], "affiliation": "", "summary": "Differentially private (DP) decentralized Federated Learning (FL) allows local users to collaborate without sharing their data with a central server", "links": { "openreview": "https://openreview.net/forum?id=YIGUv0BZCy", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-233", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mitigating Instability in High Residual Adaptive Sampling for PINNs via Langevin Dynamics", "authors": [ "Minseok Jeong", "Giup Seo", "Euiseok Hwang" ], "affiliation": "", "summary": "Recently, physics-informed neural networks (PINNs) have gained attention in the scientific community for their potential to solve partial differential equations (PDEs)", "links": { "openreview": "https://openreview.net/forum?id=hq2CkcEY7h", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-234", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Spectral Estimation with Free Decompression", "authors": [ "Siavash Ameli", "Chris van der Heide", "Liam Hodgkinson", "Michael W. Mahoney" ], "affiliation": "", "summary": "Computing eigenvalues of very large matrices is a critical task in many machine learning applications, including the evaluation of log-determinants, the trace of matrix functions, and other important", "links": { "openreview": "https://openreview.net/forum?id=2CeGVUpOd7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-235", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching", "authors": [ "Chen Chen", "Pengsheng Guo", "Liangchen Song", "Jiasen Lu", "Rui Qian", "Tsu-Jui Fu", "Xinze Wang", "Wei Liu", "Yinfei Yang", "Alex Schwing" ], "affiliation": "", "summary": "Conditional generative modeling aims to learn a conditional data distribution from samples containing data-condition pairs", "links": { "openreview": "https://openreview.net/forum?id=idnW3BiZcV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-236", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Training-Free Constrained Generation With Stable Diffusion Models", "authors": [ "Stefano Zampini", "Jacob K Christopher", "Luca Oneto", "Davide Anguita", "Ferdinando Fioretto" ], "affiliation": "", "summary": "Stable diffusion models represent the state-of-the-art in data synthesis across diverse domains and hold transformative potential for applications in science and engineering, e", "links": { "openreview": "https://openreview.net/forum?id=TrNB08KuHK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-237", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Comparator-Adaptive $\\Phi$-Regret: Improved Bounds, Simpler Algorithms, and Applications to Games", "authors": [ "Soumita Hait", "Ping Li", "Haipeng Luo", "Mengxiao Zhang" ], "affiliation": "", "summary": "In the classic expert problem, $\\Phi$-regret measures the gap between the learner's total loss and that achieved by applying the best action transformation $\\phi \\in \\Phi$", "links": { "openreview": "https://openreview.net/forum?id=rSsc9uCVBl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-238", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "X-Field: A Physically Informed Representation for 3D X-ray Reconstruction", "authors": [ "Feiran Wang", "Jiachen Tao", "Junyi Wu", "Haoxuan Wang", "Bin Duan", "Kai Wang", "Zongxin Yang", "Yan Yan" ], "affiliation": "", "summary": "X-ray imaging is indispensable in medical diagnostics, yet its use is tightly regulated due to radiation exposure", "links": { "openreview": "https://openreview.net/forum?id=S8XcHutp7Z", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-239", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Distillation Robustifies Unlearning", "authors": [ "Bruce W. Lee", "Addie Foote", "Alex Infanger", "Leni Shor", "Harish K Kamath", "Jacob Goldman-Wetzler", "Bryce Woodworth", "Alex Cloud", "Alexander Matt Turner" ], "affiliation": "", "summary": "Current LLM unlearning methods are not robust", "links": { "openreview": "https://openreview.net/forum?id=UTGjik64IK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-240", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SimWorld: An Open-ended Simulator for Agents in Physical and Social Worlds", "authors": [ "Xiaokang Ye", "Jiawei Ren", "Yan Zhuang", "Xuhong He", "Yiming Liang", "Yiqing Yang", "Mrinaal Dogra", "Xianrui Zhong", "Eric Liu", "Kevin Benavente", "Rajiv Mandya Nagaraju", "Dhruv Vivek Sharma", "Ziqiao Ma", "Tianmin Shu", "Zhiting Hu", "Lianhui Qin" ], "affiliation": "", "summary": "While LLM/VLM-powered AI agents have advanced rapidly in math, coding, and computer use, their applications in complex physical and social environments remain challenging", "links": { "openreview": "https://openreview.net/forum?id=FxCy8TvQHO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-241", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Two‑Stage Learning of Stabilizing Neural Controllers via Zubov Sampling and Iterative Domain Expansion", "authors": [ "Haoyu Li", "Xiangru Zhong", "Bin Hu", "Huan Zhang" ], "affiliation": "", "summary": "Learning-based neural network (NN) control policies have shown impressive empirical performance", "links": { "openreview": "https://openreview.net/forum?id=IT12Radlnq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-242", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Complexity of Symmetric Equilibria in Min-Max Optimization and Team Zero-Sum Games", "authors": [ "Ioannis Anagnostides", "Ioannis Panageas", "Tuomas Sandholm", "Jingming Yan" ], "affiliation": "", "summary": "We consider the problem of computing stationary points in min-max optimization, with a focus on the special case of Nash equilibria in (two-)team zero-sum games", "links": { "openreview": "https://openreview.net/forum?id=75LMvs1CjG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-243", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On Traceability in $\\ell_p$ Stochastic Convex Optimization", "authors": [ "Sasha Voitovych", "Mahdi Haghifam", "Idan Attias", "Gintare Karolina Dziugaite", "Roi Livni", "Daniel M. Roy" ], "affiliation": "", "summary": "In this paper, we investigate the necessity of traceability for accurate learning in stochastic convex optimization (SCO) under $\\ell_p$ geometries", "links": { "openreview": "https://openreview.net/forum?id=LyG7kDSsGh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-244", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Improved Representation Steering for Language Models", "authors": [ "Zhengxuan Wu", "Qinan Yu", "Aryaman Arora", "Christopher D Manning", "Christopher Potts" ], "affiliation": "", "summary": "Steering methods for language models (LMs) seek to provide fine-grained and interpretable control over model generations by variously changing model inputs, weights, or representations to adjust behav", "links": { "openreview": "https://openreview.net/forum?id=VHb883Gs1u", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-245", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FlexOLMo: Open Language Models for Flexible Data Use", "authors": [ "Weijia Shi", "Akshita Bhagia", "Kevin Farhat", "Niklas Muennighoff", "Jacob Morrison", "Evan Pete Walsh", "Dustin Schwenk", "Shayne Longpre", "Jake Poznanski", "Allyson Ettinger", "Daogao Liu", "Margaret Li", "Mike Lewis", "Wen-tau Yih", "Dirk Groeneveld", "Luca Soldaini", "Kyle Lo", "Noah A. Smith", "Luke Zettlemoyer", "Pang Wei Koh", "Hannaneh Hajishirzi", "Ali Farhadi", "Sewon Min" ], "affiliation": "", "summary": "We introduce FlexOLMo, a new class of language models (LMs) that supports (1) distributed training without data sharing, where different model parameters are independently trained on private datasets,", "links": { "openreview": "https://openreview.net/forum?id=1rUj9ZN6Bz", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-246", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Fragile Truth of Saliency: Improving LLM Input Attribution via Attention Bias Optimization", "authors": [ "Yihua Zhang", "Changsheng Wang", "Yiwei Chen", "Chongyu Fan", "Jinghan Jia", "Sijia Liu" ], "affiliation": "", "summary": "Input saliency aims to quantify the influence of input tokens on the output of large language models (LLMs), which has been widely used for prompt engineering, model interpretability, and behavior att", "links": { "openreview": "https://openreview.net/forum?id=DrUR87D4Hj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-247", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Enhancing Training Data Attribution with Representational Optimization", "authors": [ "Weiwei Sun", "Haokun Liu", "Nikhil Kandpal", "Colin Raffel", "Yiming Yang" ], "affiliation": "", "summary": "Training data attribution (TDA) methods aim to measure how training data impacts a model's predictions", "links": { "openreview": "https://openreview.net/forum?id=ESB924uT5Y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-248", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DisMo: Disentangled Motion Representations for Open-World Motion Transfer", "authors": [ "Thomas Ressler-Antal", "Frank Fundel", "Malek Ben Alaya", "Stefan Andreas Baumann", "Felix Krause", "Ming Gui", "Björn Ommer" ], "affiliation": "", "summary": "Recent advances in text-to-video (T2V) and image-to-video (I2V) models, have enabled the creation of visually compelling and dynamic videos from simple textual descriptions or initial frames", "links": { "openreview": "https://openreview.net/forum?id=jneVld5iZw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-249", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Rig3R: Rig-Aware Conditioning and Discovery for 3D Reconstruction", "authors": [ "Samuel Li", "Pujith Kachana", "Prajwal Chidananda", "Saurabh Nair", "Yasutaka Furukawa", "Matthew Brown" ], "affiliation": "", "summary": "Estimating agent pose and 3D scene structure from multi-camera rigs is a central task in embodied AI applications such as autonomous driving", "links": { "openreview": "https://openreview.net/forum?id=vEFPm6gw2s", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-250", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Cycle-Sync: Robust Global Camera Pose Estimation through Enhanced Cycle-Consistent Synchronization", "authors": [ "Shaohan Li", "Yunpeng Shi", "Gilad Lerman" ], "affiliation": "", "summary": "We introduce Cycle-Sync, a robust and global framework for estimating camera poses (both rotations and locations)", "links": { "openreview": "https://openreview.net/forum?id=M3zxsDL2Rk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-251", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Dense Associative Memory with Epanechnikov Energy", "authors": [ "Benjamin Hoover", "Zhaoyang Shi", "Krishna Balasubramanian", "Dmitry Krotov", "Parikshit Ram" ], "affiliation": "", "summary": "We propose a novel energy function for Dense Associative Memory (DenseAM) networks, the log-sum-ReLU (LSR), inspired by optimal kernel density estimation", "links": { "openreview": "https://openreview.net/forum?id=ZbQ5Zq3zA3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-252", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Searching Latent Program Spaces", "authors": [ "Matthew Macfarlane", "Clément Bonnet" ], "affiliation": "", "summary": "General intelligence requires systems that acquire new skills efficiently and generalize beyond their training distributions", "links": { "openreview": "https://openreview.net/forum?id=CsXKGIqZtr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-253", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Convergence Rates of Constrained Expected Improvement", "authors": [ "Haowei Wang", "Jingyi Wang", "Zhongxiang Dai", "Nai-Yuan Chiang", "Szu Hui Ng", "Cosmin G. Petra" ], "affiliation": "", "summary": "Constrained Bayesian optimization (CBO) methods have seen significant success in black-box optimization with constraints", "links": { "openreview": "https://openreview.net/forum?id=Dn4He1IrUT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-254", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CoT Information: Improved Sample Complexity under Chain-of-Thought Supervision", "authors": [ "Awni Altabaa", "Omar Montasser", "John Lafferty" ], "affiliation": "", "summary": "Learning complex functions that involve multi-step reasoning poses a significant challenge for standard supervised learning from input-output examples", "links": { "openreview": "https://openreview.net/forum?id=OkVQJZWGfn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-255", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Unlocking hidden biomolecular conformational landscapes in diffusion models at inference time", "authors": [ "Daniel D. Richman", "Jessica Karaguesian", "Carl-Mikael Suomivuori", "Ron O. Dror" ], "affiliation": "", "summary": "The function of biomolecules such as proteins depends on their ability to interconvert between a wide range of structures or conformations", "links": { "openreview": "https://openreview.net/forum?id=U87XyMPrZp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-256", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "QSVD: Efficient Low-rank Approximation for Unified Query-Key-Value Weight Compression in Low-Precision Vision-Language Models", "authors": [ "Yutong Wang", "Haiyu Wang", "Sai Qian Zhang" ], "affiliation": "", "summary": "Vision-Language Models (VLMs) are integral to tasks such as image captioning and visual question answering, but their high computational cost, driven by large memory footprints and processing time, li", "links": { "openreview": "https://openreview.net/forum?id=sEFDhxF1mG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-257", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Caption This, Reason That: VLMs Caught in the Middle", "authors": [ "Zihan Weng", "Lucas Gomez", "Taylor Whittington Webb", "Pouya Bashivan" ], "affiliation": "", "summary": "Vision-Language Models (VLMs) have shown remarkable progress in visual understanding in recent years", "links": { "openreview": "https://openreview.net/forum?id=m6WmeOI1AW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-258", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Universal Causal Inference in a Topos", "authors": [ "Sridhar Mahadevan" ], "affiliation": "", "summary": "In this paper, we explore the universal properties underlying causal inference by formulating it in terms of a topos", "links": { "openreview": "https://openreview.net/forum?id=TOhpnECT10", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-259", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Can We Infer Confidential Properties of Training Data from LLMs?", "authors": [ "Pengrun Huang", "Chhavi Yadav", "Kamalika Chaudhuri", "Ruihan Wu" ], "affiliation": "", "summary": "Large language models (LLMs) are increasingly fine-tuned on domain-specific datasets to support applications in fields such as healthcare, finance, and law", "links": { "openreview": "https://openreview.net/forum?id=PhIWEbewAz", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-260", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Temperature is All You Need for Generalization in Langevin Dynamics and other Markov Processes", "authors": [ "Itamar Harel", "Yonathan Wolanowsky", "Gal Vardi", "Nathan Srebro", "Daniel Soudry" ], "affiliation": "", "summary": "We analyze the generalization gap (gap between the training and test errors) when training a potentially over-parametrized model using a Markovian stochastic training algorithm, initialized from some", "links": { "openreview": "https://openreview.net/forum?id=EjkvtZwRoA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-261", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learning the Wrong Lessons: Syntactic-Domain Spurious Correlations in Language Models", "authors": [ "Chantal Shaib", "Vinith Menon Suriyakumar", "Byron C Wallace", "Marzyeh Ghassemi" ], "affiliation": "", "summary": "For an LLM to correctly respond to an instruction it must understand both the semantics and the domain (i", "links": { "openreview": "https://openreview.net/forum?id=oBikm5Rshc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-262", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Corporate Needs You to Find the Difference: Revisiting Submodular and Supermodular Ratio Optimization Problems", "authors": [ "Elfarouk Harb", "Yousef Yassin", "Chandra Chekuri" ], "affiliation": "", "summary": "We consider the following question: given a submodular or supermodular set function $f:2^V \\to \\mathbb{R}$, how should one minimize or maximize its average value $f(S)/\\ | S\\ | $ over non-empty subsets $S\\s", "links": { "openreview": "https://openreview.net/forum?id=Sf5nxMRiG7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-263", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Private Set Union with Multiple Contributions", "authors": [ "Travis Dick", "Haim Kaplan", "Alex Kulesza", "Uri Stemmer", "Ziteng Sun", "Ananda Theertha Suresh" ], "affiliation": "", "summary": "In the private set union problem each user owns a bag of at most $k$ items (from some large universe of items), and we are interested in computing the union of the items in the bags of all of the user", "links": { "openreview": "https://openreview.net/forum?id=RdNYp8ilPr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-264", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Clustering via Hedonic Games: New Concepts and Algorithms", "authors": [ "Gergely Csáji", "Alexander Gundert", "Jörg Rothe", "Ildikó Schlotter" ], "affiliation": "", "summary": "We study fundamental connections between coalition formation games and clustering, illustrating the cross-disciplinary relevance of these concepts", "links": { "openreview": "https://openreview.net/forum?id=96I0XnrjkQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-265", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Graphon Limit Hypothesis: Understanding Neural Network Pruning via Infinite Width Analysis", "authors": [ "Hoang Pham", "The-Anh Ta", "Tom Jacobs", "Rebekka Burkholz", "Long Tran-Thanh" ], "affiliation": "", "summary": "Sparse neural networks promise efficiency, yet training them effectively remains a fundamental challenge", "links": { "openreview": "https://openreview.net/forum?id=EEZLBhyer1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-266", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fast Training of Large Kernel Models with Delayed Projections", "authors": [ "Amirhesam Abedsoltan", "Siyuan Ma", "Parthe Pandit", "Mikhail Belkin" ], "affiliation": "", "summary": "Classical kernel machines have historically faced significant challenges in scaling to large datasets and model sizes—a key ingredient that has driven the success of neural networks", "links": { "openreview": "https://openreview.net/forum?id=a7hHwWnZey", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-267", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning", "authors": [ "Siyan Zhao", "Devaansh Gupta", "Qinqing Zheng", "Aditya Grover" ], "affiliation": "", "summary": "Recent large language models (LLMs) have demonstrated strong reasoning capabilities that benefits from online reinforcement learning (RL)", "links": { "openreview": "https://openreview.net/forum?id=7ZVRlBFuEv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-268", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Affine-Invariant Global Non-Asymptotic Convergence Analysis of BFGS under Self-Concordance", "authors": [ "Qiujiang Jin", "Aryan Mokhtari" ], "affiliation": "", "summary": "In this paper, we establish global non-asymptotic convergence guarantees for the BFGS quasi-Newton method without requiring strong convexity or the Lipschitz continuity of the gradient or Hessian", "links": { "openreview": "https://openreview.net/forum?id=d6UV0UNgn9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-269", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Extrapolation by Association: Length Generalization Transfer In Transformers", "authors": [ "Ziyang Cai", "Nayoung Lee", "Avi Schwarzschild", "Samet Oymak", "Dimitris Papailiopoulos" ], "affiliation": "", "summary": "Transformer language models have demonstrated impressive generalization capabilities in natural language domains, yet we lack a fine-grained understanding of how such generalization arises", "links": { "openreview": "https://openreview.net/forum?id=aLUAzLDIOc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-270", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Optimal Neural Compressors for the Rate-Distortion-Perception Tradeoff", "authors": [ "Eric Lei", "Hamed Hassani", "Shirin Saeedi Bidokhti" ], "affiliation": "", "summary": "Recent efforts in neural compression have focused on the rate-distortion-perception (RDP) tradeoff, where the perception constraint ensures the source and reconstruction distributions are close in ter", "links": { "openreview": "https://openreview.net/forum?id=ZAKpELpclI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-271", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Bridging Symmetry and Robustness: On the Role of Equivariance in Enhancing Adversarial Robustness", "authors": [ "Longwei Wang", "Ifrat Ikhtear Uddin", "KC Santosh", "Chaowei Zhang", "Xiao Qin", "Yang Zhou" ], "affiliation": "", "summary": "Adversarial examples reveal critical vulnerabilities in deep neural networks by exploiting their sensitivity to imperceptible input perturbations", "links": { "openreview": "https://openreview.net/forum?id=xDxskDUvte", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-272", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Color Conditional Generation with Sliced Wasserstein Guidance", "authors": [ "Alexander Lobashev", "Maria Larchenko", "Dmitry Guskov" ], "affiliation": "", "summary": "We propose SW-Guidance, a training-free approach for image generation conditioned on the color distribution of a reference image", "links": { "openreview": "https://openreview.net/forum?id=r1Bx58M6It", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-273", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Private Hyperparameter Tuning with Ex-Post Guarantee", "authors": [ "Badih Ghazi", "Pritish Kamath", "Alexander Knop", "Ravi Kumar", "Pasin Manurangsi", "Chiyuan Zhang" ], "affiliation": "", "summary": "The conventional approach in differential privacy (DP) literature formulates the privacy-utility tradeoff with a \"privacy-first\" perspective: for a predetermined level of privacy, a certain utility i", "links": { "openreview": "https://openreview.net/forum?id=zjMd3yfyWv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-274", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Implies B: Circuit Analysis in LLMs for Propositional Logical Reasoning", "authors": [ "Guan Zhe Hong", "Nishanth Dikkala", "Enming Luo", "Cyrus Rashtchian", "Xin Wang", "Rina Panigrahy" ], "affiliation": "", "summary": "Due to the size and complexity of modern large language models (LLMs), it has proven challenging to uncover the underlying mechanisms that models use to solve reasoning problems", "links": { "openreview": "https://openreview.net/forum?id=M0U8wUow8c", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-275", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Optimal and Provable Calibration in High-Dimensional Binary Classification: Angular Calibration and Platt Scaling", "authors": [ "Yufan Li", "Pragya Sur" ], "affiliation": "", "summary": "We study the fundamental problem of calibrating a linear binary classifier of the form \\(\\sigma(\\hat{w}^\\top x)\\), where the feature vector \\(x\\) is Gaussian, \\(\\sigma\\) is a link function, and \\(\\hat", "links": { "openreview": "https://openreview.net/forum?id=SgQAleMecy", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-276", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Reconstruction and Secrecy under Approximate Distance Queries", "authors": [ "Shay Moran", "Elizaveta Nesterova" ], "affiliation": "", "summary": "Consider the task of locating an unknown target point using approximate distance queries: in each round, a reconstructor selects a reference point and receives a noisy version of its distance to the t", "links": { "openreview": "https://openreview.net/forum?id=0vJJdEiXOb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-277", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Win Fast or Lose Slow: Balancing Speed and Accuracy in Latency-Sensitive Decisions of LLMs", "authors": [ "Hao Kang", "Qingru Zhang", "Han Cai", "Weiyuan Xu", "Tushar Krishna", "Yilun Du", "Tsachy Weissman" ], "affiliation": "", "summary": "Large language models (LLMs) have shown remarkable performance across diverse reasoning and generation tasks, and are increasingly deployed as agents in dynamic environments such as code generation an", "links": { "openreview": "https://openreview.net/forum?id=Fcs90Rwm8j", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-278", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Characterizing the Expressivity of Fixed-Precision Transformer Language Models", "authors": [ "Jiaoda Li", "Ryan Cotterell" ], "affiliation": "", "summary": "Transformer-based language models (LMs) have achieved widespread empirical success, but their theoretical expressive power remains only partially understood", "links": { "openreview": "https://openreview.net/forum?id=29LwAgLFpj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-279", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Distributional Training Data Attribution: What do Influence Functions Sample?", "authors": [ "Bruno Kacper Mlodozeniec", "Isaac Reid", "Samuel Power", "David Krueger", "Murat A Erdogdu", "Richard E. Turner", "Roger Baker Grosse" ], "affiliation": "", "summary": "Randomness is an unavoidable part of training deep learning models, yet something that traditional training data attribution algorithms fail to rigorously account for", "links": { "openreview": "https://openreview.net/forum?id=UBRFn7YKMe", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-280", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Hogwild! Inference: Parallel LLM Generation via Concurrent Attention", "authors": [ "Gleb Rodionov", "Roman Garipov", "Alina Shutova", "George Yakushev", "Erik Schultheis", "Vage Egiazarian", "Anton Sinitsin", "Denis Kuznedelev", "Dan Alistarh" ], "affiliation": "", "summary": "Large Language Models (LLMs) have demonstrated the ability to tackle increasingly complex tasks through advanced reasoning, long-form content generation, and tool use", "links": { "openreview": "https://openreview.net/forum?id=roKj4IwaVT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-281", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mind-the-Glitch: Visual Correspondence for Detecting Inconsistencies in Subject-Driven Generation", "authors": [ "Abdelrahman Eldesokey", "Aleksandar Cvejić", "Bernard Ghanem", "Peter Wonka" ], "affiliation": "", "summary": "We propose a novel approach for disentangling visual and semantic features from the backbones of pre-trained diffusion models, enabling visual correspondence in a manner analogous to the well-establis", "links": { "openreview": "https://openreview.net/forum?id=4FyNdd2b5S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-282", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability?", "authors": [ "Denis Sutter", "Julian Minder", "Thomas Hofmann", "Tiago Pimentel" ], "affiliation": "", "summary": "The concept of causal abstraction got recently popularised to demystify the opaque decision-making processes of machine learning models; in short, a neural network can be abstracted as a higher-level", "links": { "openreview": "https://openreview.net/forum?id=ZYXTLo7kCi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-283", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Gaussian Herding across Pens: An Optimal Transport Perspective on Global Gaussian Reduction for 3DGS", "authors": [ "Tao Wang", "Mengyu Li", "Geduo Zeng", "Cheng Meng", "Qiong Zhang" ], "affiliation": "", "summary": "3D Gaussian Splatting (3DGS) has emerged as a powerful technique for radiance field rendering, but it typically requires millions of redundant Gaussian primitives, overwhelming memory and rendering bu", "links": { "openreview": "https://openreview.net/forum?id=j1QkrVjNVF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-284", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Generalized Top-k Mallows Model for Ranked Choices", "authors": [ "Shahrzad Haddadan", "Sara Ahmadian" ], "affiliation": "", "summary": "The classic Mallows model is a foundational tool for modeling user preferences", "links": { "openreview": "https://openreview.net/forum?id=jVwIfsJLvh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-285", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Improving Bilinear RNN with Closed-loop Control", "authors": [ "Jiaxi Hu", "Yongqi Pan", "Jusen Du", "Disen Lan", "Xiaqiang Tang", "Qingsong Wen", "Yuxuan Liang", "Weigao Sun" ], "affiliation": "", "summary": "Recent efficient sequence modeling methods, such as Gated DeltaNet, TTT, and RWKV-7, have achieved performance improvements by supervising the recurrent memory management through the Delta learning ru", "links": { "openreview": "https://openreview.net/forum?id=jlJaRXDzCE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-286", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence", "authors": [ "Diankun Wu", "Fangfu Liu", "Yi-Hsin Hung", "Yueqi Duan" ], "affiliation": "", "summary": "Recent advancements in Multimodal Large Language Models (MLLMs) have significantly enhanced performance on 2D visual tasks", "links": { "openreview": "https://openreview.net/forum?id=RnXS7aK4rK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-287", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SANSA: Unleashing the Hidden Semantics in SAM2 for Few-Shot Segmentation", "authors": [ "Claudia Cuttano", "Gabriele Trivigno", "Giuseppe Averta", "Carlo Masone" ], "affiliation": "", "summary": "Few-shot segmentation aims to segment unseen categories from just a handful of annotated examples", "links": { "openreview": "https://openreview.net/forum?id=o8r3gOFTQo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-288", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Cost-Aware Contrastive Routing for LLMs", "authors": [ "Reza Shirkavand", "Shangqian Gao", "Peiran Yu", "Heng Huang" ], "affiliation": "", "summary": "We study cost-aware routing for large language models across diverse and dynamic pools of models", "links": { "openreview": "https://openreview.net/forum?id=4Qe2Hga43N", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-289", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Is Grokking a Computational Glass Relaxation?", "authors": [ "Xiaotian Zhang", "Yue Shang", "Entao Yang", "Ge Zhang" ], "affiliation": "", "summary": "Understanding neural network' (NN) generalizability remains a central question in deep learning research", "links": { "openreview": "https://openreview.net/forum?id=Tk5nQnTGmP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-290", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LoRATv2: Enabling Low-Cost Temporal Modeling in One-Stream Trackers", "authors": [ "Liting Lin", "Heng Fan", "Zhipeng Zhang", "Yuqing Huang", "Yaowei Wang", "Yong Xu", "Haibin Ling" ], "affiliation": "", "summary": "Transformer-based algorithms, such as LoRAT, have significantly enhanced object-tracking performance", "links": { "openreview": "https://openreview.net/forum?id=q06YjUj0FB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-291", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the necessity of adaptive regularisation: Optimal anytime online learning on $\\boldsymbol{\\ell_p}$-balls", "authors": [ "Emmeran Johnson", "David Martínez-Rubio", "Ciara Pike-Burke", "Patrick Rebeschini" ], "affiliation": "", "summary": "We study online convex optimization on $\\ell_p$-balls in $\\mathbb{R}^d$ for $p > 2$", "links": { "openreview": "https://openreview.net/forum?id=ONc9vWkwCp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-292", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Controlling Thinking Speed in Reasoning Models", "authors": [ "Zhengkai Lin", "Zhihang Fu", "Ze Chen", "Chao Chen", "Liang Xie", "Wenxiao Wang", "Deng Cai", "Zheng Wang", "Jieping Ye" ], "affiliation": "", "summary": "Human cognition is theorized to operate in two modes: fast, intuitive System 1 thinking and slow, deliberate System 2 thinking", "links": { "openreview": "https://openreview.net/forum?id=ESELaMThLN", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-293", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Vector Quantization in the Brain: Grid-like Codes in World Models", "authors": [ "Xiangyuan Peng", "Xingsi Dong", "Si Wu" ], "affiliation": "", "summary": "We propose Grid-like Code Quantization (GCQ), a brain-inspired method for compressing observation-action sequences into discrete representations using grid-like patterns in attractor dynamics", "links": { "openreview": "https://openreview.net/forum?id=M44RvNMZs4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-294", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "IA-GGAD: Zero-shot Generalist Graph Anomaly Detection via Invariant and Affinity Learning", "authors": [ "Xiong Zhang", "Zhenli He", "Changlong Fu", "Cheng Xie" ], "affiliation": "", "summary": "Generalist Graph Anomaly Detection (GGAD) extends traditional Graph Anomaly Detection (GAD) from one-for-one to one-for-all scenarios, posing significant challenges due to Feature Space Shift (FSS) an", "links": { "openreview": "https://openreview.net/forum?id=Cggdvyt8ik", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-295", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Revisiting Generative Infrared and Visible Image Fusion Based on Human Cognitive Laws", "authors": [ "Lin Guo", "Xiaoqing Luo", "Wei Xie", "Zhancheng Zhang", "Hui Li", "Rui Wang", "Zhenhua Feng", "Xiaoning Song" ], "affiliation": "", "summary": "Existing infrared and visible image fusion methods often face the dilemma of balancing modal information", "links": { "openreview": "https://openreview.net/forum?id=wvcYIEaD5X", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-296", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MonoLift: Learning 3D Manipulation Policies from Monocular RGB via Distillation", "authors": [ "Ziru Wang", "Mengmeng Wang", "Guang Dai", "Yongliu Long", "Jingdong Wang" ], "affiliation": "", "summary": "Although learning 3D manipulation policies from monocular RGB images is lightweight and deployment-friendly, the lack of structural information often leads to inaccurate action estimation", "links": { "openreview": "https://openreview.net/forum?id=wZzC5rpDY1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-297", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SoFar: Language-Grounded Orientation Bridges Spatial Reasoning and Object Manipulation", "authors": [ "Zekun Qi", "Wenyao Zhang", "Yufei Ding", "Runpei Dong", "XinQiang Yu", "Jingwen Li", "Lingyun Xu", "Baoyu Li", "Xialin He", "Guofan Fan", "Jiazhao Zhang", "Jiawei He", "Jiayuan Gu", "Xin Jin", "Kaisheng Ma", "Zhizheng Zhang", "He Wang", "Li Yi" ], "affiliation": "", "summary": "While spatial reasoning has made progress in object localization relationships, it often overlooks object orientation—a key factor in 6-DoF fine-grained manipulation", "links": { "openreview": "https://openreview.net/forum?id=kmv7yg6QXv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-298", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Conditional Representation Learning for Customized Tasks", "authors": [ "Honglin Liu", "Chao Sun", "Peng Hu", "Yunfan Li", "Xi Peng" ], "affiliation": "", "summary": "Conventional representation learning methods learn a universal representation that primarily captures dominant semantics, which may not always align with customized downstream tasks", "links": { "openreview": "https://openreview.net/forum?id=F1wDPNLvTb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-299", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Angular Steering: Behavior Control via Rotation in Activation Space", "authors": [ "Hieu M. Vu", "Tan Minh Nguyen" ], "affiliation": "", "summary": "Controlling specific behaviors in large language models while preserving their general capabilities is a central challenge for safe and reliable artificial intelligence deployment", "links": { "openreview": "https://openreview.net/forum?id=dGi2d5yDs4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-300", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "COOPERA: Continual Open-Ended Human-Robot Assistance", "authors": [ "Chenyang Ma", "Kai Lu", "Ruta Desai", "Xavier Puig", "Andrew Markham", "Niki Trigoni" ], "affiliation": "", "summary": "To understand and collaborate with humans, robots must account for individual human traits, habits, and activities over time", "links": { "openreview": "https://openreview.net/forum?id=wOSZVnYH5w", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-301", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Joint Hierarchical Representation Learning of Samples and Features via Informed Tree-Wasserstein Distance", "authors": [ "Ya-Wei Eileen Lin", "Ronald R. Coifman", "Gal Mishne", "Ronen Talmon" ], "affiliation": "", "summary": "High-dimensional data often exhibit hierarchical structures in both modes: samples and features", "links": { "openreview": "https://openreview.net/forum?id=VanvIu0KZU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-302", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Detecting Generated Images by Fitting Natural Image Distributions", "authors": [ "Yonggang Zhang", "Jun Nie", "Xinmei Tian", "Mingming Gong", "Kun Zhang", "Bo Han" ], "affiliation": "", "summary": "The increasing realism of generated images has raised significant concerns about their potential misuse, necessitating robust detection methods", "links": { "openreview": "https://openreview.net/forum?id=27xTIAFbc6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-303", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Agnostic Learning under Targeted Poisoning: Optimal Rates and the Role of Randomness", "authors": [ "Bogdan Chornomaz", "Yonatan Koren", "Shay Moran", "Tom Waknine" ], "affiliation": "", "summary": "We study the problem of learning in the presence of an adversary that can corrupt an $\\eta$ fraction of the training examples with the goal of causing failure on a specific test point", "links": { "openreview": "https://openreview.net/forum?id=K3xaVpSHkV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-304", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ORIGAMISPACE: Benchmarking Multimodal LLMs in Multi-Step Spatial Reasoning with Mathematical Constraints", "authors": [ "Rui Xu", "Dakuan Lu", "Zicheng Zhao", "Xiaoyu Tan", "Xintao Wang", "Siyu Yuan", "Jiangjie Chen", "Xu Yinghui" ], "affiliation": "", "summary": "Spatial reasoning is a key capability in the field of artificial intelligence, especially crucial in areas such as robotics, computer vision, and natural language understanding", "links": { "openreview": "https://openreview.net/forum?id=y7ahj9RoXQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-305", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Establishing Linear Surrogate Regret Bounds for Convex Smooth Losses via Convolutional Fenchel–Young Losses", "authors": [ "Yuzhou Cao", "Han Bao", "Lei Feng", "Bo An" ], "affiliation": "", "summary": "Surrogate regret bounds, also known as excess risk bounds, bridge the gap between the convergence rates of surrogate and target losses", "links": { "openreview": "https://openreview.net/forum?id=A4Xx9irvpp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-306", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Co-Reinforcement Learning for Unified Multimodal Understanding and Generation", "authors": [ "Jingjing Jiang", "Chongjie Si", "Jun Luo", "Hanwang Zhang", "Chao Ma" ], "affiliation": "", "summary": "This paper presents a pioneering exploration of reinforcement learning (RL) via group relative policy optimization for unified multimodal large language models (ULMs), aimed at simultaneously reinforc", "links": { "openreview": "https://openreview.net/forum?id=aDa0xEFDu1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-307", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Structured Linear CDEs: Maximally Expressive and Parallel-in-Time Sequence Models", "authors": [ "Benjamin Walker", "Lingyi Yang", "Nicola Muca Cirone", "Cristopher Salvi", "Terry Lyons" ], "affiliation": "", "summary": "This work introduces Structured Linear Controlled Differential Equations (SLiCEs), a unifying framework for sequence models with structured, input-dependent state-transition matrices that retain the m", "links": { "openreview": "https://openreview.net/forum?id=HKDyRDzy1E", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-308", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding", "authors": [ "Xiaoqian Shen", "Wenxuan Zhang", "Jun Chen", "Mohamed Elhoseiny" ], "affiliation": "", "summary": "Understanding and reasoning over long videos pose significant challenges for large video language models (LVLMs) due to the difficulty in processing intensive video tokens beyond context window and re", "links": { "openreview": "https://openreview.net/forum?id=5xPvWat3IX", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-309", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Privacy amplification by random allocation", "authors": [ "Moshe Shenfeld", "Vitaly Feldman" ], "affiliation": "", "summary": "We consider the privacy amplification properties of a sampling scheme in which a user's data is used in $k$ steps chosen randomly and uniformly from a sequence (or set) of $t$ steps", "links": { "openreview": "https://openreview.net/forum?id=MiPIjE5onj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-310", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Tackling Biased Evaluators in Dueling Bandits", "authors": [ "Ming Tang", "Yuxuan Zhou", "Chao Huang" ], "affiliation": "", "summary": "In dueling bandits, an agent explores and exploits choices (i", "links": { "openreview": "https://openreview.net/forum?id=ThgoX1dMeM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-311", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Advanced Sign Language Video Generation with Compressed and Quantized Multi-Condition Tokenization", "authors": [ "Cong Wang", "Zexuan Deng", "Zhiwei Jiang", "Yafeng Yin", "Fei Shen", "Zifeng Cheng", "Shiping Ge", "Shiwei Gan", "Qing Gu" ], "affiliation": "", "summary": "Sign Language Video Generation (SLVG) seeks to generate identity-preserving sign language videos from spoken language texts", "links": { "openreview": "https://openreview.net/forum?id=6FHvr5hJdd", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-312", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transstratal Adversarial Attack: Compromising Multi-Layered Defenses in Text-to-Image Models", "authors": [ "Chunlong Xie", "Kangjie Chen", "Shangwei Guo", "Shudong Zhang", "Tianwei Zhang", "Tao Xiang" ], "affiliation": "", "summary": "Modern Text-to-Image (T2I) models deploy multi-layered defenses to block Not-Safe-For-Work (NSFW) content generation", "links": { "openreview": "https://openreview.net/forum?id=JIpKkzSqly", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-313", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SORTeD Rashomon Sets of Sparse Decision Trees: Anytime Enumeration", "authors": [ "Elif Arslan", "Jacobus G. M. van der Linden", "Serge Hoogendoorn", "Marco Rinaldi", "Emir Demirović" ], "affiliation": "", "summary": "Sparse decision tree learning provides accurate and interpretable predictive models that are ideal for high-stakes applications by finding the single most accurate tree within a (soft) size limit", "links": { "openreview": "https://openreview.net/forum?id=Gibq7Wa7Bq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-314", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TGA: True-to-Geometry Avatar Dynamic Reconstruction", "authors": [ "Bo Guo", "Sijia Wen", "Ziwei Wang", "Yifan Zhao" ], "affiliation": "", "summary": "Recent advances in 3D Gaussian Splatting (3DGS) have improved the visual fidelity of dynamic avatar reconstruction", "links": { "openreview": "https://openreview.net/forum?id=EyFrTjaYU3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-315", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Neighbor-aware Contrastive Disambiguation for Cross-Modal Hashing with Redundant Annotations", "authors": [ "Chao Su", "Likang Peng", "Yuan Sun", "Dezhong Peng", "Xi Peng", "Xu Wang" ], "affiliation": "", "summary": "Cross-modal hashing aims to efficiently retrieve information across different modalities by mapping data into compact hash codes", "links": { "openreview": "https://openreview.net/forum?id=Bi1udlTjMb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-316", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Minimax Adaptive Online Nonparametric Regression over Besov spaces", "authors": [ "Paul Liautaud", "Pierre Gaillard", "Olivier Wintenberger" ], "affiliation": "", "summary": "We study online adversarial regression with convex losses against a rich class of continuous yet highly irregular competitor functions,% prediction rules, modeled by Besov spaces $B_{pq}^s$ with gene", "links": { "openreview": "https://openreview.net/forum?id=nKuFQhKZtt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-317", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UniSite: The First Cross-Structure Dataset and Learning Framework for End-to-End Ligand Binding Site Detection", "authors": [ "Jigang Fan", "QuanLin Wu", "Shengjie Luo", "Liwei Wang" ], "affiliation": "", "summary": "The detection of ligand binding sites for proteins is a fundamental step in Structure-Based Drug Design", "links": { "openreview": "https://openreview.net/forum?id=APXcX7z1Bi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-318", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "URDF-Anything: Constructing Articulated Objects with 3D Multimodal Language Model", "authors": [ "Zhe Li", "Xiang Bai", "Jieyu Zhang", "Zhuangzhe Wu", "Che Xu", "Ying Li", "Chengkai Hou", "Shanghang Zhang" ], "affiliation": "", "summary": "Constructing accurate digital twins of articulated objects is essential for robotic simulation training and embodied AI world model building, yet historically requires painstaking manual modeling or m", "links": { "openreview": "https://openreview.net/forum?id=g3EF5XsapH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-319", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "L2DGCN: Learnable Enhancement and Label Selection Dynamic Graph Convolutional Networks for Mitigating Degree Bias", "authors": [ "jingxiao zhang", "Shifei Ding", "Jian Zhang", "Lili Guo", "Xuan Li" ], "affiliation": "", "summary": "Graph Neural Networks (GNNs) are powerful models for node classification, but their performance is heavily reliant on manually labeled data, which is often costly and results in insufficient labeling", "links": { "openreview": "https://openreview.net/forum?id=PoIhCjqzn0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-320", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Memo: Training Memory-Efficient Embodied Agents with Reinforcement Learning", "authors": [ "Gunshi Gupta", "Karmesh Yadav", "Zsolt Kira", "Yarin Gal", "Rahaf Aljundi" ], "affiliation": "", "summary": "To enable embodied agents to operate effectively over extended timeframes, it is crucial to develop models that form and access memories to stay contextualized in their environment", "links": { "openreview": "https://openreview.net/forum?id=9eIntNc69t", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-321", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Achilles' Heel of Mamba: Essential difficulties of the Mamba architecture demonstrated by synthetic data", "authors": [ "Tianyi Chen", "Pengxiao Lin", "Zhiwei Wang", "Zhi-Qin John Xu" ], "affiliation": "", "summary": "State Space Models (SSMs) have emerged as promising alternatives to attention mechanisms, with the Mamba architecture demonstrating impressive performance and linear complexity for processing long seq", "links": { "openreview": "https://openreview.net/forum?id=4p28lkk44b", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-322", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RepLDM: Reprogramming Pretrained Latent Diffusion Models for High-Quality, High-Efficiency, High-Resolution Image Generation", "authors": [ "Boyuan Cao", "Jiaxin Ye", "Yujie Wei", "Hongming Shan" ], "affiliation": "", "summary": "While latent diffusion models (LDMs), such as Stable Diffusion, are designed for high-resolution image generation, they often struggle with significant structural distortions when generating images at", "links": { "openreview": "https://openreview.net/forum?id=QwXpn5IPKk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-323", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Understanding LLM Behaviors via Compression: Data Generation, Knowledge Acquisition and Scaling Laws", "authors": [ "Zhixuan Pan", "Shaowen Wang", "Liao Pengfei", "Jian Li" ], "affiliation": "", "summary": "Large Language Models (LLMs) have demonstrated remarkable capabilities across numerous tasks, yet principled explanations for their underlying mechanisms and several phenomena, such as scaling laws, h", "links": { "openreview": "https://openreview.net/forum?id=853SwC2dMZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-324", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Differentiable Hierarchical Visual Tokenization", "authors": [ "Marius Aasan", "Martine Hjelkrem-Tan", "Nico Catalano", "Changkyu Choi", "Adín Ramírez Rivera" ], "affiliation": "", "summary": "Vision Transformers rely on fixed patch tokens that ignore the spatial and semantic structure of images", "links": { "openreview": "https://openreview.net/forum?id=y8VWYf5cVI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-325", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Trajectory Graph Learning: Aligning with Long Trajectories in Reinforcement Learning Without Reward Design", "authors": [ "Yunfan Li", "Eric Liu", "Lin Yang" ], "affiliation": "", "summary": "Reinforcement learning (RL) often relies on manually designed reward functions, which are difficult to specify and can lead to issues such as reward hacking and suboptimal behavior", "links": { "openreview": "https://openreview.net/forum?id=QtnCPZMxYg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-326", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Meta CLIP 2: A Worldwide Scaling Recipe", "authors": [ "Yung-Sung Chuang", "Yang Li", "Dong Wang", "Ching-Feng Yeh", "Kehan Lyu", "Ramya Raghavendra", "James R. Glass", "LIFEI HUANG", "Jason E Weston", "Luke Zettlemoyer", "Xinlei Chen", "Zhuang Liu", "Saining Xie", "Wen-tau Yih", "Shang-Wen Li", "Hu Xu" ], "affiliation": "", "summary": "Contrastive Language-Image Pretraining (CLIP) is a popular foundation model, supporting from zero-shot classification, retrieval to encoders for multimodal large language models (MLLMs)", "links": { "openreview": "https://openreview.net/forum?id=aYRNINhNGV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-327", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Policy Compatible Skill Incremental Learning via Lazy Learning Interface", "authors": [ "Daehee Lee", "Dongsu Lee", "TaeYoon Kwack", "Wonje Choi", "Honguk Woo" ], "affiliation": "", "summary": "Skill Incremental Learning (SIL) is the process by which an embodied agent expands and refines its skill set over time by leveraging experience gained through interaction with its environment or by th", "links": { "openreview": "https://openreview.net/forum?id=xmYT1JqVpj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-328", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Estimating cognitive biases with attention-aware inverse planning", "authors": [ "Sounak Banerjee", "Daphne Cornelisse", "Deepak Edakkattil Gopinath", "Emily Sumner", "Jonathan DeCastro", "Guy Rosman", "Eugene Vinitsky", "Mark K Ho" ], "affiliation": "", "summary": "People's goal-directed behaviors are influenced by their cognitive biases, and autonomous systems that interact with people should be aware of this", "links": { "openreview": "https://openreview.net/forum?id=lNPo3FAMsl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-329", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "An Efficient Orlicz-Sobolev Approach for Transporting Unbalanced Measures on a Graph", "authors": [ "Tam Le", "Truyen Nguyen", "Hideitsu Hino", "Kenji Fukumizu" ], "affiliation": "", "summary": "We investigate optimal transport (OT) for measures on graph metric spaces with different total masses", "links": { "openreview": "https://openreview.net/forum?id=VymXLPX6Ps", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-330", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DeepDiver: Adaptive Web-Search Intensity Scaling via Reinforcement Learning", "authors": [ "Wenxuan Shi", "Haochen Tan", "Chuqiao Kuang", "Xiaoguang Li", "Hanting Chen", "Xiaozhe Ren", "Yasheng Wang", "Lu Hou", "Lifeng Shang" ], "affiliation": "", "summary": "Information seeking demands iterative evidence gathering and reflective reasoning, yet large language models (LLMs) still struggle with it in open-web question answering", "links": { "openreview": "https://openreview.net/forum?id=CqLWckpTbG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-331", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MetaGS: A Meta-Learned Gaussian-Phong Model for Out-of-Distribution 3D Scene Relighting", "authors": [ "Yumeng He", "Yunbo Wang" ], "affiliation": "", "summary": "Out-of-distribution (OOD) 3D relighting requires novel view synthesis under unseen lighting conditions that differ significantly from the observed images", "links": { "openreview": "https://openreview.net/forum?id=UXc87Orcri", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-332", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards a Pairwise Ranking Model with Orderliness and Monotonicity for Label Enhancement", "authors": [ "Yunan Lu", "Xixi Zhang", "Yaojin Lin", "Weiwei Li", "Lei Yang", "Xiuyi Jia" ], "affiliation": "", "summary": "Label distribution in recent years has been applied in a diverse array of complex decision-making tasks", "links": { "openreview": "https://openreview.net/forum?id=4WQ5Qgpl2F", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-333", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Spend Wisely: Maximizing Post-Training Gains in Iterative Synthetic Data Bootstrapping", "authors": [ "Pu Yang", "Yunzhen Feng", "Ziyuan Chen", "Yuhang Wu", "Zhuoyuan Li" ], "affiliation": "", "summary": "Modern foundation models often undergo iterative ``bootstrapping'' in their post-training phase: a model generates synthetic data, an external verifier filters out low-quality samples, and the high-qu", "links": { "openreview": "https://openreview.net/forum?id=aqpHTPC63N", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-334", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fast-Slow Thinking GRPO for Large Vision-Language Model Reasoning", "authors": [ "Wenyi Xiao", "Leilei Gan" ], "affiliation": "", "summary": "When applying reinforcement learning—typically through GRPO—to large vision-language model reasoning struggles to effectively scale reasoning length or generates verbose outputs across all tasks with", "links": { "openreview": "https://openreview.net/forum?id=MI1uT5rReV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-335", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Exploration via Feature Perturbation in Contextual Bandits", "authors": [ "Seouh-won Yi", "Min-hwan Oh" ], "affiliation": "", "summary": "We propose *feature perturbation*, a simple yet effective exploration strategy for contextual bandits that injects randomness directly into feature inputs, instead of randomizing unknown parameters or", "links": { "openreview": "https://openreview.net/forum?id=gAddPMjmUc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-336", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Instance-Optimality for Private KL Distribution Estimation", "authors": [ "Jiayuan Ye", "Vitaly Feldman", "Kunal Talwar" ], "affiliation": "", "summary": "We study the fundamental problem of estimating an unknown discrete distribution $p$ over $d$ symbols, given $n$ i", "links": { "openreview": "https://openreview.net/forum?id=K0FbK2GOGj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-337", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DMWM: Dual-Mind World Model with Long-Term Imagination", "authors": [ "Lingyi Wang", "Rashed Shelim", "Walid Saad", "Naren Ramakrishnan" ], "affiliation": "", "summary": "Imagination in world models is crucial for enabling agents to learn long-horizon policy in a sample-efficient manner", "links": { "openreview": "https://openreview.net/forum?id=Bzlt5tPFT6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-338", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TimeWak: Temporal Chained-Hashing Watermark for Time Series Data", "authors": [ "Zhi Wen Soi", "Chaoyi Zhu", "Fouad Abiad", "Aditya Shankar", "Jeroen M. Galjaard", "Huijuan Wang", "Lydia Y. Chen" ], "affiliation": "", "summary": "Synthetic time series generated by diffusion models enable sharing privacy-sensitive datasets, such as patients' functional MRI records", "links": { "openreview": "https://openreview.net/forum?id=akhhwQh6UV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-339", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface", "authors": [ "Hao Tang", "Chen-Wei Xie", "Haiyang Wang", "Xiaoyi Bao", "Tingyu Weng", "Pandeng Li", "Yun Zheng", "Liwei Wang" ], "affiliation": "", "summary": "Generalist models have achieved remarkable success in both language and vision-language tasks, showcasing the potential of unified modeling", "links": { "openreview": "https://openreview.net/forum?id=8omLr8BtjL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-340", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DenseDPO: Fine-Grained Temporal Preference Optimization for Video Diffusion Models", "authors": [ "Ziyi Wu", "Anil Kag", "Ivan Skorokhodov", "Willi Menapace", "Ashkan Mirzaei", "Igor Gilitschenski", "Sergey Tulyakov", "Aliaksandr Siarohin" ], "affiliation": "", "summary": "Direct Preference Optimization (DPO) has recently been applied as a post‑training technique for text-to-video diffusion models", "links": { "openreview": "https://openreview.net/forum?id=YFa7eULIeN", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-341", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Precise Asymptotics and Refined Regret of Variance-Aware UCB", "authors": [ "Yingying Fan", "Yuxuan Han", "Jinchi Lv", "Xiaocong XU", "Zhengyuan Zhou" ], "affiliation": "", "summary": "In this paper, we study the behavior of the Upper Confidence Bound-Variance (UCB-V) algorithm for the Multi-Armed Bandit (MAB) problems, a variant of the canonical Upper Confidence Bound (UCB) algorit", "links": { "openreview": "https://openreview.net/forum?id=A0M3apV5zI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-342", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "InfMasking: Unleashing Synergistic Information by Contrastive Multimodal Interactions", "authors": [ "Liangjian Wen", "Qun Dai", "Jianzhuang Liu", "Jiangtao Zheng", "Yong Dai", "Dongkai Wang", "zhao kang", "Jun Wang", "Zenglin Xu", "Jiang Duan" ], "affiliation": "", "summary": "In multimodal representation learning, synergistic interactions between modalities not only provide complementary information but also create unique outcomes through specific interaction patterns that", "links": { "openreview": "https://openreview.net/forum?id=6AbM9UG4aD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-343", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Extracting task-relevant preserved dynamics from contrastive aligned neural recordings", "authors": [ "Yiqi Jiang", "Kaiwen Sheng", "Yujia Gao", "E. Kelly Buchanan", "Yu Shikano", "Seung Je Woo", "Yixiu Zhao", "Tony Hyun Kim", "Fatih Dinc", "Scott Linderman", "Mark Schnitzer" ], "affiliation": "", "summary": "Recent work indicates that low-dimensional dynamics of neural and behavioral data are often preserved across days and subjects", "links": { "openreview": "https://openreview.net/forum?id=uvTea5Rfek", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-344", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Orochi: Versatile Biomedical Image Processor", "authors": [ "Gaole Dai", "Chenghao Zhou", "Yu Zhou", "Rongyu Zhang", "Yuan Zhang", "Chengkai Hou", "Tiejun Huang", "Jianxu Chen", "Shanghang Zhang" ], "affiliation": "", "summary": "Deep learning has emerged as a pivotal tool for accelerating research in the life sciences, with the low-level processing of biomedical images (e", "links": { "openreview": "https://openreview.net/forum?id=Rtd6GoJcoT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-345", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "WISA: World simulator assistant for physics-aware text-to-video generation", "authors": [ "Jing Wang", "Ao Ma", "Ke Cao", "Jun Zheng", "Jiasong Feng", "Zhanjie Zhang", "Wanyuan Pang", "Xiaodan Liang" ], "affiliation": "", "summary": "Recent advances in text-to-video (T2V) generation, exemplified by models such as Sora and Kling, have demonstrated strong potential for constructing world simulators", "links": { "openreview": "https://openreview.net/forum?id=4jWuS5hye1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-346", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Conservative classifiers do consistently well with improving agents: characterizing statistical and online learning", "authors": [ "Dravyansh Sharma", "Alec Sun" ], "affiliation": "", "summary": "Machine learning is now ubiquitous in societal decision-making, for example in evaluating job candidates or loan applications, and it is increasingly important to take into account how classified agen", "links": { "openreview": "https://openreview.net/forum?id=Zc5Ntjt5Ul", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-347", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "KORGym: A Dynamic Game Platform for LLM Reasoning Evaluation", "authors": [ "Jiajun Shi", "Jian Yang", "Jiaheng Liu", "Xingyuan Bu", "Jiangjie Chen", "Junting Zhou", "Kaijing Ma", "Zhoufutu Wen", "Bingli Wang", "Yancheng He", "Liang Song", "Hualei Zhu", "Shilong Li", "Xingjian Wang", "Wei Zhang", "Ruibin Yuan", "Yifan Yao", "Wenjun Yang", "Yunli Wang", "Siyuan Fang", "Siyu Yuan", "Qianyu He", "Xiangru Tang", "Yingshui Tan", "Wangchunshu Zhou", "Zhaoxiang Zhang", "Zhoujun Li", "Wenhao Huang", "Ge Zhang" ], "affiliation": "", "summary": "Recent advancements in large language models (LLMs) underscore the need for more comprehensive evaluation methods to accurately assess their reasoning capabilities", "links": { "openreview": "https://openreview.net/forum?id=uAeqQePu4c", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-348", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learnable Sampler Distillation for Discrete Diffusion Models", "authors": [ "Feiyang Fu", "Tongxian Guo", "Zhaoqiang Liu" ], "affiliation": "", "summary": "Discrete diffusion models (DDMs) have shown powerful generation ability for discrete data modalities like text and molecules", "links": { "openreview": "https://openreview.net/forum?id=gMHLQASj11", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-349", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Hardness of Approximating Distributions with Tractable Probabilistic Models", "authors": [ "John Leland", "YooJung Choi" ], "affiliation": "", "summary": "A fundamental challenge in probabilistic modeling is to balance expressivity and inference efficiency", "links": { "openreview": "https://openreview.net/forum?id=0GvEaa9prl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-350", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MoESD: Unveil Speculative Decoding's Potential for Accelerating Sparse MoE", "authors": [ "Zongle Huang", "Lei Zhu", "ZongYuan Zhan", "Ting Hu", "Weikai Mao", "Xianzhi Yu", "Yongpan Liu", "Tianyu Zhang" ], "affiliation": "", "summary": "Large Language Models (LLMs) have achieved remarkable success across many applications, with Mixture of Experts (MoE) models demonstrating great potential", "links": { "openreview": "https://openreview.net/forum?id=FAeU7516MR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-351", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement", "authors": [ "Xiyao Wang", "Zhengyuan Yang", "Chao Feng", "Hongjin Lu", "Linjie Li", "Chung-Ching Lin", "Kevin Lin", "Furong Huang", "Lijuan Wang" ], "affiliation": "", "summary": "We introduce ThinkLite-VL, a family of visual reasoning models that achieve state-of-the-art (SoTA) performance using an order of magnitude fewer training samples, relying purely on reinforcement fine", "links": { "openreview": "https://openreview.net/forum?id=PHu9xJeAum", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-352", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Deno-IF: Unsupervised Noisy Visible and Infrared Image Fusion Method", "authors": [ "Han Xu", "Yuyang Li", "Yunfei Deng", "Jiayi Ma", "Guangcan Liu" ], "affiliation": "", "summary": "Most image fusion methods are designed for ideal scenarios and struggle to handle noise", "links": { "openreview": "https://openreview.net/forum?id=36cKp4tsHF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-353", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ReCon: Region-Controllable Data Augmentation with Rectification and Alignment for Object Detection", "authors": [ "Haowei Zhu", "Tianxiang Pan", "Rui Qin", "Jun-Hai Yong", "Bin Wang" ], "affiliation": "", "summary": "The scale and quality of datasets are crucial for training robust perception models", "links": { "openreview": "https://openreview.net/forum?id=2zjH76SmiF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-354", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Repurposing Marigold for Zero-Shot Metric Depth Estimation via Defocus Blur Cues", "authors": [ "Chinmay Talegaonkar", "Nikhil Gandudi Suresh", "Zachary Novack", "Yash Belhe", "Priyanka Nagasamudra", "Nicholas Antipa" ], "affiliation": "", "summary": "Recent monocular metric depth estimation (MMDE) methods have made notable progress towards zero-shot generalization", "links": { "openreview": "https://openreview.net/forum?id=imO353Gyrl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-355", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PARTONOMY: Large Multimodal Models with Part-Level Visual Understanding", "authors": [ "Ansel Blume", "Jeonghwan Kim", "Hyeonjeong Ha", "Elen Chatikyan", "Xiaomeng Jin", "Khanh Duy Nguyen", "Nanyun Peng", "Kai-Wei Chang", "Derek Hoiem", "Heng Ji" ], "affiliation": "", "summary": "Real-world objects are composed of distinctive, object-specific parts", "links": { "openreview": "https://openreview.net/forum?id=yjLew3Nd7z", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-356", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Composite Flow Matching for Reinforcement Learning with Shifted-Dynamics Data", "authors": [ "Lingkai Kong", "Haichuan Wang", "Tonghan Wang", "GUOJUN XIONG", "Milind Tambe" ], "affiliation": "", "summary": "Incorporating pre-collected offline data can substantially improve the sample efficiency of reinforcement learning (RL), but its benefits can break down when the transition dynamics in the offline dat", "links": { "openreview": "https://openreview.net/forum?id=7cPDOBWTbM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-357", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Asymmetric Duos: Sidekicks Improve Uncertainty", "authors": [ "Tim G. Zhou", "Evan Shelhamer", "Geoff Pleiss" ], "affiliation": "", "summary": "The go-to strategy to apply deep networks in settings where uncertainty informs decisions—ensembling multiple training runs with random initializations—is ill-suited for the extremely large-scale mode", "links": { "openreview": "https://openreview.net/forum?id=9CzRx5MZct", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-358", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Robust learning of halfspaces under log-concave marginals", "authors": [ "Jane Lange", "Arsen Vasilyan" ], "affiliation": "", "summary": "We say that a classifier is $\\text{\\emph{adversarially robust}}$ to perturbations of norm $r$ if, with high probability over a point $x$ drawn from the input distribution, there is no point within dis", "links": { "openreview": "https://openreview.net/forum?id=ZrCQGVpQrl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-359", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Implicit Bias of Spectral Descent and Muon on Multiclass Separable Data", "authors": [ "Chen Fan", "Mark Schmidt", "Christos Thrampoulidis" ], "affiliation": "", "summary": "Different gradient-based methods for optimizing overparameterized models can all achieve zero training error yet converge to distinctly different solutions inducing different generalization properties", "links": { "openreview": "https://openreview.net/forum?id=Zn2ajV1kTQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-360", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Lost in Transmission: When and Why LLMs Fail to Reason Globally", "authors": [ "Tobias Schnabel", "Kiran Tomlinson", "Adith Swaminathan", "Jennifer Neville" ], "affiliation": "", "summary": "Despite their many successes, transformer-based large language models (LLMs) continue to struggle with tasks that require complex reasoning over large parts of their input", "links": { "openreview": "https://openreview.net/forum?id=MaJ3ASZ0NI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-361", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On Transferring Transferability: Towards a Theory for Size Generalization", "authors": [ "Eitan Levin", "Yuxin Ma", "Mateo Diaz Diaz", "Soledad Villar" ], "affiliation": "", "summary": "Many modern learning tasks require models that can take inputs of varying sizes", "links": { "openreview": "https://openreview.net/forum?id=Iic9I2nHdZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-362", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "When Worse is Better: Navigating the Compression Generation Trade-off In Visual Tokenization", "authors": [ "Vivek Ramanujan", "Kushal Tirumala", "Armen Aghajanyan", "Luke Zettlemoyer", "Ali Farhadi" ], "affiliation": "", "summary": "Current image generation methods are based on a two-stage training approach", "links": { "openreview": "https://openreview.net/forum?id=o8hWyJIgAV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-363", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FlowFeat: Pixel-Dense Embedding of Motion Profiles", "authors": [ "Nikita Araslanov", "Anna Ribic", "Daniel Cremers" ], "affiliation": "", "summary": "Dense and versatile image representations underpin the success of virtually all computer vision applications", "links": { "openreview": "https://openreview.net/forum?id=gZsmYwFHci", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-364", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Generative Leap: Tight Sample Complexity for Efficiently Learning Gaussian Multi-Index Models", "authors": [ "Alex Damian", "Jason D. Lee", "Joan Bruna" ], "affiliation": "", "summary": "In this work we consider generic Gaussian Multi-index models, in which the labels only depend on the (Gaussian) $d$-dimensional inputs through their projection onto a low-dimensional $r = O_d(1)$ subs", "links": { "openreview": "https://openreview.net/forum?id=5X6PL4906S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-365", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Incremental Sequence Classification with Temporal Consistency", "authors": [ "Lucas Maystre", "Gabriel Barello", "Tudor Berariu", "Aleix Cambray", "Rares Dolga", "Alvaro Ortega Gonzalez", "Andrei Cristian Nica", "David Barber" ], "affiliation": "", "summary": "We address the problem of incremental sequence classification, where predictions are updated as new elements in the sequence are revealed", "links": { "openreview": "https://openreview.net/forum?id=bTssV4Cnjn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-366", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LODGE: Level-of-Detail Large-Scale Gaussian Splatting with Efficient Rendering", "authors": [ "Jonas Kulhanek", "Marie-Julie Rakotosaona", "Fabian Manhardt", "Christina Tsalicoglou", "Michael Niemeyer", "Torsten Sattler", "Songyou Peng", "Federico Tombari" ], "affiliation": "", "summary": "In this work, we present a novel level-of-detail (LOD) method for 3D Gaussian Splatting that enables real-time rendering of large-scale scenes on memory-constrained devices", "links": { "openreview": "https://openreview.net/forum?id=Iqu63cYI3z", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-367", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Solving Neural Min-Max Games: The Role of Architecture, Initialization & Dynamics", "authors": [ "Deep Patel", "Emmanouil-Vasileios Vlatakis-Gkaragkounis" ], "affiliation": "", "summary": "Many emerging applications—such as adversarial training, AI alignment, and robust optimization—can be framed as zero-sum games between neural nets, with von Neumann–Nash equilibria (NE) capturing the", "links": { "openreview": "https://openreview.net/forum?id=5xdbWUdM87", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-368", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Provable Gradient Editing of Deep Neural Networks", "authors": [ "Zhe Tao", "Aditya V. Thakur" ], "affiliation": "", "summary": "In explainable AI, DNN gradients are used to interpret the prediction; in safety-critical control systems, gradients could encode safety constraints; in scientific-computing applications, gradients co", "links": { "openreview": "https://openreview.net/forum?id=1ffIkWo0yq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-369", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models", "authors": [ "Sreyan Ghosh", "Arushi Goel", "Jaehyeon Kim", "Sonal Kumar", "Zhifeng Kong", "Sang-gil Lee", "Chao-Han Huck Yang", "Ramani Duraiswami", "Dinesh Manocha", "Rafael Valle", "Bryan Catanzaro" ], "affiliation": "", "summary": "We present Audio Flamingo 3 (AF3), a fully open state-of-the-art (SOTA) large audio-language model that advances reasoning and understanding across speech, sound, and music", "links": { "openreview": "https://openreview.net/forum?id=FjByDpDVIO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-370", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RGB-Only Supervised Camera Parameter Optimization in Dynamic Scenes", "authors": [ "Fang Li", "Hao Zhang", "Narendra Ahuja" ], "affiliation": "", "summary": "Although COLMAP has long remained the predominant method for camera parameter optimization in static scenes, it is constrained by its lengthy runtime and reliance on ground truth (GT) motion masks for", "links": { "openreview": "https://openreview.net/forum?id=0fZoqVoc0o", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-371", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Integration Matters for Learning PDEs with Backward SDEs", "authors": [ "Sungje Park", "Stephen Tu" ], "affiliation": "", "summary": "Backward stochastic differential equation (BSDE)-based deep learning methods provide an alternative to Physics-Informed Neural Networks (PINNs) for solving high-dimensional partial differential equati", "links": { "openreview": "https://openreview.net/forum?id=k4jg1QCw0e", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-372", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Unleashing Hour-Scale Video Training for Long Video-Language Understanding", "authors": [ "Jingyang Lin", "Jialian Wu", "Ximeng Sun", "Ze Wang", "Jiang Liu", "Yusheng Su", "Xiaodong Yu", "Hao Chen", "Jiebo Luo", "Zicheng Liu", "Emad Barsoum" ], "affiliation": "", "summary": "Recent long-form video-language understanding benchmarks have driven progress in video large multimodal models (Video-LMMs)", "links": { "openreview": "https://openreview.net/forum?id=2ptM76yNzZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-373", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Neural Entropy", "authors": [ "Akhil Premkumar" ], "affiliation": "", "summary": "We explore the connection between deep learning and information theory through the paradigm of diffusion models", "links": { "openreview": "https://openreview.net/forum?id=f6AYwCvynr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-374", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Locality in Image Diffusion Models Emerges from Data Statistics", "authors": [ "Artem Lukoianov", "Chenyang Yuan", "Justin Solomon", "Vincent Sitzmann" ], "affiliation": "", "summary": "Recent work has shown that the generalization ability of image diffusion models arises from the locality properties of the trained neural network", "links": { "openreview": "https://openreview.net/forum?id=skunuOdavO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-375", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Regularized least squares learning with heavy-tailed noise is minimax optimal", "authors": [ "Mattes Mollenhauer", "Nicole Mücke", "Dimitri Meunier", "Arthur Gretton" ], "affiliation": "", "summary": "This paper examines the performance of ridge regression in reproducing kernel Hilbert spaces in the presence of noise that exhibits a finite number of higher moments", "links": { "openreview": "https://openreview.net/forum?id=TjQP5hc3WC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-376", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On Feasible Rewards in Multi-Agent Inverse Reinforcement Learning", "authors": [ "Till Freihaut", "Giorgia Ramponi" ], "affiliation": "", "summary": "Multi-agent inverse reinforcement learning (MAIRL) aims to recover agent reward functions from expert demonstrations", "links": { "openreview": "https://openreview.net/forum?id=qu6mRbSnUs", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-377", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MonarchAttention: Zero-Shot Conversion to Fast, Hardware-Aware Structured Attention", "authors": [ "Can Yaras", "Alec S Xu", "Pierre Abillama", "Changwoo Lee", "Laura Balzano" ], "affiliation": "", "summary": "Transformers have achieved state-of-the-art performance across various tasks, but suffer from a notable quadratic complexity in sequence length due to the attention mechanism", "links": { "openreview": "https://openreview.net/forum?id=XfHfTqeXfZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-378", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness", "authors": [ "Sahar Abdelnabi", "Ahmed Salem" ], "affiliation": "", "summary": "Reasoning-focused LLMs sometimes alter their behavior when they detect that they are being evaluated—which can lead them to optimize for test-passing performance or to comply more readily with harmful", "links": { "openreview": "https://openreview.net/forum?id=ccPts3Df2q", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-379", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Measuring and Controlling Solution Degeneracy across Task-Trained Recurrent Neural Networks", "authors": [ "Ann Huang", "Satpreet Harcharan Singh", "Flavio Martinelli", "Kanaka Rajan" ], "affiliation": "", "summary": "Task-trained recurrent neural networks (RNNs) are widely used in neuroscience and machine learning to model dynamical computations", "links": { "openreview": "https://openreview.net/forum?id=R0LqbSgZjP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-380", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Reinforcement Learning with Imperfect Transition Predictions: A Bellman-Jensen Approach", "authors": [ "Chenbei Lu", "Zaiwei Chen", "Tongxin Li", "Chenye Wu", "Adam Wierman" ], "affiliation": "", "summary": "Traditional reinforcement learning (RL) assumes the agents make decisions based on Markov decision processes (MDPs) with one-step transition models", "links": { "openreview": "https://openreview.net/forum?id=DYuPwwDy9n", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-381", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AI Research Agents for Machine Learning: Search, Exploration, and Generalization in MLE-bench", "authors": [ "Edan Toledo", "Karen Hambardzumyan", "Martin Josifoski", "RISHI HAZRA", "Nicolas Baldwin", "Alexis Audran-Reiss", "Michael Kuchnik", "Despoina Magka", "Minqi Jiang", "Alisia Maria Lupidi", "Andrei Lupu", "Roberta Raileanu", "Tatiana Shavrina", "Kelvin Niu", "Jean-Christophe Gagnon-Audet", "Michael Shvartsman", "Shagun Sodhani", "Alexander H Miller", "Abhishek Charnalia", "Derek Dunfield", "Carole-Jean Wu", "Pontus Stenetorp", "Nicola Cancedda", "Jakob Nicolaus Foerster", "Yoram Bachrach" ], "affiliation": "", "summary": "AI research agents are demonstrating great potential to accelerate scientific progress by automating the design, implementation, and training of machine learning models", "links": { "openreview": "https://openreview.net/forum?id=RwfrdKSgCE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-382", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Differentiable Cyclic Causal Discovery Under Unmeasured Confounders", "authors": [ "Muralikrishnna Guruswamy Sethuraman", "Faramarz Fekri" ], "affiliation": "", "summary": "Understanding causal relationships between variables is fundamental across scientific disciplines", "links": { "openreview": "https://openreview.net/forum?id=95z3psF9zJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-383", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ELECTRA: A Cartesian Network for 3D Charge Density Prediction with Floating Orbitals", "authors": [ "Jonas Elsborg", "Luca Thiede", "Alan Aspuru-Guzik", "Tejs Vegge", "Arghya Bhowmik" ], "affiliation": "", "summary": "We present the Electronic Tensor Reconstruction Algorithm (ELECTRA) - an equivariant model for predicting electronic charge densities using floating orbitals", "links": { "openreview": "https://openreview.net/forum?id=Rz9ISbLxf0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-384", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Compositional Neural Network Verification via Assume-Guarantee Reasoning", "authors": [ "Hai Duong", "David Shriver", "ThanhVu Nguyen", "Matthew B. Dwyer" ], "affiliation": "", "summary": "Verifying the behavior of neural networks is necessary if developers are to confidently deploy them as parts of mission-critical systems", "links": { "openreview": "https://openreview.net/forum?id=WbpXT0WL9S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-385", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A machine learning approach that beats Rubik's cubes", "authors": [ "Alexander Chervov", "Kirill Khoruzhii", "Nikita Bukhal", "Jalal Naghiyev", "Vladislav Zamkovoy", "Ivan Koltsov", "Lyudmila Cheldieva", "Arsenii Sychev", "Arsenii Lenin", "Mark Obozov", "Egor Urvanov", "Alexey M. Romanov" ], "affiliation": "", "summary": "The paper proposes a novel machine learning-based approach to the pathfinding problem on extremely large graphs", "links": { "openreview": "https://openreview.net/forum?id=31CaYYw1Xz", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-386", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ENMA: Tokenwise Autoregression for Continuous Neural PDE Operators", "authors": [ "Armand Kassaï Koupaï", "Lise Le Boudec", "Louis Serrano", "Patrick Gallinari" ], "affiliation": "", "summary": "Solving time-dependent parametric partial differential equations (PDEs) remains a fundamental challenge for neural solvers, particularly when generalizing across a wide range of physical parameters an", "links": { "openreview": "https://openreview.net/forum?id=3CYXSMFv55", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-387", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Critical Batch Size Revisited: A Simple Empirical Approach to Large-Batch Language Model Training", "authors": [ "William Merrill", "Shane Arora", "Dirk Groeneveld", "Hannaneh Hajishirzi" ], "affiliation": "", "summary": "The right batch size is important when training language models at scale: a large batch size is necessary for fast training, but a batch size that is *too large* will harm token efficiency", "links": { "openreview": "https://openreview.net/forum?id=XUKUx7Xu89", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-388", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Shift Before You Learn: Enabling Low-Rank Representations in Reinforcement Learning", "authors": [ "Bastien Dubail", "Stefan Stojanovic", "Alexandre Proutiere" ], "affiliation": "", "summary": "Low-rank structure is a common implicit assumption in many modern reinforcement learning (RL) algorithms", "links": { "openreview": "https://openreview.net/forum?id=loVvFhfsDf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-389", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Stochastic Optimization in Semi-Discrete Optimal Transport: Convergence Analysis and Minimax Rate", "authors": [ "Ferdinand Genans", "Antoine Godichon-Baggioni", "François-Xavier Vialard", "Olivier Wintenberger" ], "affiliation": "", "summary": "We investigate the semi-discrete Optimal Transport (OT) problem, where a continuous source measure $\\mu$ is transported to a discrete target measure $\\nu$, with particular attention to the OT map appr", "links": { "openreview": "https://openreview.net/forum?id=fGfl6dqQVf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-390", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transferable Black-Box One-Shot Forging of Watermarks via Image Preference Models", "authors": [ "Tomas Soucek", "Sylvestre-Alvise Rebuffi", "Pierre Fernandez", "Nikola Jovanović", "Hady Elsahar", "Valeriu Lacatusu", "Tuan A. Tran", "Alexandre Mourachko" ], "affiliation": "", "summary": "Recent years have seen a surge in interest in digital content watermarking techniques, driven by the proliferation of generative models and increased legal pressure", "links": { "openreview": "https://openreview.net/forum?id=yb5JOOmfxA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-391", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Non-Asymptotic Analysis Of Data Augmentation For Precision Matrix Estimation", "authors": [ "Lucas Morisset", "Adrien Hardy", "Alain Oliviero Durmus" ], "affiliation": "", "summary": "This paper addresses the problem of inverse covariance (also known as precision matrix) estimation in high-dimensional settings", "links": { "openreview": "https://openreview.net/forum?id=UwtFuWbW6B", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-392", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Improving Perturbation-based Explanations by Understanding the Role of Uncertainty Calibration", "authors": [ "Thomas Decker", "Volker Tresp", "Florian Buettner" ], "affiliation": "", "summary": "Perturbation-based explanations are widely utilized to enhance the transparency of machine-learning models in practice", "links": { "openreview": "https://openreview.net/forum?id=AjOl3iahHd", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-393", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ConTextTab: A Semantics-Aware Tabular In-Context Learner", "authors": [ "Marco Spinaci", "Marek Polewczyk", "Maximilian Schambach", "Sam Thelin" ], "affiliation": "", "summary": "Tabular in-context learning (ICL) has recently achieved state-of-the-art (SOTA) performance on several tabular prediction tasks", "links": { "openreview": "https://openreview.net/forum?id=kGMRb4jbTP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-394", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Some Optimizers are More Equal: Understanding the Role of Optimizers in Group Fairness", "authors": [ "Mojtaba Kolahdouzi", "Hatice Gunes", "Ali Etemad" ], "affiliation": "", "summary": "We study whether and how the choice of optimization algorithm can impact group fairness in deep neural networks", "links": { "openreview": "https://openreview.net/forum?id=s3WyfnHw6B", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-395", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Less is More: Improving LLM Alignment via Preference Data Selection", "authors": [ "Xun Deng", "Han Zhong", "Rui Ai", "Fuli Feng", "Zheng Wang", "Xiangnan He" ], "affiliation": "", "summary": "Direct Preference Optimization (DPO) has emerged as a promising approach for aligning large language models with human preferences", "links": { "openreview": "https://openreview.net/forum?id=R2ZJSjLDJC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-396", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Stable Gradients for Stable Learning at Scale in Deep Reinforcement Learning", "authors": [ "Roger Creus Castanyer", "Johan Obando-Ceron", "Lu Li", "Pierre-Luc Bacon", "Glen Berseth", "Aaron Courville", "Pablo Samuel Castro" ], "affiliation": "", "summary": "Scaling deep reinforcement learning networks is challenging and often results in degraded performance, yet the root causes of this failure mode remain poorly understood", "links": { "openreview": "https://openreview.net/forum?id=Vqj65VeDOu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-397", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SpecMER: Fast Protein Generation with K-mer Guided Speculative Decoding", "authors": [ "Thomas Walton", "Darin Tsui", "Aryan Musharaf", "Amirali Aghazadeh" ], "affiliation": "", "summary": "Autoregressive models have transformed protein engineering by enabling the generation of novel protein sequences beyond those found in nature", "links": { "openreview": "https://openreview.net/forum?id=2sG4ebgqBd", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-398", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SpecEdge: Scalable Edge-Assisted Serving Framework for Interactive LLMs", "authors": [ "Jinwoo Park", "Seunggeun Cho", "Dongsu Han" ], "affiliation": "", "summary": "Large language models (LLMs) power many modern applications, but serving them at scale remains costly and resource-intensive", "links": { "openreview": "https://openreview.net/forum?id=4QVLKwgg3S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-399", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DEXTER: Diffusion-Guided EXplanations with TExtual Reasoning for Vision Models", "authors": [ "Simone Carnemolla", "Matteo Pennisi", "Sarinda Samarasinghe", "Giovanni Bellitto", "Simone Palazzo", "Daniela Giordano", "Mubarak Shah", "Concetto Spampinato" ], "affiliation": "", "summary": "Understanding and explaining the behavior of machine learning models is essential for building transparent and trustworthy AI systems", "links": { "openreview": "https://openreview.net/forum?id=baBhSzaSHI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-400", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "What One Cannot, Two Can: Two-Layer Transformers Provably Represent Induction Heads on Any-Order Markov Chains", "authors": [ "Chanakya Ekbote", "Ashok Vardhan Makkuva", "Marco Bondaschi", "Nived Rajaraman", "Michael Gastpar", "Jason D. Lee", "Paul Pu Liang" ], "affiliation": "", "summary": "In-context learning (ICL) is a hallmark capability of transformers, through which trained models learn to adapt to new tasks by leveraging information from the input context", "links": { "openreview": "https://openreview.net/forum?id=nYg6Qzm5xS", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-401", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Sample-Adaptivity Tradeoff in On-Demand Sampling", "authors": [ "Nika Haghtalab", "Omar Montasser", "Mingda Qiao" ], "affiliation": "", "summary": "We study the tradeoff between sample complexity and round complexity in *on-demand sampling*, where the learning algorithm adaptively samples from $k$ distributions over a limited number of rounds", "links": { "openreview": "https://openreview.net/forum?id=gaHjGx1cMh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-402", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Online Strategic Classification With Noise and Partial Feedback", "authors": [ "Tianrun Zhao", "Xiaojie Mao", "Yong Liang" ], "affiliation": "", "summary": "In this paper, we study an online strategic classification problem, where a principal aims to learn an accurate binary linear classifier from sequentially arriving agents", "links": { "openreview": "https://openreview.net/forum?id=Oo9KXaM6Mu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-403", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Partition-Then-Adapt: Combating Prediction Bias for Reliable Multi-Modal Test-Time Adaptation", "authors": [ "Guowei Wang", "Fan Lyu", "Changxing Ding" ], "affiliation": "", "summary": "Existing test-time adaptation (TTA) methods primarily focus on scenarios involving domain shifts in a single modality", "links": { "openreview": "https://openreview.net/forum?id=T6RkYsuoMW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-404", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Strategic Hypothesis Testing", "authors": [ "Yatong Chen", "Safwan Hossain", "Yiling Chen" ], "affiliation": "", "summary": "We examine hypothesis testing within a principal-agent framework, where a strategic agent, holding private beliefs about the effectiveness of a product, submits data to a principal who decides on appr", "links": { "openreview": "https://openreview.net/forum?id=px1OlCcqkj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-405", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Accelerating Optimization via Differentiable Stopping Time", "authors": [ "Zhonglin Xie", "Yiman Fong", "Haoran Yuan", "Zaiwen Wen" ], "affiliation": "", "summary": "A common approach for accelerating optimization algorithms is to minimize the loss achieved in a fixed time, which enables a differentiable framework with respect to the algorithm's hyperparameters", "links": { "openreview": "https://openreview.net/forum?id=L51U5RSFBo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-406", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Tight Generalization Bounds for Large-Margin Halfspaces", "authors": [ "Kasper Green Larsen", "Natascha Schalburg" ], "affiliation": "", "summary": "We prove the first generalization bound for large-margin halfspaces that is asymptotically tight in the tradeoff between the margin, the fraction of training points with the given margin, the failure", "links": { "openreview": "https://openreview.net/forum?id=wAq0ZLxrGq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-407", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Word-Level Emotional Expression Control in Zero-Shot Text-to-Speech Synthesis", "authors": [ "Tianrui Wang", "Haoyu Wang", "Meng Ge", "Cheng Gong", "Chunyu Qiang", "Ziyang Ma", "Zikang Huang", "Guanrou Yang", "Xiaobao Wang", "EngSiong Chng", "Xie Chen", "Longbiao Wang", "Jianwu Dang" ], "affiliation": "", "summary": "While emotional text-to-speech (TTS) has made significant progress, most existing research remains limited to utterance-level emotional expression and fails to support word-level control", "links": { "openreview": "https://openreview.net/forum?id=SYcggdxX6W", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-408", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UMoE: Unifying Attention and FFN with Shared Experts", "authors": [ "Yuanhang Yang", "Chaozheng Wang", "Jing Li" ], "affiliation": "", "summary": "Sparse Mixture of Experts (MoE) architectures have emerged as a promising approach for scaling Transformer models", "links": { "openreview": "https://openreview.net/forum?id=2Z0OFReqkT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-409", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RF-Agent: Automated Reward Function Design via Language Agent Tree Search", "authors": [ "Ning Gao", "Xiuhui Zhang", "Xingyu Jiang", "Mukang You", "Mohan Zhang", "Yue Deng" ], "affiliation": "", "summary": "Designing efficient reward functions for low-level control tasks is a challenging problem", "links": { "openreview": "https://openreview.net/forum?id=dZ94ZS410X", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-410", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GraLoRA: Granular Low-Rank Adaptation for Parameter-Efficient Fine-Tuning", "authors": [ "Yeonjoon Jung", "Daehyun Ahn", "Hyungjun Kim", "Taesu Kim", "Eunhyeok Park" ], "affiliation": "", "summary": "Low-Rank Adaptation (LoRA) is a popular method for parameter-efficient fine-tuning (PEFT) of generative models, valued for its simplicity and effectiveness", "links": { "openreview": "https://openreview.net/forum?id=8wvOMQ2Olw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-411", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Quantization-Free Autoregressive Action Transformer", "authors": [ "Ziyad Sheebaelhamd", "Michael Tschannen", "Michael Muehlebach", "Claire Vernade" ], "affiliation": "", "summary": "Current transformer-based imitation learning approaches introduce discrete action representations and train an autoregressive transformer decoder on the resulting latent code", "links": { "openreview": "https://openreview.net/forum?id=3a18D8IeQ1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-412", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LLM-Explorer: A Plug-in Reinforcement Learning Policy Exploration Enhancement Driven by Large Language Models", "authors": [ "Qianyue Hao", "Yiwen Song", "Qingmin Liao", "Jian Yuan", "Yong Li" ], "affiliation": "", "summary": "Policy exploration is critical in reinforcement learning (RL), where existing approaches include $\\epsilon$-greedy, Gaussian process, etc", "links": { "openreview": "https://openreview.net/forum?id=VA5P0rUZPx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-413", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "BayeSQP: Bayesian Optimization through Sequential Quadratic Programming", "authors": [ "Paul Brunzema", "Sebastian Trimpe" ], "affiliation": "", "summary": "We introduce BayeSQP, a novel algorithm for general black-box optimization that merges the structure of sequential quadratic programming with concepts from Bayesian optimization", "links": { "openreview": "https://openreview.net/forum?id=FEugj28qhC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-414", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Practical do-Shapley Explanations with Estimand-Agnostic Causal Inference", "authors": [ "Álvaro Parafita", "Tomas Garriga", "Axel Brando", "Francisco J. Cazorla" ], "affiliation": "", "summary": "Among explainability techniques, SHAP stands out as one of the most popular, but often overlooks the causal structure of the problem", "links": { "openreview": "https://openreview.net/forum?id=aTiMLVePXi", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-415", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AlphaZero Neural Scaling and Zipf's Law: a Tale of Board Games and Power Laws", "authors": [ "Oren Neumann", "Claudius Gros" ], "affiliation": "", "summary": "Neural scaling laws are observed in a range of domains, to date with no universal understanding of why they occur", "links": { "openreview": "https://openreview.net/forum?id=IMmkDMqFMU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-416", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Set Smoothness Unlocks Clarke Hyper-stationarity in Bilevel Optimization", "authors": [ "He Chen", "Jiajin Li", "Anthony Man-Cho So" ], "affiliation": "", "summary": "Solving bilevel optimization (BLO) problems to global optimality is generally intractable", "links": { "openreview": "https://openreview.net/forum?id=QBnfYm6Naa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-417", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Computational Advantage of Depth in Learning High-Dimensional Hierarchical Targets", "authors": [ "Yatin Dandi", "Luca Pesce", "Lenka Zdeborova", "Florent Krzakala" ], "affiliation": "", "summary": "Understanding the advantages of deep neural networks trained by gradient descent (GD) compared to shallow models remains an open theoretical challenge", "links": { "openreview": "https://openreview.net/forum?id=5JcDVsV8pf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-418", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Thoughts Are All Over the Place: On the Underthinking of Long Reasoning Models", "authors": [ "Yue Wang", "Qiuzhi Liu", "Jiahao Xu", "Tian Liang", "Xingyu Chen", "Zhiwei He", "Linfeng Song", "Dian Yu", "Juntao Li", "Zhuosheng Zhang", "Rui Wang", "Zhaopeng Tu", "Haitao Mi", "Dong Yu" ], "affiliation": "", "summary": "Long reasoning models (LRMs) such as OpenAI's o1 and DeepSeek's R1 have demonstrated remarkable abilities in complex reasoning tasks by scaling test-time compute and exhibiting human-like deep thinkin", "links": { "openreview": "https://openreview.net/forum?id=WcUo7Z2Jnh", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-419", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OPTFM: A Scalable Multi-View Graph Transformer for Hierarchical Pre-Training in Combinatorial Optimization", "authors": [ "Hao Yuan", "Wenli Ouyang", "Changwen Zhang", "Congrui Li", "Yong Sun" ], "affiliation": "", "summary": "Foundation Models (FMs) have demonstrated remarkable success in fields like computer vision and natural language processing, yet their application to combinatorial optimization remains underexplored", "links": { "openreview": "https://openreview.net/forum?id=24tuzE5KZc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-420", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ESCA: Contextualizing Embodied Agents via Scene-Graph Generation", "authors": [ "Jiani Huang", "Amish Sethi", "Matthew Kuo", "Mayank Keoliya", "Neelay Velingker", "JungHo Jung", "Ser-Nam Lim", "Ziyang Li", "Mayur Naik" ], "affiliation": "", "summary": "Multi-modal large language models (MLLMs) are making rapid progress toward general-purpose embodied agents", "links": { "openreview": "https://openreview.net/forum?id=cjjPn1EIwq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-421", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Regret Bounds for Adversarial Contextual Bandits with General Function Approximation and Delayed Feedback", "authors": [ "Orin Levy", "Liad Erez", "Alon Cohen", "Yishay Mansour" ], "affiliation": "", "summary": "We present regret minimization algorithms for the contextual multi-armed bandit (CMAB) problem over $K$ actions in the presence of delayed feedback, a scenario where loss observations arrive with dela", "links": { "openreview": "https://openreview.net/forum?id=McPNQEDEZE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-422", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Multi-Table Learning: A Novel Paradigm for Complementarity Quantification and Integration", "authors": [ "Junyu Zhang", "Lizhong Ding", "MinghongZhang", "Ye Yuan", "Xingcan Li", "Pengqi Li", "Tihang Xi", "Guoren Wang", "Changsheng Li" ], "affiliation": "", "summary": "Multi-table data integrate various entities and attributes, with potential interconnections between them", "links": { "openreview": "https://openreview.net/forum?id=KlzLtQV64O", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-423", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "QFFT, Question-Free Fine-Tuning for Adaptive Reasoning", "authors": [ "Wanlong Liu", "Junxiao Xu", "Fei Yu", "Yukang Lin", "Ke Ji", "Wenyu Chen", "Lifeng Shang", "Yasheng Wang", "Yan Xu", "Benyou Wang" ], "affiliation": "", "summary": "Recent advancements in Long Chain-of-Thought (CoT) reasoning models have improved performance on complex tasks, but they suffer from overthinking, which generates redundant reasoning steps, especially", "links": { "openreview": "https://openreview.net/forum?id=CrBWOjZoKc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-424", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Option-aware Temporally Abstracted Value for Offline Goal-Conditioned Reinforcement Learning", "authors": [ "Hongjoon Ahn", "Heewoong Choi", "Jisu Han", "Taesup Moon" ], "affiliation": "", "summary": "Offline goal-conditioned reinforcement learning (GCRL) offers a practical learning paradigm in which goal-reaching policies are trained from abundant state–action trajectory datasets without additiona", "links": { "openreview": "https://openreview.net/forum?id=gfXBNBKx02", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-425", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Functional Scaling Laws in Kernel Regression: Loss Dynamics and Learning Rate Schedules", "authors": [ "Binghui Li", "Fengling Chen", "Zixun Huang", "Lean Wang", "Lei Wu" ], "affiliation": "", "summary": "Scaling laws have emerged as a unifying lens for understanding and guiding the training of large language models (LLMs)", "links": { "openreview": "https://openreview.net/forum?id=dpllevHMbc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-426", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GSRF: Complex-Valued 3D Gaussian Splatting for Efficient Radio-Frequency Data Synthesis", "authors": [ "Kang Yang", "Gaofeng Dong", "Sijie Ji", "Wan Du", "Mani Srivastava" ], "affiliation": "", "summary": "Synthesizing radio-frequency (RF) data given the transmitter and receiver positions, e", "links": { "openreview": "https://openreview.net/forum?id=E3oNDQ8e9r", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-427", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Sharp Gaussian approximations for Decentralized Federated Learning", "authors": [ "Soham Bonnerjee", "Sayar Karmakar", "Wei Biao Wu" ], "affiliation": "", "summary": "Federated Learning has gained traction in privacy-sensitive collaborative environments, with local SGD emerging as a key optimization method in decentralized settings", "links": { "openreview": "https://openreview.net/forum?id=b7waOsMnq8", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-428", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DAPO : Improving Multi-Step Reasoning Abilities of Large Language Models with Direct Advantage-Based Policy Optimization", "authors": [ "Jiacai Liu", "Chaojie Wang", "Chris Yuhao Liu", "Liang Zeng", "Rui Yan", "Yiwen Sun", "Yang Liu" ], "affiliation": "", "summary": "The role of reinforcement learning (RL) in enhancing the reasoning of large language models (LLMs) is becoming increasingly significant", "links": { "openreview": "https://openreview.net/forum?id=77eEDRhPkQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-429", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MGUP: A Momentum-Gradient Alignment Update Policy for Stochastic Optimization", "authors": [ "Da Chang", "Ganzhao Yuan" ], "affiliation": "", "summary": "Efficient optimization is essential for training large language models", "links": { "openreview": "https://openreview.net/forum?id=TDFSKAspoQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-430", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Unveiling the Power of Multiple Gossip Steps: A Stability-Based Generalization Analysis in Decentralized Training", "authors": [ "Qinglun Li", "Yingqi Liu", "Miao Zhang", "Xiaochun Cao", "Quanjun Yin", "Li Shen" ], "affiliation": "", "summary": "Decentralized training removes the centralized server, making it a communication-efficient approach that can significantly improve training efficiency, but it often suffers from degraded performance c", "links": { "openreview": "https://openreview.net/forum?id=V8dGVO5Xpg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-431", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Enhancing Contrastive Learning with Variable Similarity", "authors": [ "Haowen Cui", "Shuo Chen", "Jun Li", "Jian Yang" ], "affiliation": "", "summary": "Contrastive learning has achieved remarkable success in self-supervised learning by pretraining a generalizable feature representation based on the augmentation invariance", "links": { "openreview": "https://openreview.net/forum?id=oUf6lhK52B", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-432", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Orient Anything V2: Unifying Orientation and Rotation Understanding", "authors": [ "Zehan Wang", "Ziang Zhang", "Jiayang Xu", "Jialei Wang", "Tianyu Pang", "Chao Du", "Hengshuang Zhao", "Zhou Zhao" ], "affiliation": "", "summary": "This work presents Orient Anything V2, an enhanced foundation model for unified understanding of object 3D orientation and rotation from single or paired images", "links": { "openreview": "https://openreview.net/forum?id=n3armuTFit", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-433", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Closer Look at Graph Transformers: Cross-Aggregation and Beyond", "authors": [ "Jiaming Zhuo", "Ziyi Ma", "Yintong Lu", "Yuwei Liu", "Kun Fu", "Di Jin", "Chuan Wang", "Wenning Wu", "Zhen Wang", "Xiaochun Cao", "Liang Yang" ], "affiliation": "", "summary": "Graph Transformers (GTs), which effectively capture long-range dependencies and structural biases simultaneously, have recently emerged as promising alternatives to traditional Graph Neural Networks (", "links": { "openreview": "https://openreview.net/forum?id=7FhWZFoVem", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-434", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution", "authors": [ "Zhanyi Sun", "Shuran Song" ], "affiliation": "", "summary": "Visuomotor policies trained via behavior cloning are vulnerable to covariate shift, where small deviations from expert trajectories can compound into failure", "links": { "openreview": "https://openreview.net/forum?id=FUd016XD4d", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-435", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models", "authors": [ "Tyler A. Chang", "Ben Bergen" ], "affiliation": "", "summary": "In Transformer language models, activation vectors transform from current token embeddings to next token predictions as they pass through the model", "links": { "openreview": "https://openreview.net/forum?id=0TD3eO46gk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-436", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Minimax-Optimal Univariate Function Selection in Sparse Additive Models: Rates, Adaptation, and the Estimation-Selection Gap", "authors": [ "Shixiang Liu" ], "affiliation": "", "summary": "The sparse additive model (SpAM) offers a trade-off between interpretability and flexibility, and hence is a powerful model for high-dimensional research", "links": { "openreview": "https://openreview.net/forum?id=QudbVyFaTu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-437", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Continuous Thought Machines", "authors": [ "Luke Nicholas Darlow", "Ciaran Regan", "Sebastian Risi", "Jeffrey Seely", "Llion Jones" ], "affiliation": "", "summary": "Biological brains demonstrate complex neural activity, where neural dynamics are critical to how brains process information", "links": { "openreview": "https://openreview.net/forum?id=y0wDflmpLk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-438", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Virus Infection Attack on LLMs: Your Poisoning Can Spread \"VIA\" Synthetic Data", "authors": [ "Zi Liang", "Qingqing Ye", "Xuan Liu", "Yanyun Wang", "Jianliang Xu", "Haibo Hu" ], "affiliation": "", "summary": "Synthetic data refers to artificial samples generated by models", "links": { "openreview": "https://openreview.net/forum?id=QYJt0pX0zJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-439", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Robust Neural Rendering in the Wild with Asymmetric Dual 3D Gaussian Splatting", "authors": [ "Chengqi Li", "Zhihao Shi", "Yangdi Lu", "Wenbo He", "Xiangyu Xu" ], "affiliation": "", "summary": "3D reconstruction from in-the-wild images remains a challenging task due to inconsistent lighting conditions and transient distractors", "links": { "openreview": "https://openreview.net/forum?id=jPaM3AiFLq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-440", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Frame Context Packing and Drift Prevention in Next-Frame-Prediction Video Diffusion Models", "authors": [ "Lvmin Zhang", "Shengqu Cai", "Muyang Li", "Gordon Wetzstein", "Maneesh Agrawala" ], "affiliation": "", "summary": "We present a neural network structure, FramePack, to train next-frame (or next-frame-section) prediction models for video generation", "links": { "openreview": "https://openreview.net/forum?id=J8JCF64aEn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-441", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CoLT: The conditional localization test for assessing the accuracy of neural posterior estimates", "authors": [ "Tianyu Chen", "Vansh Bansal", "James G. Scott" ], "affiliation": "", "summary": "We consider the problem of validating whether a neural posterior estimate $q(\\theta \\mid x)$ is an accurate approximation to the true, unknown true posterior $p(\\theta \\mid x)$", "links": { "openreview": "https://openreview.net/forum?id=761hggw1Wx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-442", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Predictive Preference Learning from Human Interventions", "authors": [ "Haoyuan Cai", "Zhenghao Peng", "Bolei Zhou" ], "affiliation": "", "summary": "Learning from human involvement aims to incorporate the human subject to monitor and correct agent behavior errors", "links": { "openreview": "https://openreview.net/forum?id=ErEaq1UNaQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-443", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transformers for Mixed-type Event Sequences", "authors": [ "Felix Draxler", "Yang Meng", "Kai Nelson", "Lukas Laskowski", "Yibo Yang", "Theofanis Karaletsos", "Stephan Mandt" ], "affiliation": "", "summary": "Event sequences appear widely in domains such as medicine, finance, and remote sensing, yet modeling them is challenging due to their heterogeneity: sequences often contain multiple event types with d", "links": { "openreview": "https://openreview.net/forum?id=MtwsRjPZhf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-444", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ROGR: Relightable 3D Objects using Generative Relighting", "authors": [ "Jiapeng Tang", "Matthew Jacob Levine", "Dor Verbin", "Stephan J. Garbin", "Matthias Nießner", "Ricardo Martin Brualla", "Pratul P. Srinivasan", "Philipp Henzler" ], "affiliation": "", "summary": "We introduce ROGR, a novel approach that reconstructs a relightable 3D model of an object captured from multiple views, driven by a generative relighting model that simulates the effects of placing th", "links": { "openreview": "https://openreview.net/forum?id=VRwcEVRcC9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-445", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Hamiltonian Descent Algorithms for Optimization: Accelerated Rates via Randomized Integration Time", "authors": [ "Qiang Fu", "Andre Wibisono" ], "affiliation": "", "summary": "We study the Hamiltonian flow for optimization (HF-opt), which simulates the Hamiltonian dynamics for some integration time and resets the velocity to $0$ to decrease the objective function; this is t", "links": { "openreview": "https://openreview.net/forum?id=nW6SMcfDq3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-446", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Cost-aware LLM-based Online Dataset Annotation", "authors": [ "Eray Can Elumar", "Cem Tekin", "Osman Yagan" ], "affiliation": "", "summary": "Recent advances in large language models (LLMs) have enabled automated dataset labeling with minimal human supervision", "links": { "openreview": "https://openreview.net/forum?id=3AdTRYA2uJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-447", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scaling Unlocks Broader Generation and Deeper Functional Understanding of Proteins", "authors": [ "Aadyot Bhatnagar", "Sarthak Jain", "Joel Beazer", "Samuel C. Curran", "Alexander M. Hoffnagle", "Kyle Shan Ching", "Michael Martyn", "Stephen Nayfach", "Jeffrey A. Ruffolo", "Ali Madani" ], "affiliation": "", "summary": "Generative protein language models (PLMs) are powerful tools for designing proteins purpose-built to solve problems in medicine, agriculture, and industrial processes", "links": { "openreview": "https://openreview.net/forum?id=yvGL2HP7pU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-448", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "STARFlow: Scaling Latent Normalizing Flows for High-resolution Image Synthesis", "authors": [ "Jiatao Gu", "Tianrong Chen", "David Berthelot", "Huangjie Zheng", "Yuyang Wang", "Ruixiang ZHANG", "Laurent Dinh", "Miguel Ángel Bautista", "Joshua M. Susskind", "Shuangfei Zhai" ], "affiliation": "", "summary": "We present STARFlow, a scalable generative model based on normalizing flows that achieves strong performance on high-resolution image synthesis", "links": { "openreview": "https://openreview.net/forum?id=3YguS2rxdk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-449", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Flow Equivariant Recurrent Neural Networks", "authors": [ "T. Anderson Keller" ], "affiliation": "", "summary": "Data arrives at our senses as a continuous stream, smoothly transforming from one instant to the next", "links": { "openreview": "https://openreview.net/forum?id=N1KPOlcN6P", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-450", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AdaReasoner: Adaptive Reasoning Enables More Flexible Thinking", "authors": [ "Xiangqi Wang", "Yue Huang", "Yanbo Wang", "Xiaonan Luo", "Kehan Guo", "Yujun Zhou", "Xiangliang Zhang" ], "affiliation": "", "summary": "LLMs often need effective configurations, like temperature and reasoning steps, to handle tasks requiring sophisticated reasoning and problem-solving, ranging from joke generation to mathematical reas", "links": { "openreview": "https://openreview.net/forum?id=VjjJlJ5qik", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-451", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LABridge: Text–Image Latent Alignment Framework via Mean-Conditioned OU Process", "authors": [ "Huiyang Shao", "Xin Xia", "Yuxi Ren", "XING WANG", "Xuefeng Xiao" ], "affiliation": "", "summary": "Diffusion models have emerged as state‑of‑the‑art in image synthesis", "links": { "openreview": "https://openreview.net/forum?id=CABcYH1wKM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-452", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Achieving $\\tilde{\\mathcal{O}}(1/N)$ Optimality Gap in Restless Bandits through Gaussian Approximation", "authors": [ "Chen YAN", "Weina Wang", "Lei Ying" ], "affiliation": "", "summary": "We study the finite-horizon Restless Multi-Armed Bandit (RMAB) problem with $N$ homogeneous arms", "links": { "openreview": "https://openreview.net/forum?id=TQNlIQIrcK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-453", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Surprising Effectiveness of Large Learning Rates under Standard Width Scaling", "authors": [ "Moritz Haas", "Sebastian Bordt", "Ulrike von Luxburg", "Leena Chennuru Vankadara" ], "affiliation": "", "summary": "Scaling limits, such as infinite-width limits, serve as promising theoretical tools to study large-scale models", "links": { "openreview": "https://openreview.net/forum?id=hTxnm6H93P", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-454", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Flash Invariant Point Attention", "authors": [ "Andrew Liu", "Axel Elaldi", "Nicholas T Franklin", "Nathan Russell", "Gurinder S. Atwal", "Yih-En Andrew Ban", "Olivia Viessmann" ], "affiliation": "", "summary": "Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models", "links": { "openreview": "https://openreview.net/forum?id=gKsG5qR3Bt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-455", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SmallKV: Small Model Assisted Compensation of KV Cache Compression for Efficient LLM Inference", "authors": [ "Yi Zhao", "Yajuan Peng", "Nguyen Cam-Tu", "Zuchao Li", "Wang Xiaoliang", "hai zhao", "Xiaoming Fu" ], "affiliation": "", "summary": "KV cache eviction has emerged as an effective solution to alleviate resource constraints faced by LLMs in long-context scenarios", "links": { "openreview": "https://openreview.net/forum?id=0BVrpXMr5Y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-456", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs", "authors": [ "Jan Hagnberger", "Daniel Musekamp", "Mathias Niepert" ], "affiliation": "", "summary": "Solving time-dependent Partial Differential Equations (PDEs) using a densely discretized spatial domain is a fundamental problem in various scientific and engineering disciplines, including modeling c", "links": { "openreview": "https://openreview.net/forum?id=0r4yzkvt9j", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-457", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Empirical Power of Goodness-of-Fit Tests in Watermark Detection", "authors": [ "Weiqing He", "Xiang Li", "Tianqi Shang", "Li Shen", "Weijie J Su", "Qi Long" ], "affiliation": "", "summary": "Large language models (LLMs) raise concerns about content authenticity and integrity because they can generate human-like text at scale", "links": { "openreview": "https://openreview.net/forum?id=YES7VDXPV8", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-458", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Balancing Multimodal Training Through Game-Theoretic Regularization", "authors": [ "Konstantinos Kontras", "Thomas Strypsteen", "Christos Chatzichristos", "Paul Pu Liang", "Matthew B. Blaschko", "Maarten De Vos" ], "affiliation": "", "summary": "Multimodal learning holds the promise for richer information extraction by capturing dependencies across data sources", "links": { "openreview": "https://openreview.net/forum?id=auiURbhoYx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-459", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Bipolar Self-attention for Spiking Transformers", "authors": [ "Shuai Wang", "Malu Zhang", "Jingya Wang", "Dehao Zhang", "Yimeng Shan", "Jieyuan Zhang", "Yichen Xiao", "Honglin Cao", "Haonan Zhang", "Zeyu Ma", "Yang Yang", "Haizhou Li" ], "affiliation": "", "summary": "Harnessing the event-driven characteristic, Spiking Neural Networks (SNNs) present a promising avenue toward energy-efficient Transformer architectures", "links": { "openreview": "https://openreview.net/forum?id=nG45z7lJ7D", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-460", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Online Functional Tensor Decomposition via Continual Learning for Streaming Data Completion", "authors": [ "Xi Zhang", "Yanyi Li", "Yisi Luo", "Qi Xie", "Deyu Meng" ], "affiliation": "", "summary": "Online tensor decompositions are powerful and proven techniques that address the challenges in processing high-velocity streaming tensor data, such as traffic flow and weather system", "links": { "openreview": "https://openreview.net/forum?id=RPuTB28HsK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-461", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TrajMamba: An Efficient and Semantic-rich Vehicle Trajectory Pre-training Model", "authors": [ "Yichen Liu", "Yan Lin", "Shengnan Guo", "Zeyu Zhou", "Youfang Lin", "Huaiyu Wan" ], "affiliation": "", "summary": "Vehicle GPS trajectories record how vehicles move over time, storing valuable travel semantics, including movement patterns and travel purposes", "links": { "openreview": "https://openreview.net/forum?id=pRYGjhirkY", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-462", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Tensor Product Attention Is All You Need", "authors": [ "Yifan Zhang", "Yifeng Liu", "Huizhuo Yuan", "Zhen Qin", "Yang Yuan", "Quanquan Gu", "Andrew C Yao" ], "affiliation": "", "summary": "Scaling language models to handle longer input sequences typically necessitates large key-value (KV) caches, resulting in substantial memory overhead during inference", "links": { "openreview": "https://openreview.net/forum?id=ECTxVRFhUa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-463", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TreeSynth: Synthesizing Diverse Data from Scratch via Tree-Guided Subspace Partitioning", "authors": [ "Sheng Wang", "Pengan CHEN", "Jingqi Zhou", "Qintong Li", "Jingwei Dong", "Jiahui Gao", "Boyang XUE", "Jiyue Jiang", "Lingpeng Kong", "Chuan Wu" ], "affiliation": "", "summary": "Model customization necessitates high-quality and diverse datasets, but acquiring such data remains time-consuming and labor-intensive", "links": { "openreview": "https://openreview.net/forum?id=wmweEDugTZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-464", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DeCaFlow: A deconfounding causal generative model", "authors": [ "Alejandro Almodóvar", "Adrián Javaloy", "Juan Parras", "Santiago Zazo", "Isabel Valera" ], "affiliation": "", "summary": "We introduce DeCaFlow, a deconfounding causal generative model", "links": { "openreview": "https://openreview.net/forum?id=mOYGK7Hw9Y", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-465", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transfer Learning for Benign Overfitting in High-Dimensional Linear Regression", "authors": [ "Yeichan Kim", "Ilmun Kim", "Seyoung Park" ], "affiliation": "", "summary": "Transfer learning is a key component of modern machine learning, enhancing the performance of target tasks by leveraging diverse data sources", "links": { "openreview": "https://openreview.net/forum?id=arXNS7T90z", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-466", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LLM Meeting Decision Trees on Tabular Data", "authors": [ "Hangting Ye", "Jinmeng Li", "He Zhao", "Dandan Guo", "Yi Chang" ], "affiliation": "", "summary": "Tabular data have been playing a vital role in diverse real-world fields, including healthcare, finance, etc", "links": { "openreview": "https://openreview.net/forum?id=SRDF3RV0KP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-467", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AgentBreeder: Mitigating the AI Safety Risks of Multi-Agent Scaffolds via Self-Improvement", "authors": [ "J Rosser", "Jakob Nicolaus Foerster" ], "affiliation": "", "summary": "Scaffolding Large Language Models (LLMs) into multi-agent systems often improves performance on complex tasks, but the safety impact of such scaffolds has not been thoroughly explored", "links": { "openreview": "https://openreview.net/forum?id=mlU9KqdZUS", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-468", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learnable Burst-Encodable Time-of-Flight Imaging for High-Fidelity Long-Distance Depth Sensing", "authors": [ "Manchao Bao", "Shengjiang Fang", "Tao Yue", "Xuemei Hu" ], "affiliation": "", "summary": "Long-distance depth imaging holds great promise for applications such as autonomous driving and robotics", "links": { "openreview": "https://openreview.net/forum?id=zL4ifL17bU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-469", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MigGPT: Harnessing Large Language Models for Automated Migration of Out-of-Tree Linux Kernel Patches Across Versions", "authors": [ "Pucheng Dang", "Di Huang", "Dong Li", "Kang Chen", "Yuanbo Wen", "Qi Guo", "Xing Hu" ], "affiliation": "", "summary": "Out-of-tree kernel patches are essential for adapting the Linux kernel to new hardware or enabling specific functionalities", "links": { "openreview": "https://openreview.net/forum?id=7Z25QbOv4a", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-470", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FP4 All the Way: Fully Quantized Training of Large Language Models", "authors": [ "Brian Chmiel", "Maxim Fishman", "Ron Banner", "Daniel Soudry" ], "affiliation": "", "summary": "We demonstrate, for the first time, fully quantized training (FQT) of large language models (LLMs) using predominantly 4-bit floating-point (FP4) precision for weights, activations, and gradients on d", "links": { "openreview": "https://openreview.net/forum?id=kuzye4EPLR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-471", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Efficient Knowledge Transfer in Federated Recommendation for Joint Venture Ecosystem", "authors": [ "Yichen Li", "Yijing Shan", "YI LIU", "Haozhao Wang", "Cheng Wang", "wangshi.ww", "Yi Wang", "Ruixuan Li" ], "affiliation": "", "summary": "The current Federated Recommendation System (FedRS) focuses on personalized recommendation services and assumes clients are personalized IoT devices (e", "links": { "openreview": "https://openreview.net/forum?id=ZWOe1kkufx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-472", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning", "authors": [ "Haozhe Wang", "Chao Qu", "Zuming Huang", "Wei Chu", "Fangzhen Lin", "Wenhu Chen" ], "affiliation": "", "summary": "Recently, slow-thinking systems like GPT-o1 and DeepSeek-R1 have demonstrated great potential in solving challenging problems through explicit reflection", "links": { "openreview": "https://openreview.net/forum?id=4oYxzssbVg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-473", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Spectral Graph Neural Networks are Incomplete on Graphs with a Simple Spectrum", "authors": [ "Snir Hordan", "Maya Bechler-Speicher", "Gur Lifshitz", "Nadav Dym" ], "affiliation": "", "summary": "Spectral features are widely incorporated within Graph Neural Networks (GNNs) to improve their expressive power, or their ability to distinguish among non-isomorphic graphs", "links": { "openreview": "https://openreview.net/forum?id=3dnG7LcKxT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-474", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ShapeLLM-Omni: A Native Multimodal LLM for 3D Generation and Understanding", "authors": [ "Junliang Ye", "Zhengyi Wang", "Ruowen Zhao", "Shenghao Xie", "Jun Zhu" ], "affiliation": "", "summary": "Recently, the powerful text-to-image capabilities of GPT-4o have led to growing appreciation for native multimodal large language models", "links": { "openreview": "https://openreview.net/forum?id=muWdWcMvpW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-475", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Implicit Bias of Structured State Space Models Can Be Poisoned With Clean Labels", "authors": [ "Yonatan Slutzky", "Yotam Alexander", "Noam Razin", "Nadav Cohen" ], "affiliation": "", "summary": "Neural networks are powered by an implicit bias: a tendency of gradient descent to fit training data in a way that generalizes to unseen data", "links": { "openreview": "https://openreview.net/forum?id=3UaItHVjyE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-476", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Multiverse: Your Language Models Secretly Decide How to Parallelize and Merge Generation", "authors": [ "Xinyu Yang", "Yuwei An", "Hongyi Liu", "Tianqi Chen", "Beidi Chen" ], "affiliation": "", "summary": "Autoregressive Large Language Models (AR-LLMs) frequently exhibit implicit parallelism in sequential generation", "links": { "openreview": "https://openreview.net/forum?id=r9YDEErKXU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-477", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "BevSplat: Resolving Height Ambiguity via Feature-Based Gaussian Primitives for Weakly-Supervised Cross-View Localization", "authors": [ "Qiwei Wang", "Wu Shaoxun", "Yujiao Shi" ], "affiliation": "", "summary": "This paper addresses the problem of weakly supervised cross-view localization, where the goal is to estimate the pose of a ground camera relative to a satellite image with noisy ground truth annotatio", "links": { "openreview": "https://openreview.net/forum?id=Ig5mtZ8etr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-478", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Conflict-Aware Knowledge Editing in the Wild: Semantic-Augmented Graph Representation for Unstructured Text", "authors": [ "Zhange Zhang", "Zhicheng Geng", "Yuqing Ma", "Tianbo Wang", "Kai Lv", "Xianglong Liu" ], "affiliation": "", "summary": "Large Language Models (LLMs) have demonstrated broad applications but suffer from issues like hallucinations, erroneous outputs and outdated knowledge", "links": { "openreview": "https://openreview.net/forum?id=wHm5J9uanV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-479", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Vision Transformers with Self-Distilled Registers", "authors": [ "Zipeng Yan", "Yinjie Chen", "Chong Zhou", "Bo Dai", "Andrew Luo" ], "affiliation": "", "summary": "Vision Transformers (ViTs) have emerged as the dominant architecture for visual processing tasks, demonstrating excellent scalability with increased training data and model size", "links": { "openreview": "https://openreview.net/forum?id=VsDsRqaMJv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-480", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "EraseFlow: Learning Concept Erasure Policies via GFlowNet-Driven Alignment", "authors": [ "Naga Sai Abhiram kusumba", "Maitreya Patel", "Kyle Min", "Changhoon Kim", "Chitta Baral", "Yezhou Yang" ], "affiliation": "", "summary": "Erasing harmful or proprietary concepts from powerful text‑to‑image generators is an emerging safety requirement, yet current ``concept erasure'' techniques either collapse image quality, rely on brit", "links": { "openreview": "https://openreview.net/forum?id=igB289kbej", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-481", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Boundary-Value PDEs Meet Higher-Order Differential Topology-aware GNNs", "authors": [ "Yunfeng Liao", "Yangxin Wu", "Xiucheng Li" ], "affiliation": "", "summary": "Recent advances in graph neural network (GNN)-based neural operators have demonstrated significant progress in solving partial differential equations (PDEs) by effectively representing computational m", "links": { "openreview": "https://openreview.net/forum?id=PluDA8DEar", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-482", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos", "authors": [ "Zhen Xu", "Zhengqin Li", "Zhao Dong", "Xiaowei Zhou", "Richard Newcombe", "Zhaoyang Lv" ], "affiliation": "", "summary": "We propose 4DGT, a 4D Gaussian-based Transformer model for dynamic scene reconstruction, trained entirely on real-world monocular posed videos", "links": { "openreview": "https://openreview.net/forum?id=qMRFNxioPC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-483", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning", "authors": [ "Yinjie Wang", "Ling Yang", "Ye Tian", "Ke Shen", "Mengdi Wang" ], "affiliation": "", "summary": "Mathematical reasoning in large language models has been successfully incentivized through reinforcement learning with verifiable rewards, leading to improved one-shot precision", "links": { "openreview": "https://openreview.net/forum?id=wPdBe9zxNr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-484", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "How many measurements are enough? Bayesian recovery in inverse problems with general distributions", "authors": [ "Ben Adcock", "Nick Huang" ], "affiliation": "", "summary": "We study the sample complexity of Bayesian recovery for solving inverse problems with general prior, forward operator and noise distributions", "links": { "openreview": "https://openreview.net/forum?id=IIiRwgkZcm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-485", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Diffusion Generative Modeling on Lie Group Representations", "authors": [ "Marco Bertolini", "Tuan Le", "Djork-Arné Clevert" ], "affiliation": "", "summary": "We introduce a novel class of score-based diffusion processes that operate directly in the representation space of Lie groups", "links": { "openreview": "https://openreview.net/forum?id=Jom8tNYuQI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-486", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "EGGS: Exchangeable 2D/3D Gaussian Splatting for Geometry-Appearance Balanced Novel View Synthesis", "authors": [ "Yancheng Zhang", "Guangyu Sun", "Chen Chen" ], "affiliation": "", "summary": "Novel view synthesis (NVS) is crucial in computer vision and graphics, with wide applications in AR, VR, and autonomous driving", "links": { "openreview": "https://openreview.net/forum?id=25C8oC1pb2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-487", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CURE: Concept Unlearning via Orthogonal Representation Editing in Diffusion Models", "authors": [ "Shristi Das Biswas", "Arani Roy", "Kaushik Roy" ], "affiliation": "", "summary": "As Text-to-Image models continue to evolve, so does the risk of generating unsafe, copyrighted, or privacy-violating content", "links": { "openreview": "https://openreview.net/forum?id=zprMrpiLgT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-488", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Projection-based Lyapunov method for fully heterogeneous weakly-coupled MDPs", "authors": [ "XiangCheng Zhang", "Yige Hong", "Weina Wang" ], "affiliation": "", "summary": "Heterogeneity poses a fundamental challenge for many real-world large-scale decision-making problems but remains largely understudied", "links": { "openreview": "https://openreview.net/forum?id=alw3e1Qa7I", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-489", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Low-degree evidence for computational transition of recovery rate in stochastic block model", "authors": [ "Jingqiu Ding", "Yiding Hua", "Lucas Slot", "David Steurer" ], "affiliation": "", "summary": "We investigate implications of the (extended) low-degree conjecture (recently formalized in [moitra et al2023]) in the context of the symmetric stochastic block model", "links": { "openreview": "https://openreview.net/forum?id=fBNaGVMDD9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-490", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Nonlinear Laplacians: Tunable principal component analysis under directional prior information", "authors": [ "Yuxin Ma", "Dmitriy Kunisky" ], "affiliation": "", "summary": "We introduce a new family of algorithms for detecting and estimating a rank-one signal from a noisy observation under prior information about that signal's direction, focusing on examples where the si", "links": { "openreview": "https://openreview.net/forum?id=hLZrqDFugY", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-491", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Chain-of-Zoom: Extreme Super-Resolution via Scale Autoregression and Preference Alignment", "authors": [ "Bryan Sangwoo Kim", "Jeongsol Kim", "Jong Chul Ye" ], "affiliation": "", "summary": "Modern single-image super-resolution (SISR) models deliver photo-realistic results at the scale factors on which they are trained, but collapse when asked to magnify far beyond that regime", "links": { "openreview": "https://openreview.net/forum?id=I8S4ASqO5H", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-492", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Personalized Decision Modeling: Utility Optimization or Textualized-Symbolic Reasoning", "authors": [ "Yibo Zhao", "Yang Zhao", "Hongru Du", "Hao Frank Yang" ], "affiliation": "", "summary": "Decision-making models for individuals, particularly in high-stakes scenarios like vaccine uptake, often diverge from population optimal predictions", "links": { "openreview": "https://openreview.net/forum?id=RDt0crdC7N", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-493", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "What are you sinking? A geometric approach on attention sink", "authors": [ "Valeria Ruscio", "Umberto Nanni", "Fabrizio Silvestri" ], "affiliation": "", "summary": "Attention sink (AS) is a consistent pattern in transformer attention maps where certain tokens (often special tokens or positional anchors) disproportionately attract attention from other tokens", "links": { "openreview": "https://openreview.net/forum?id=OiC78C68sJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-494", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "KARMA: Leveraging Multi-Agent LLMs for Automated Knowledge Graph Enrichment", "authors": [ "Yuxing Lu", "Wei Wu", "Xukai Zhao", "Rui Peng", "Jinzhuo Wang" ], "affiliation": "", "summary": "Maintaining comprehensive and up-to-date knowledge graphs (KGs) is critical for modern AI systems, but manual curation struggles to scale with the rapid growth of scientific literature", "links": { "openreview": "https://openreview.net/forum?id=k0wyi4cOGy", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-495", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MetaMind: Modeling Human Social Thoughts with Metacognitive Multi-Agent Systems", "authors": [ "Xuanming Zhang", "Yuxuan Chen", "Samuel Yeh", "Sharon Li" ], "affiliation": "", "summary": "Human social interactions depend on the ability to infer others' unspoken intentions, emotions, and beliefs—a cognitive skill grounded in the psychological concept of Theory of Mind (ToM)", "links": { "openreview": "https://openreview.net/forum?id=rGMaZkn1ve", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-496", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Enhancing CLIP Robustness via Cross-Modality Alignment", "authors": [ "Xingyu Zhu", "Beier Zhu", "Shuo Wang", "Kesen Zhao", "Hanwang Zhang" ], "affiliation": "", "summary": "Vision-language models (VLMs) such as CLIP demonstrate strong generalization in zero-shot classification but remain highly vulnerable to adversarial perturbations", "links": { "openreview": "https://openreview.net/forum?id=t77EZLjvd5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-497", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Rethinking Entropy in Test-Time Adaptation: The Missing Piece from Energy Duality", "authors": [ "Mincheol Park", "Heeji Won", "Won Woo Ro", "Suhyun Kim" ], "affiliation": "", "summary": "Test-time adaptation (TTA) aims to preserve model performance under distribution shifts", "links": { "openreview": "https://openreview.net/forum?id=BKYFAutCDZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-498", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Optimal Construction of Unbiased Gradient Estimators for Zeroth-Order Optimization", "authors": [ "Shaocong Ma", "Heng Huang" ], "affiliation": "", "summary": "Zeroth-order optimization (ZOO) is an important framework for stochastic optimization when gradients are unavailable or expensive to compute", "links": { "openreview": "https://openreview.net/forum?id=rVT1GK60Nt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-499", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Gradient Variance Reveals Failure Modes in Flow-Based Generative Models", "authors": [ "Teodora Reu", "Sixtine Dromigny", "Michael M. Bronstein", "Francisco Vargas" ], "affiliation": "", "summary": "Rectified Flows learn ODE vector fields whose trajectories are straight between source and target distributions, enabling near one-step inference", "links": { "openreview": "https://openreview.net/forum?id=pVaqdFlUAO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-500", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception", "authors": [ "Runjian Chen", "Hyoungseob Park", "Bo Zhang", "Wenqi Shao", "Ping Luo", "Alex Wong" ], "affiliation": "", "summary": "Labeling LiDAR point clouds is notoriously time-and-energy-consuming, which spurs recent unsupervised 3D representation learning methods to alleviate the labeling burden in LiDAR perception via pretra", "links": { "openreview": "https://openreview.net/forum?id=AHccBzULR7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-501", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Boosting Generative Image Modeling via Joint Image-Feature Synthesis", "authors": [ "Theodoros Kouzelis", "Efstathios Karypidis", "Ioannis Kakogeorgiou", "Spyros Gidaris", "Nikos Komodakis" ], "affiliation": "", "summary": "Latent diffusion models (LDMs) dominate high-quality image generation, yet integrating representation learning with generative modeling remains a challenge", "links": { "openreview": "https://openreview.net/forum?id=i4qAfV04rZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-502", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Token is Worth over 1,000 Tokens: Efficient Knowledge Distillation through Low-Rank Clone", "authors": [ "Jitai Hao", "Qiang Huang", "Hao Liu", "Xinyan Xiao", "Zhaochun Ren", "Jun Yu" ], "affiliation": "", "summary": "Training high-performing Small Language Models (SLMs) remains computationally expensive, even with knowledge distillation and pruning from larger teacher models", "links": { "openreview": "https://openreview.net/forum?id=LVDRJE4xQ2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-503", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SparseMVC: Probing Cross-view Sparsity Variations for Multi-view Clustering", "authors": [ "Ruimeng Liu", "Xin Zou", "Chang Tang", "Xiao Zheng", "Xingchen Hu", "Kun Sun", "Xinwang Liu" ], "affiliation": "", "summary": "Existing multi-view clustering methods employ various strategies to address data-level sparsity and view-level dynamic fusion", "links": { "openreview": "https://openreview.net/forum?id=cvJvk6oYfC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-504", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Unbiased Prototype Consistency Learning for Multi-Modal and Multi-Task Object Re-Identification", "authors": [ "Zhongao Zhou", "Bin Yang", "Wenke Huang", "Jun Chen", "Mang Ye" ], "affiliation": "", "summary": "In object re-identification (ReID) task, both cross-modal and multi-modal retrieval methods have achieved notable progress", "links": { "openreview": "https://openreview.net/forum?id=WWa5x1WnEw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-505", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MJ-Video: Benchmarking and Rewarding Video Generation with Fine-Grained Video Preference", "authors": [ "Haibo Tong", "Zhaoyang Wang", "Zhaorun Chen", "Haonian Ji", "Shi Qiu", "Siwei Han", "Kexin Geng", "Zhongkai Xue", "Yiyang Zhou", "Peng Xia", "Mingyu Ding", "Rafael Rafailov", "Chelsea Finn", "Huaxiu Yao" ], "affiliation": "", "summary": "Recent advancements in video generation have significantly improved the ability to synthesize videos from text instructions", "links": { "openreview": "https://openreview.net/forum?id=56C0n6zSpC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-506", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fast Projection-Free Approach (without Optimization Oracle) for Optimization over Compact Convex Set", "authors": [ "Chenghao Liu", "Enming Liang", "Minghua Chen" ], "affiliation": "", "summary": "Projection-free first-order methods, e", "links": { "openreview": "https://openreview.net/forum?id=bP5cU0OYSn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-507", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search", "authors": [ "Yuichi Inoue", "Kou Misaki", "Yuki Imajuku", "So Kuroki", "Taishi Nakamura", "Takuya Akiba" ], "affiliation": "", "summary": "Recent advances demonstrate that increasing inference-time computation can significantly boost the reasoning capabilities of large language models (LLMs)", "links": { "openreview": "https://openreview.net/forum?id=jAsr5GHt3P", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-508", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RepoMaster: Autonomous Exploration and Understanding of GitHub Repositories for Complex Task Solving", "authors": [ "Huacan Wang", "Ziyi Ni", "Shuo Zhang", "Shuo Lu", "Sen Hu", "Ziyang He", "Chen Hu", "Jiaye Lin", "Yifu Guo", "Yuntao Du", "Pin Lyu" ], "affiliation": "", "summary": "The ultimate goal of code agents is to solve complex tasks autonomously", "links": { "openreview": "https://openreview.net/forum?id=aSfBbhUJAa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-509", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Hierarchical Shortest-Path Graph Kernel Network", "authors": [ "Jiaxin Wang", "Wenxuan Tu", "Jieren Cheng" ], "affiliation": "", "summary": "Graph kernels have emerged as a fundamental and widely adopted technique in graph machine learning", "links": { "openreview": "https://openreview.net/forum?id=721bDIvjen", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-510", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Differential Privacy on Fully Dynamic Streams", "authors": [ "Yuan Qiu", "Ke Yi" ], "affiliation": "", "summary": "A fundamental problem in differential privacy is to release privatized answers to a class of linear queries with small error", "links": { "openreview": "https://openreview.net/forum?id=piM21sPyVL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-511", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Shortcut Features as Top Eigenfunctions of NTK: A Linear Neural Network Case and More", "authors": [ "Jinwoo Lim", "Suhyun Kim", "Soo-Mook Moon" ], "affiliation": "", "summary": "One of the chronic problems of deep-learning models is shortcut learning", "links": { "openreview": "https://openreview.net/forum?id=TYroQXu6X0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-512", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SuffixDecoding: Extreme Speculative Decoding for Emerging AI Applications", "authors": [ "Gabriele Oliaro", "Zhihao Jia", "Daniel F Campos", "Aurick Qiao" ], "affiliation": "", "summary": "Speculative decoding is widely adopted to reduce latency in large language model (LLM) inference by leveraging smaller draft models capable of handling diverse user tasks", "links": { "openreview": "https://openreview.net/forum?id=uwL0vbeEVn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-513", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Efficient Prompt Compression with Evaluator Heads for Long-Context Transformer Inference", "authors": [ "Weizhi Fei", "Xueyan Niu", "XIE GUOQING", "Yingqing Liu", "Bo Bai", "Wei Han" ], "affiliation": "", "summary": "Although applications involving long-context inputs are crucial for the effective utilization of large language models (LLMs), they also result in increased computational costs and reduced performance", "links": { "openreview": "https://openreview.net/forum?id=yOs12gdsaL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-514", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting", "authors": [ "Kai He", "Ruofan Liang", "Jacob Munkberg", "Jon Hasselgren", "Nandita Vijaykumar", "Alexander Keller", "Sanja Fidler", "Igor Gilitschenski", "Zan Gojcic", "Zian Wang" ], "affiliation": "", "summary": "We address the challenge of relighting a single image or video, a task that demands precise scene intrinsic understanding and high-quality light transport synthesis", "links": { "openreview": "https://openreview.net/forum?id=1bO9wIdyKa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-515", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs", "authors": [ "Xiaomin Li", "Zhou Yu", "Zhiwei Zhang", "Xupeng Chen", "Ziji Zhang", "Yingying Zhuang", "Narayanan Sadagopan", "Anurag Beniwal" ], "affiliation": "", "summary": "Reasoning-enhanced large language models (RLLMs), whether explicitly trained for reasoning or prompted via chain-of-thought (CoT), have achieved state-of-the-art performance on many complex reasoning", "links": { "openreview": "https://openreview.net/forum?id=w5uUvxp81b", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-516", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On Agnostic PAC Learning in the Small Error Regime", "authors": [ "Julian Asilis", "Mikael Møller Høgsgaard", "Grigoris Velegkas" ], "affiliation": "", "summary": "Binary classification in the classic PAC model exhibits a curious phenomenon: Empirical Risk Minimization (ERM) learners are suboptimal in the realizable case yet optimal in the agnostic case", "links": { "openreview": "https://openreview.net/forum?id=6LOgOsIcXe", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-517", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Wavelet Canonical Coherence for Nonstationary Signals", "authors": [ "Haibo Wu", "Marina I. Knight", "Keiland W. Cooper", "Norbert J. Fortin", "Hernando Ombao" ], "affiliation": "", "summary": "Understanding the evolving dependence between two sets of multivariate signals is fundamental in neuroscience and other domains where sub-networks in a system interact dynamically over time", "links": { "openreview": "https://openreview.net/forum?id=uUIgxjWkCI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-518", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Language Modeling by Language Models", "authors": [ "Junyan Cheng", "Peter Clark", "Kyle Richardson" ], "affiliation": "", "summary": "*Can we leverage LLMs to model the process of discovering novel language model (LM) architectures?* Inspired by real research, we propose a multi-agent LLM approach that simulates the conventional sta", "links": { "openreview": "https://openreview.net/forum?id=VrCdsZBbIg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-519", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Plasticity as the Mirror of Empowerment", "authors": [ "David Abel", "Michael Bowling", "Andre Barreto", "Will Dabney", "Shi Dong", "Steven Stenberg Hansen", "Anna Harutyunyan", "Khimya Khetarpal", "Clare Lyle", "Razvan Pascanu", "Georgios Piliouras", "Doina Precup", "Jonathan Richens", "Mark Rowland", "Tom Schaul", "Satinder Singh" ], "affiliation": "", "summary": "Agents are minimally entities that are influenced by their past observations and act to influence future observations", "links": { "openreview": "https://openreview.net/forum?id=eOZFqyE9Ok", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-520", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PiKE: Adaptive Data Mixing for Large-Scale Multi-Task Learning Under Low Gradient Conflicts", "authors": [ "Zeman Li", "Yuan Deng", "Peilin Zhong", "Meisam Razaviyayn", "Vahab Mirrokni" ], "affiliation": "", "summary": "Modern foundation models are trained on diverse datasets to enhance generalization across tasks and domains", "links": { "openreview": "https://openreview.net/forum?id=xNJenVNmzL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-521", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Language Models can Self-Improve at State-Value Estimation for Better Search", "authors": [ "Ethan Mendes", "Alan Ritter" ], "affiliation": "", "summary": "Collecting ground-truth rewards or human demonstrations for multi-step reasoning tasks is often prohibitively expensive, especially in interactive domains such as web tasks", "links": { "openreview": "https://openreview.net/forum?id=W2874Arl4g", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-522", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scalable Cross-View Sample Alignment for Multi-View Clustering with View Structure Similarity", "authors": [ "Jun Wang", "Zhenglai Li", "Chang Tang", "Suyuan Liu", "Hao Yu", "Chuan Tang", "Miaomiao Li", "Xinwang Liu" ], "affiliation": "", "summary": "Most existing multi-view clustering methods aim to generate a consensus partition across all views, based on the assumption that all views share the same sample arrangement", "links": { "openreview": "https://openreview.net/forum?id=oysfr9yqUI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-523", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Multidimensional Bayesian Utility Maximization: Tight Approximations to Welfare", "authors": [ "Kira Goldner", "Taylor Lundy" ], "affiliation": "", "summary": "We initiate the study of multidimensional Bayesian utility maximization, focusing on the unit-demand setting where values are i", "links": { "openreview": "https://openreview.net/forum?id=SX6nL00JvM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-524", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Decomposing stimulus-specific sensory neural information via diffusion models", "authors": [ "Steeve Laquitaine", "Simone Azeglio", "Carlo Paris", "Ulisse Ferrari", "Matthew Chalk" ], "affiliation": "", "summary": "A central question in sensory neuroscience is how much, but also what information neurons transmit about the world", "links": { "openreview": "https://openreview.net/forum?id=Dt5vRmUjAv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-525", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Any-stepsize Gradient Descent for Separable Data under Fenchel–Young Losses", "authors": [ "Han Bao", "Shinsaku Sakaue", "Yuki Takezawa" ], "affiliation": "", "summary": "The gradient descent (GD) has been one of the most common optimizer in machine learning", "links": { "openreview": "https://openreview.net/forum?id=D6aCr4RRdt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-526", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation", "authors": [ "Toshinori Kitamura", "Arnob Ghosh", "Tadashi Kozuno", "Wataru Kumagai", "Kazumi Kasaura", "Kenta Hoshino", "Yohei Hosoe", "Yutaka Matsuo" ], "affiliation": "", "summary": "We study the reinforcement learning (RL) problem in a constrained Markov decision process (CMDP), where an agent explores the environment to maximize the expected cumulative reward while satisfying a", "links": { "openreview": "https://openreview.net/forum?id=rgoSyTCTkn", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-527", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "To Distill or Decide? Understanding the Algorithmic Trade-off in Partially Observable RL", "authors": [ "Yuda Song", "Dhruv Rohatgi", "Aarti Singh", "Drew Bagnell" ], "affiliation": "", "summary": "Partial observability is a notorious challenge in reinforcement learning (RL), due to the need to learn complex, history-dependent policies", "links": { "openreview": "https://openreview.net/forum?id=iEgaS6wbLa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-528", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A learnability analysis on neuro-symbolic learning", "authors": [ "Hao-Yuan He", "Ming Li" ], "affiliation": "", "summary": "This paper presents a comprehensive theoretical analysis of the learnability of neuro-symbolic (NeSy) tasks within hybrid systems", "links": { "openreview": "https://openreview.net/forum?id=FrdX7K4Gli", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-529", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Self-Perturbed Anomaly-Aware Graph Dynamics for Multivariate Time-Series Anomaly Detection", "authors": [ "Jinyu Cai", "Yuan Xie", "Glynnis Lim", "Yifang Yin", "Roger Zimmermann", "See-Kiong Ng" ], "affiliation": "", "summary": "Detecting anomalies in multivariate time-series data is an essential task across various domains, yet there are unresolved challenges such as (1) severe class imbalance between normal and anomalous da", "links": { "openreview": "https://openreview.net/forum?id=hJJnwcvE2M", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-530", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Restoring Pruned Large Language Models via Lost Component Compensation", "authors": [ "Zijian Feng", "Hanzhang Zhou", "Zixiao Zhu", "Tianjiao Li", "Chua Jia Jim Deryl", "Mak Lee Onn", "Gee Wah Ng", "Kezhi Mao" ], "affiliation": "", "summary": "Pruning is a widely used technique to reduce the size and inference cost of large language models (LLMs), but it often causes performance degradation", "links": { "openreview": "https://openreview.net/forum?id=cECo8tetzF", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-531", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Building Model/Prompt-Transferable Attackers against Large Vision-Language Models", "authors": [ "Xiaowen Cai", "Daizong Liu", "Xiaoye Qu", "Xiang Fang", "Jianfeng Dong", "Keke Tang", "Pan Zhou", "Lichao Sun", "Wei Hu" ], "affiliation": "", "summary": "Although Large Vision-Language Models (LVLMs) exhibit impressive multimodal capabilities, their vulnerability to adversarial examples has raised serious security concerns", "links": { "openreview": "https://openreview.net/forum?id=TyW1V1KukG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-532", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GaussianFusion: Gaussian-Based Multi-Sensor Fusion for End-to-End Autonomous Driving", "authors": [ "Shuai Liu", "Quanmin Liang", "Zefeng Li", "Boyang Li", "Kai Huang" ], "affiliation": "", "summary": "Multi-sensor fusion is crucial for improving the performance and robustness of end-to-end autonomous driving systems", "links": { "openreview": "https://openreview.net/forum?id=LBo4e6Y7Zg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-533", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SIU3R: Simultaneous Scene Understanding and 3D Reconstruction Beyond Feature Alignment", "authors": [ "Qi Xu", "Dongxu Wei", "Lingzhe Zhao", "Wenpu Li", "Zhangchi Huang", "Shunping Ji", "Peidong Liu" ], "affiliation": "", "summary": "Simultaneous understanding and 3D reconstruction plays an important role in developing end-to-end embodied intelligent systems", "links": { "openreview": "https://openreview.net/forum?id=GtImvTta8x", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-534", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "HYPERION: Fine-Grained Hypersphere Alignment for Robust Federated Graph Learning", "authors": [ "Guancheng Wan", "Xiaoran Shang", "Yuxin Wu", "Guibin Zhang", "Jinhe Bi", "Liangtao Zheng", "Xin Lin", "Yue Liu", "Yanbiao Ma", "Wenke Huang", "Bo Du" ], "affiliation": "", "summary": "Robust Federated Graph Learning (FGL) provides an effective decentralized framework for training Graph Neural Networks (GNNs) in noisy-label environments", "links": { "openreview": "https://openreview.net/forum?id=TZB6YT8Owr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-535", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Brain-Inspired fMRI-to-Text Decoding via Incremental and Wrap-Up Language Modeling", "authors": [ "Wentao Lu", "Dong Nie", "Pengcheng Xue", "Zheng Cui", "Piji Li", "Daoqiang Zhang", "Xuyun Wen" ], "affiliation": "", "summary": "Decoding natural language text from non-invasive brain signals, such as functional magnetic resonance imaging (fMRI), remains a central challenge in brain-computer interface research", "links": { "openreview": "https://openreview.net/forum?id=REIo9ZLSYo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-536", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decoding", "authors": [ "Yiming Wang", "Pei Zhang", "Siyuan Huang", "Baosong Yang", "Zhuosheng Zhang", "Fei Huang", "Rui Wang" ], "affiliation": "", "summary": "Test-time scaling enhances large language model performance by allocating additional compute resources during decoding", "links": { "openreview": "https://openreview.net/forum?id=BcKYVmh3yH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-537", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PoE-World: Compositional World Modeling with Products of Programmatic Experts", "authors": [ "Wasu Top Piriyakulkij", "Yichao Liang", "Hao Tang", "Adrian Weller", "Marta Kryven", "Kevin Ellis" ], "affiliation": "", "summary": "Learning how the world works is central to building AI agents that can adapt to complex environments", "links": { "openreview": "https://openreview.net/forum?id=obwRcksFZw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-538", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Uni-LoRA: One Vector is All You Need", "authors": [ "Kaiyang Li", "Shaobo Han", "Qing Su", "Wei Li", "Zhipeng Cai", "Shihao Ji" ], "affiliation": "", "summary": "Low-Rank Adaptation (LoRA) has become the de facto parameter-efficient fine-tuning (PEFT) method for large language models (LLMs) by constraining weight updates to low-rank matrices", "links": { "openreview": "https://openreview.net/forum?id=hzBqQZK2iV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-539", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Reinforcement Learning for Out-of-Distribution Reasoning in LLMs: An Empirical Study on Diagnosis-Related Group Coding", "authors": [ "Hanyin Wang", "Zhenbang Wu", "Gururaj J. Kolar", "Hariprasad Reddy Korsapati", "Brian Bartlett", "Bryan Hull", "Jimeng Sun" ], "affiliation": "", "summary": "Diagnosis-Related Group (DRG) codes are essential for hospital reimbursement and operations but require labor-intensive assignment", "links": { "openreview": "https://openreview.net/forum?id=0jvnfH0WYV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-540", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Stochastic Process Learning via Operator Flow Matching", "authors": [ "Yaozhong Shi", "Zachary E Ross", "Domniki Asimaki", "Kamyar Azizzadenesheli" ], "affiliation": "", "summary": "Expanding on neural operators, we propose a novel framework for stochastic process learning across arbitrary domains", "links": { "openreview": "https://openreview.net/forum?id=Quvnn2o17a", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-541", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Strategic Costs of Perceived Bias in Fair Selection", "authors": [ "L. Elisa Celis", "Lingxiao Huang", "Milind Sohoni", "Nisheeth K. Vishnoi" ], "affiliation": "", "summary": "Meritocratic systems, from admissions to hiring, aim to impartially reward skill and effort", "links": { "openreview": "https://openreview.net/forum?id=W8xcKoJcrl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-542", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Closer Look at Model Collapse: From a Generalization-to-Memorization Perspective", "authors": [ "Lianghe Shi", "Meng Wu", "Huijie Zhang", "Zekai Zhang", "Molei Tao", "Qing Qu" ], "affiliation": "", "summary": "The widespread use of diffusion models has led to an abundance of AI-generated data, raising concerns about model collapse---a phenomenon in which recursive iterations of training on synthetic data le", "links": { "openreview": "https://openreview.net/forum?id=6xCcjYa97j", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-543", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AdaSPEC: Selective Knowledge Distillation for Efficient Speculative Decoders", "authors": [ "Yuezhou Hu", "Jiaxin Guo", "Xinyu Feng", "Tuo Zhao" ], "affiliation": "", "summary": "Speculative Decoding (SD) accelerates large language model inference by employing a small draft model to generate predictions, which are then verified by a larger target model", "links": { "openreview": "https://openreview.net/forum?id=zNLlglSOwD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-544", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Depth-Width Tradeoffs for Transformers on Graph Tasks", "authors": [ "Gilad Yehudai", "Clayton Sanford", "Maya Bechler-Speicher", "Orr Fischer", "Ran Gilad-Bachrach", "Amir Globerson" ], "affiliation": "", "summary": "Transformers have revolutionized the field of machine learning", "links": { "openreview": "https://openreview.net/forum?id=A2pmNL7L1E", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-545", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Refinement Methods for Distributed Distribution Estimation under $\\ell^p$-Losses", "authors": [ "Deheng Yuan", "Tao Guo", "Zhongyi Huang" ], "affiliation": "", "summary": "Consider the communication-constrained estimation of discrete distributions under $\\ell^p$ losses, where each distributed terminal holds multiple independent samples and uses limited number of bits to", "links": { "openreview": "https://openreview.net/forum?id=0SYkQ50imt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-546", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Measuring and Guiding Monosemanticity", "authors": [ "Ruben Härle", "Felix Friedrich", "Manuel Brack", "Björn Deiseroth", "Stephan Waeldchen", "Patrick Schramowski", "Kristian Kersting" ], "affiliation": "", "summary": "There is growing interest in leveraging mechanistic interpretability and controllability to better understand and influence the internal dynamics of large language models (LLMs)", "links": { "openreview": "https://openreview.net/forum?id=REHjkmWdQL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-547", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Variational Transdimensional Inference", "authors": [ "Laurence Davies", "Dan MacKinlay", "Rafael Oliveira", "Scott A Sisson" ], "affiliation": "", "summary": "The expressiveness of flow-based models combined with stochastic variational inference (SVI) has expanded the application of optimization-based Bayesian inference to highly complex problems", "links": { "openreview": "https://openreview.net/forum?id=KqhMpsWiz2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-548", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scaling Laws For Scalable Oversight", "authors": [ "Joshua Engels", "David D. Baek", "Subhash Kantamneni", "Max Tegmark" ], "affiliation": "", "summary": "Scalable oversight, the process by which weaker AI systems supervise stronger ones, has been proposed as a key strategy to control future superintelligent systems", "links": { "openreview": "https://openreview.net/forum?id=u1j6RqH8nM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-549", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Near-Optimal Algorithm for Decentralized Convex-Concave Finite-Sum Minimax Optimization", "authors": [ "Hongxu Chen", "Ke Wei", "Haishan Ye", "Luo Luo" ], "affiliation": "", "summary": "In this paper, we study the distributed convex-concave finite-sum minimax optimization over the network, and a decentralized variance-reduced optimistic gradient method with stochastic mini-batch size", "links": { "openreview": "https://openreview.net/forum?id=raZEmZ48h4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-550", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learning with Calibration: Exploring Test-Time Computing of Spatio-Temporal Forecasting", "authors": [ "Wei Chen", "Yuxuan Liang" ], "affiliation": "", "summary": "Spatio-temporal forecasting is crucial in many domains, such as transportation, meteorology, and energy", "links": { "openreview": "https://openreview.net/forum?id=Zapn9l2LMY", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-551", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FPSAttention: Training-Aware FP8 and Sparsity Co-Design for Fast Video Diffusion", "authors": [ "Akide Liu", "Zeyu Zhang", "Zhexin Li", "Xuehai Bai", "Yuanjie Xing", "Yizeng Han", "Jiasheng Tang", "Jichao Wu", "Mingyang Yang", "Weihua Chen", "Jiahao He", "Yuanyu He", "Fan Wang", "Gholamreza Haffari", "Bohan Zhuang" ], "affiliation": "", "summary": "Diffusion generative models have become the standard for producing high-quality, coherent video content, yet their slow inference speeds and high computational demands hinder practical deployment", "links": { "openreview": "https://openreview.net/forum?id=T62TYoF8R3", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-552", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition", "authors": [ "Jongseo Lee", "Wooil Lee", "Gyeong-Moon Park", "Seong Tae Kim", "Jinwoo Choi" ], "affiliation": "", "summary": "Effective explanations of video action recognition models should disentangle how movements unfold over time from the surrounding spatial context", "links": { "openreview": "https://openreview.net/forum?id=paRLw86ONU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-553", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Physics-Driven Spatiotemporal Modeling for AI-Generated Video Detection", "authors": [ "Shuhai Zhang", "ZiHao Lian", "Jiahao Yang", "Daiyuan Li", "Guoxuan Pang", "Feng Liu", "Bo Han", "Shutao Li", "Mingkui Tan" ], "affiliation": "", "summary": "AI-generated videos have achieved near-perfect visual realism (e", "links": { "openreview": "https://openreview.net/forum?id=HiBoJLCyEo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-554", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Robust Graph Condensation via Classification Complexity Mitigation", "authors": [ "Jiayi Luo", "Qingyun Sun", "Beining Yang", "Haonan Yuan", "Xingcheng Fu", "Yanbiao Ma", "Jianxin Li", "Philip S. Yu" ], "affiliation": "", "summary": "Graph condensation (GC) has gained significant attention for its ability to synthesize smaller yet informative graphs", "links": { "openreview": "https://openreview.net/forum?id=vATe64ktAo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-555", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CausalPFN: Amortized Causal Effect Estimation via In-Context Learning", "authors": [ "Vahid Balazadeh", "Hamidreza Kamkari", "Valentin Thomas", "Junwei Ma", "Bingru Li", "Jesse C. Cresswell", "Rahul Krishnan" ], "affiliation": "", "summary": "Causal effect estimation from observational data is fundamental across various applications", "links": { "openreview": "https://openreview.net/forum?id=RblaNJGx8C", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-556", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "High-order Equivariant Flow Matching for Density Functional Theory Hamiltonian Prediction", "authors": [ "Seongsu Kim", "Nayoung Kim", "Dongwoo Kim", "Sungsoo Ahn" ], "affiliation": "", "summary": "Density functional theory (DFT) is a fundamental method for simulating quantum chemical properties, but it remains expensive due to the iterative self-consistent field (SCF) process required to solve", "links": { "openreview": "https://openreview.net/forum?id=iFIjNXb0Y5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-557", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "🎧MOSPA: Human Motion Generation Driven by Spatial Audio", "authors": [ "Shuyang Xu", "Zhiyang Dou", "Mingyi Shi", "Liang Pan", "Leo Ho", "Jingbo Wang", "Yuan Liu", "Cheng Lin", "Yuexin Ma", "Wenping Wang", "Taku Komura" ], "affiliation": "", "summary": "Enabling virtual humans to dynamically and realistically respond to diverse auditory stimuli remains a key challenge in character animation, demanding the integration of perceptual modeling and motion", "links": { "openreview": "https://openreview.net/forum?id=X2r9D46kvI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-558", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Absolute Zero: Reinforced Self-play Reasoning with Zero Data", "authors": [ "Andrew Zhao", "Yiran Wu", "Yang Yue", "Tong Wu", "Quentin Xu", "Yang Yue", "Matthieu Lin", "Shenzhi Wang", "Qingyun Wu", "Zilong Zheng", "Gao Huang" ], "affiliation": "", "summary": "Reinforcement learning with verifiable rewards (RLVR) has shown promise in enhancing the reasoning capabilities of large language models by learning directly from rule-based outcome rewards", "links": { "openreview": "https://openreview.net/forum?id=neZSGqhxDa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-559", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Graph–Smoothed Bayesian Black-Box Shift Estimator and Its Information Geometry", "authors": [ "Masanari Kimura" ], "affiliation": "", "summary": "Label shift adaptation aims to recover target class priors when the labelled source distribution $P$ and the unlabelled target distribution $Q$ share $P(X \\mid Y) = Q(X \\mid Y)$ but $P(Y) \\neq Q(Y)$", "links": { "openreview": "https://openreview.net/forum?id=Vws7eXQXsa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-560", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Vision Transformers Don't Need Trained Registers", "authors": [ "Nicholas Jiang", "Amil Dravid", "Alexei A Efros", "Yossi Gandelsman" ], "affiliation": "", "summary": "We investigate the mechanism underlying a previously identified phenomenon in Vision Transformers -- the emergence of high-norm tokens that lead to noisy attention maps (Darcet et al", "links": { "openreview": "https://openreview.net/forum?id=bA02DmQN5d", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-561", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Object-centric 3D Motion Field for Robot Learning from Human Videos", "authors": [ "Zhao-Heng Yin", "Sherry Yang", "Pieter Abbeel" ], "affiliation": "", "summary": "Learning robot control policies from human videos is a promising direction for scaling up robot learning", "links": { "openreview": "https://openreview.net/forum?id=kp9B9iQDIt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-562", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation", "authors": [ "Shuo Yang", "Haocheng Xi", "Yilong Zhao", "Muyang Li", "Jintao Zhang", "Han Cai", "Yujun Lin", "Xiuyu Li", "Chenfeng Xu", "Kelly Peng", "Jianfei Chen", "Song Han", "Kurt Keutzer", "Ion Stoica" ], "affiliation": "", "summary": "Diffusion Transformers (DiTs) are essential for video generation but suffer from significant latency due to the quadratic complexity of attention", "links": { "openreview": "https://openreview.net/forum?id=WPU17d1l7R", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-563", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Aggregation Hides Out-of-Distribution Generalization Failures from Spurious Correlations", "authors": [ "Olawale Elijah Salaudeen", "Haoran Zhang", "Kumail Alhamoud", "Sara Beery", "Marzyeh Ghassemi" ], "affiliation": "", "summary": "Benchmarks for out-of-distribution (OOD) generalization often reveal a strong positive correlation between in-distribution (ID) and OOD accuracy across models, a phenomenon known as “accuracy-on-the-l", "links": { "openreview": "https://openreview.net/forum?id=w97lDmoD0U", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-564", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mean-Field Sampling for Cooperative Multi-Agent Reinforcement Learning", "authors": [ "Emile Timothy Anand", "Ishani Karmarkar", "Guannan Qu" ], "affiliation": "", "summary": "Designing efficient algorithms for multi-agent reinforcement learning (MARL) is fundamentally challenging because the size of the joint state and action spaces grows exponentially in the number of age", "links": { "openreview": "https://openreview.net/forum?id=CTsdZ3j6dR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-565", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "EAG3R: Event-Augmented 3D Geometry Estimation for Dynamic and Extreme-Lighting Scenes", "authors": [ "Xiaoshan Wu", "Yifei Yu", "Xiaoyang Lyu", "Yi-Hua Huang", "Bo Wang", "Baoheng Zhang", "Zhongrui Wang", "XIAOJUAN QI" ], "affiliation": "", "summary": "Robust 3D geometry estimation from videos is critical for applications such as autonomous navigation, SLAM, and 3D scene reconstruction", "links": { "openreview": "https://openreview.net/forum?id=Lf0W2gmNBg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-566", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LaViDa: A Large Diffusion Language Model for Multimodal Understanding", "authors": [ "Shufan Li", "Konstantinos Kallidromitis", "Hritik Bansal", "Akash Gokul", "Yusuke Kato", "Kazuki Kozuka", "Jason Kuen", "Zhe Lin", "Kai-Wei Chang", "Aditya Grover" ], "affiliation": "", "summary": "Modern Vision-Language Models (VLMs) can solve a wide range of tasks requiring visual reasoning", "links": { "openreview": "https://openreview.net/forum?id=6WnBITpnzD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-567", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Theory-Driven Label-Specific Representation for Incomplete Multi-View Multi-Label Learning", "authors": [ "Quanjiang Li", "Tianxiang Xu", "Tingjin Luo", "Yan Zhong", "Yang Li", "Yiyun Zhou", "Chenping Hou" ], "affiliation": "", "summary": "Multi-view multi-label learning typically suffers from dual data incompleteness due to limitations in feature storage and annotation costs", "links": { "openreview": "https://openreview.net/forum?id=0Az25lvdT2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-568", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scalable Fingerprinting of Large Language Models", "authors": [ "Anshul Nasery", "Jonathan Hayase", "Creston Brooks", "Peiyao Sheng", "Himanshu Tyagi", "Pramod Viswanath", "Sewoong Oh" ], "affiliation": "", "summary": "Model fingerprinting has emerged as a powerful tool for model owners to identify their shared model given API access", "links": { "openreview": "https://openreview.net/forum?id=CRyOyiVvvJ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-569", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Can Knowledge-Graph-based Retrieval Augmented Generation Really Retrieve What You Need?", "authors": [ "Junchi Yu", "Yujie Liu", "Jindong Gu", "Philip Torr", "Dongzhan Zhou" ], "affiliation": "", "summary": "Retrieval-Augmented Generation (RAG) based on knowledge graphs (KGs) enhances large language models (LLMs) by providing structured and interpretable external knowledge", "links": { "openreview": "https://openreview.net/forum?id=po0eyoYFUa", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-570", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Decomposing Interventional Causality into Synergistic, Redundant, and Unique Components", "authors": [ "Abel Jansma" ], "affiliation": "", "summary": "We introduce a novel framework for decomposing interventional causal effects into synergistic, redundant, and unique components, building on the intuition of Partial Information Decomposition (PID) an", "links": { "openreview": "https://openreview.net/forum?id=yPnEvPq3kV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-571", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Long-Tailed Recognition via Information-Preservable Two-Stage Learning", "authors": [ "Fudong Lin", "Xu Yuan" ], "affiliation": "", "summary": "The imbalance (or long-tail) is the nature of many real-world data distributions, which often induces the undesirable bias of deep classification models toward frequent classes, resulting in poor perf", "links": { "openreview": "https://openreview.net/forum?id=UBsYf2lyNE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-572", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mozart: Modularized and Efficient MoE Training on 3.5D Wafer-Scale Chiplet Architectures", "authors": [ "Shuqing Luo", "Ye Han", "Pingzhi Li", "Jiayin Qin", "Jie Peng", "Yang Katie Zhao", "Yu Cao", "Tianlong Chen" ], "affiliation": "", "summary": "Mixture-of-Experts (MoE) architecture offers enhanced efficiency for Large Language Models (LLMs) with modularized computation, yet its inherent sparsity poses significant hardware deployment challeng", "links": { "openreview": "https://openreview.net/forum?id=zWHKKspghT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-573", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Data Mixing Can Induce Phase Transitions in Knowledge Acquisition", "authors": [ "Xinran Gu", "Kaifeng Lyu", "Jiazheng Li", "Jingzhao Zhang" ], "affiliation": "", "summary": "Large Language Models (LLMs) are typically trained on data mixtures: most data come from web scrapes, while a small portion is curated from high-quality sources with dense domain-specific knowledge", "links": { "openreview": "https://openreview.net/forum?id=tQZK5frjVU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-574", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models", "authors": [ "Wenda Li", "Huijie Zhang", "Qing Qu" ], "affiliation": "", "summary": "The widespread use of AI-generated content from diffusion models has raised significant concerns regarding misinformation and copyright infringement", "links": { "openreview": "https://openreview.net/forum?id=Vj56Z9yNCr", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-575", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Offline Guarded Safe Reinforcement Learning for Medical Treatment Optimization Strategies", "authors": [ "Runze Yan", "Xun Shen", "Akifumi Wachi", "Sebastien Gros", "Anni Zhao", "Xiao Hu" ], "affiliation": "", "summary": "When applying offline reinforcement learning (RL) in healthcare scenarios, the out-of-distribution (OOD) issues pose significant risks, as inappropriate generalization beyond clinical expertise can re", "links": { "openreview": "https://openreview.net/forum?id=4P6Mployhf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-576", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "From Counterfactuals to Trees: Competitive Analysis of Model Extraction Attacks", "authors": [ "Awa Khouna", "Julien Ferry", "Thibaut Vidal" ], "affiliation": "", "summary": "The advent of Machine Learning as a Service (MLaaS) has heightened the trade-off between model explainability and security", "links": { "openreview": "https://openreview.net/forum?id=6vcgsrK6pN", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-577", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Repo2Run: Automated Building Executable Environment for Code Repository at Scale", "authors": [ "Ruida Hu", "Chao Peng", "XinchenWang", "Junjielong Xu", "Cuiyun Gao" ], "affiliation": "", "summary": "Scaling up executable code data is significant for improving language models’ software engineering capability", "links": { "openreview": "https://openreview.net/forum?id=fZsd3KLMje", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-578", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Dynamic 3D Reconstruction of Hand-Instrument Interaction in Ophthalmic Surgery", "authors": [ "Ming Hu", "Zhengdi Yu", "Feilong Tang", "Kaiwen Chen", "Yulong Li", "Imran Razzak", "Junjun He", "Tolga Birdal", "Kaijing Zhou", "Zongyuan Ge" ], "affiliation": "", "summary": "Accurate 3D reconstruction of hands and instruments is critical for vision-based analysis of ophthalmic microsurgery, yet progress has been hampered by the lack of realistic, large-scale datasets and", "links": { "openreview": "https://openreview.net/forum?id=pOJBw1YQgL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-579", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Unlocking Dataset Distillation with Diffusion Models", "authors": [ "Brian Bernhard Moser", "Federico Raue", "Sebastian Palacio", "Stanislav Frolov", "Andreas Dengel" ], "affiliation": "", "summary": "Dataset distillation seeks to condense datasets into smaller but highly representative synthetic samples", "links": { "openreview": "https://openreview.net/forum?id=c6O18DyBBx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-580", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Vision-centric Token Compression in Large Language Model", "authors": [ "Ling Xing", "Alex Jinpeng Wang", "Rui Yan", "Xiangbo Shu", "Jinhui Tang" ], "affiliation": "", "summary": "Real-world applications are stretching context windows to hundreds of thousand of tokens while Large Language Models (LLMs) swell from billions to trillions of parameters", "links": { "openreview": "https://openreview.net/forum?id=YdggdEL41C", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-581", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "VITA-1.5: Towards GPT-4o Level Real-Time Vision and Speech Interaction", "authors": [ "Chaoyou Fu", "Haojia Lin", "Xiong Wang", "YiFan Zhang", "Yunhang Shen", "Xiaoyu Liu", "Haoyu Cao", "Zuwei Long", "Heting Gao", "Ke Li", "Long MA", "Xiawu Zheng", "Rongrong Ji", "Xing Sun", "Caifeng Shan", "Ran He" ], "affiliation": "", "summary": "Recent Multimodal Large Language Models (MLLMs) have typically focused on integrating visual and textual modalities, with less emphasis placed on the role of speech in enhancing interaction", "links": { "openreview": "https://openreview.net/forum?id=8PUzLga3lU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-582", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "EDELINE: Enhancing Memory in Diffusion-based World Models via Linear-Time Sequence Modeling", "authors": [ "Jia-Hua Lee", "Bor-Jiun Lin", "Wei-Fang Sun", "Chun-Yi Lee" ], "affiliation": "", "summary": "World models represent a promising approach for training reinforcement learning agents with significantly improved sample efficiency", "links": { "openreview": "https://openreview.net/forum?id=ph1V6n7BSv", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-583", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "UniTok: a Unified Tokenizer for Visual Generation and Understanding", "authors": [ "Chuofan Ma", "Yi Jiang", "Junfeng Wu", "Jihan Yang", "Xin Yu", "Zehuan Yuan", "BINGYUE PENG", "XIAOJUAN QI" ], "affiliation": "", "summary": "Visual generative and understanding models typically rely on distinct tokenizers to process images, presenting a key challenge for unifying them within a single framework", "links": { "openreview": "https://openreview.net/forum?id=f6aOPkGE8L", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-584", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "NormFit: A Lightweight Solution for Few-Shot Federated Learning with Non-IID Data", "authors": [ "Azadeh Motamedi", "Jae-Mo Kang", "Il-Min Kim" ], "affiliation": "", "summary": "Vision–Language Models (VLMs) have recently attracted considerable attention in Federated Learning (FL) due to their strong and robust performance", "links": { "openreview": "https://openreview.net/forum?id=LP4Q7tPMbs", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-585", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Comprehensive Scene Understanding: Integrating First and Third-Person Views for LVLMs", "authors": [ "Insu Lee", "Wooje Park", "Jaeyun Jang", "Minyoung Noh", "Kyuhong Shim", "Byonghyo Shim" ], "affiliation": "", "summary": "Large vision-language models (LVLMs) are increasingly deployed in interactive applications such as virtual and augmented reality, where a first-person (egocentric) view captured by head-mounted camera", "links": { "openreview": "https://openreview.net/forum?id=m5wrqqcWbN", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-586", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Accelerating Visual-Policy Learning through Parallel Differentiable Simulation", "authors": [ "Haoxiang You", "Yilang Liu", "Ian Abraham" ], "affiliation": "", "summary": "In this work, we propose a computationally efficient algorithm for visual policy learning that leverages differentiable simulation and first-order analytical policy gradients", "links": { "openreview": "https://openreview.net/forum?id=4frj038M6W", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-587", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Online Prediction with Limited Selectivity", "authors": [ "Licheng Liu", "Mingda Qiao" ], "affiliation": "", "summary": "Selective prediction [Dru13, QV19] models the scenario where a forecaster freely decides on the prediction window that their forecast spans", "links": { "openreview": "https://openreview.net/forum?id=HFT821Q83J", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-588", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Generative Trajectory Stitching through Diffusion Composition", "authors": [ "Yunhao Luo", "Utkarsh Aashu Mishra", "Yilun Du", "Danfei Xu" ], "affiliation": "", "summary": "Effective trajectory stitching for long-horizon planning is a significant challenge in robotic decision-making", "links": { "openreview": "https://openreview.net/forum?id=VCTt5DXiBe", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-589", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "VLMs have Tunnel Vision: Evaluating Nonlocal Visual Reasoning in Leading VLMs", "authors": [ "Shmuel Berman", "Jia Deng" ], "affiliation": "", "summary": "Vision Language Models (VLMs) excel at complex visual tasks such as VQA and chart understanding, yet recent work suggests they struggle with simple perceptual tests", "links": { "openreview": "https://openreview.net/forum?id=bRWkBD2BfK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-590", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Neighborhood Self-Dissimilarity Attention for Medical Image Segmentation", "authors": [ "Chen Junren", "Rui Chen", "Wang-wei", "Junlong Cheng", "Gang Liang", "zhanglei-scu", "Liangyin Chen" ], "affiliation": "", "summary": "Medical image segmentation based on neural networks is pivotal in promoting digital health equity", "links": { "openreview": "https://openreview.net/forum?id=tBhEHymG1m", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-591", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Learning to Factorize Spatio-Temporal Foundation Models", "authors": [ "Siru Zhong", "Junjie Qiu", "Yangyu Wu", "Xingchen Zou", "Zhongwen Rao", "Bin Yang", "Chenjuan Guo", "Hao Xu", "Yuxuan Liang" ], "affiliation": "", "summary": "Spatio-Temporal Foundation Models (STFMs) promise zero/few-shot generalization across various datasets, yet joint spatio-temporal pretraining is computationally prohibitive and struggles with domain-s", "links": { "openreview": "https://openreview.net/forum?id=d4CZoiaXeC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-592", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Angles Don’t Lie: Unlocking Training‑Efficient RL Through the Model’s Own Signals", "authors": [ "Qinsi Wang", "Jinghan Ke", "Hancheng Ye", "Yueqian Lin", "Yuzhe Fu", "Jianyi Zhang", "Kurt Keutzer", "Chenfeng Xu", "Yiran Chen" ], "affiliation": "", "summary": "Current Reinforcement Fine-tuning (RFT) paradigms for Large Language Models (LLMs) suffer from sample inefficiency due to the redundant exposure of identical queries under uniform data sampling", "links": { "openreview": "https://openreview.net/forum?id=KGt0F2yjBz", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-593", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Generalizable Insights for Graph Transformers in Theory and Practice", "authors": [ "Timo Stoll", "Luis Müller", "Christopher Morris" ], "affiliation": "", "summary": "Graph Transformers (GTs) have shown strong empirical performance, yet current architectures vary widely in their use of attention mechanisms, positional embeddings (PEs), and expressivity", "links": { "openreview": "https://openreview.net/forum?id=ROfYsQ2KNV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-594", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Principled Data Augmentation for Learning to Solve Quadratic Programming Problems", "authors": [ "Chendi Qian", "Christopher Morris" ], "affiliation": "", "summary": "Linear and quadratic optimization are crucial in numerous real-world applications, ranging from training machine learning models to solving integer linear programs", "links": { "openreview": "https://openreview.net/forum?id=n5NZqzAITL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-595", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Dual Data Alignment Makes AI-Generated Image Detector Easier Generalizable", "authors": [ "Ruoxin Chen", "Junwei Xi", "Zhiyuan Yan", "Ke-Yue Zhang", "Shuang Wu", "Jingyi Xie", "Xu Chen", "Lei Xu", "Isabel Guan", "Taiping Yao", "Shouhong Ding" ], "affiliation": "", "summary": "The rapid increase in AI-generated images (AIGIs) underscores the need for detection methods", "links": { "openreview": "https://openreview.net/forum?id=C39ShJwtD5", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-596", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CSBrain: A Cross-scale Spatiotemporal Brain Foundation Model for EEG Decoding", "authors": [ "Yuchen Zhou", "Jiamin Wu", "Zichen Ren", "Zhouheng Yao", "Weiheng Lu", "Kunyu Peng", "Qihao Zheng", "Chunfeng Song", "Wanli Ouyang", "Chao Gou" ], "affiliation": "", "summary": "Understanding and decoding human brain activity from electroencephalography (EEG) signals is a fundamental problem in neuroscience and artificial intelligence, with applications ranging from cognition", "links": { "openreview": "https://openreview.net/forum?id=agcXjEHmyW", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-597", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Flattening Hierarchies with Policy Bootstrapping", "authors": [ "John Luoyu Zhou", "Jonathan Kao" ], "affiliation": "", "summary": "Offline goal-conditioned reinforcement learning (GCRL) is a promising approach for pretraining generalist policies on large datasets of reward-free trajectories, akin to the self-supervised objectives", "links": { "openreview": "https://openreview.net/forum?id=KaD2Dw8Ahz", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-598", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "HopaDIFF: Holistic-Partial Aware Fourier Conditioned Diffusion for Referring Human Action Segmentation in Multi-Person Scenarios", "authors": [ "Kunyu Peng", "Junchao Huang", "Xiangsheng Huang", "Di Wen", "Junwei Zheng", "Yufan Chen", "Kailun Yang", "Jiamin Wu", "Chongqing Hao", "Rainer Stiefelhagen" ], "affiliation": "", "summary": "Action segmentation is a core challenge in high-level video understanding, aiming to partition untrimmed videos into segments and assign each a label from a predefined action set", "links": { "openreview": "https://openreview.net/forum?id=xOqCKB8XIl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-599", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Geometry Meets Incentives: Sample-Efficient Incentivized Exploration with Linear Contexts", "authors": [ "Benjamin Schiffer", "Mark Sellke" ], "affiliation": "", "summary": "In the incentivized exploration model, a principal aims to explore and learn over time by interacting with a sequence of self-interested agents", "links": { "openreview": "https://openreview.net/forum?id=nwlX15Wnr9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-600", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GeoRemover: Removing Objects and Their Causal Visual Artifacts", "authors": [ "Zixin Zhu", "Haoxiang Li", "Xuelu Feng", "He Wu", "Chunming Qiao", "Junsong Yuan" ], "affiliation": "", "summary": "Towards intelligent image editing, object removal should eliminate both the target object and its causal visual artifacts, such as shadows and reflections", "links": { "openreview": "https://openreview.net/forum?id=RnfyqrkOxD", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-601", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "The Primacy of Magnitude in Low-Rank Adaptation", "authors": [ "Zicheng Zhang", "Haoran Li", "Yifeng Zhang", "Guoqiang Gong", "Jiaxing Wang", "Junxing Hu", "pengzhang liu", "Qixia Jiang" ], "affiliation": "", "summary": "Low-Rank Adaptation (LoRA) offers a parameter-efficient paradigm for tuning large models", "links": { "openreview": "https://openreview.net/forum?id=s4LnWgjacg", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-602", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GeRaF: Neural Geometry Reconstruction from Radio Frequency Signals", "authors": [ "Jiachen Lu", "Hailan Shanbhag", "Haitham Al Hassanieh" ], "affiliation": "", "summary": "GeRaF is the first method to use neural implicit learning for near-range 3D geometry reconstruction from radio frequency (RF) signals", "links": { "openreview": "https://openreview.net/forum?id=z3PMVmzoya", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-603", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Spatial Understanding from Videos: Structured Prompts Meet Simulation Data", "authors": [ "Haoyu Zhang", "Meng Liu", "Zaijing Li", "Haokun Wen", "Weili Guan", "Yaowei Wang", "Liqiang Nie" ], "affiliation": "", "summary": "Visual-spatial understanding, the ability to infer object relationships and layouts from visual input, is fundamental to downstream tasks such as robotic navigation and embodied interaction", "links": { "openreview": "https://openreview.net/forum?id=SBYCu5uJJf", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-604", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SATURN: SAT-based Reinforcement Learning to Unleash LLMs Reasoning", "authors": [ "Huanyu Liu", "Ge Li", "Jia Li", "Hao Zhu", "Kechi Zhang", "Yihong Dong" ], "affiliation": "", "summary": "How to design reinforcement learning (RL) tasks that effectively unleash the reasoning capability of large language models (LLMs) remains an open question", "links": { "openreview": "https://openreview.net/forum?id=Sct4sajCi6", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-605", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "JavisGPT: A Unified Multi-modal LLM for Sounding-Video Comprehension and Generation", "authors": [ "Kai Liu", "Jungang Li", "Yuchong Sun", "Shengqiong Wu", "jianzhang gao", "Daoan Zhang", "Wei Zhang", "Sheng Jin", "Sicheng Yu", "Geng Zhan", "Jiayi Ji", "Fan Zhou", "Liang Zheng", "Shuicheng YAN", "Hao Fei", "Tat-Seng Chua" ], "affiliation": "", "summary": "This paper presents JavisGPT, the first unified multimodal large language model (MLLM) for joint audio-video (JAV) comprehension and generation", "links": { "openreview": "https://openreview.net/forum?id=MZoOpD9NHV", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-606", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Dimension-adapted Momentum Outscales SGD", "authors": [ "Damien Ferbach", "Katie Everett", "Gauthier Gidel", "Elliot Paquette", "Courtney Paquette" ], "affiliation": "", "summary": "We investigate scaling laws for stochastic momentum algorithms on the power law random features model, parameterized by data complexity, target complexity, and model size", "links": { "openreview": "https://openreview.net/forum?id=t4aN2G7Ucc", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-607", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "From Experts to a Generalist: Toward General Whole-Body Control for Humanoid Robots", "authors": [ "Yuxuan Wang", "Ming Yang", "Ziluo Ding", "Yu Zhang", "Weishuai Zeng", "Xinrun Xu", "Haobin Jiang", "Zongqing Lu" ], "affiliation": "", "summary": "Achieving general agile whole-body control on humanoid robots remains a major challenge due to diverse motion demands and data conflicts", "links": { "openreview": "https://openreview.net/forum?id=ZBSkyMwdEB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-608", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LoRAShop: Training-Free Multi-Concept Image Generation and Editing with Rectified Flow Transformers", "authors": [ "Yusuf Dalva", "Hidir Yesiltepe", "Pinar Yanardag" ], "affiliation": "", "summary": "We introduce LoRAShop, the first framework for multi-concept image generation and editing with LoRA models", "links": { "openreview": "https://openreview.net/forum?id=VlvtStQN34", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-609", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction", "authors": [ "Jiahe Li", "Jiawei Zhang", "Youmin Zhang", "Xiao Bai", "Jin Zheng", "Xiaohan Yu", "Lin Gu" ], "affiliation": "", "summary": "Reconstructing accurate surfaces with radiance fields has achieved remarkable progress in recent years", "links": { "openreview": "https://openreview.net/forum?id=EtqwyqJrJO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-610", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Transformer Copilot: Learning from The Mistake Log in LLM Fine-tuning", "authors": [ "Jiaru Zou", "Yikun Ban", "Zihao Li", "Yunzhe Qi", "Ruizhong Qiu", "Ling Yang", "Jingrui He" ], "affiliation": "", "summary": "Large language models are typically adapted to downstream tasks through supervised fine-tuning on domain-specific data", "links": { "openreview": "https://openreview.net/forum?id=MRvxlTlkNQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-611", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "scMRDR: A scalable and flexible framework for unpaired single-cell multi-omics data integration", "authors": [ "Jianle Sun", "Chaoqi Liang", "Ran Wei", "Peng Zheng", "LEI BAI", "Wanli Ouyang", "Hongliang Yan", "Peng Ye" ], "affiliation": "", "summary": "Advances in single-cell sequencing have enabled high-resolution profiling of diverse molecular modalities, while integrating unpaired multi-omics single-cell data remains challenging", "links": { "openreview": "https://openreview.net/forum?id=tI04KmK27S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-612", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Imitation Beyond Expectation Using Pluralistic Stochastic Dominance", "authors": [ "Ali Farajzadeh", "Danyal Saeed", "Syed M Abbas", "Rushit N. Shah", "Aadirupa Saha", "Brian D Ziebart" ], "affiliation": "", "summary": "Imitation learning seeks policies reflecting the values of demonstrated behaviors", "links": { "openreview": "https://openreview.net/forum?id=YX5DHa9OfX", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-613", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Cloud4D: Estimating Cloud Properties at a High Spatial and Temporal Resolution", "authors": [ "Jacob Lin", "Edward Gryspeerdt", "Ronald Clark" ], "affiliation": "", "summary": "There has been great progress in improving numerical weather prediction and climate models using machine learning", "links": { "openreview": "https://openreview.net/forum?id=g2AAvmBwkS", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-614", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "CLiFT: Compressive Light-Field Tokens for Compute Efficient and Adaptive Neural Rendering", "authors": [ "Zhengqing Wang", "Yuefan Wu", "Jiacheng Chen", "Fuyang Zhang", "Yasutaka Furukawa" ], "affiliation": "", "summary": "This paper proposes a neural rendering approach that represents a scene as \"compressed light-field tokens (CLiFTs)\", retaining rich appearance and geometric information of a scene", "links": { "openreview": "https://openreview.net/forum?id=MGJVhzWa2s", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-615", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Adaptive Neighborhood-Constrained Q Learning for Offline Reinforcement Learning", "authors": [ "Yixiu Mao", "Yun Qu", "Cheems Wang", "Xiangyang Ji" ], "affiliation": "", "summary": "Offline reinforcement learning (RL) suffers from extrapolation errors induced by out-of-distribution (OOD) actions", "links": { "openreview": "https://openreview.net/forum?id=qgi5TfBXBw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-616", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "To Think or Not To Think: A Study of Thinking in Rule-Based Visual Reinforcement Fine-Tuning", "authors": [ "Ming Li", "Jike Zhong", "Shitian Zhao", "Yuxiang Lai", "Haoquan Zhang", "Wang Bill Zhu", "Kaipeng Zhang" ], "affiliation": "", "summary": "This paper investigates the role of explicit thinking process in rule-based reinforcement fine-tuning (RFT) for multi-modal large language models (MLLMs)", "links": { "openreview": "https://openreview.net/forum?id=YexxvBGwQM", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-617", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DexGarmentLab: Dexterous Garment Manipulation Environment with Generalizable Policy", "authors": [ "Yuran Wang", "Ruihai Wu", "Yue Chen", "Jiarui Wang", "Jiaqi Liang", "Ziyu Zhu", "Haoran Geng", "Jitendra Malik", "Pieter Abbeel", "Hao Dong" ], "affiliation": "", "summary": "Garment manipulation is a critical challenge due to the diversity in garment categories, geometries, and deformations", "links": { "openreview": "https://openreview.net/forum?id=ZZ09oX2Xpo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-618", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "STITCH-OPE: Trajectory Stitching with Guided Diffusion for Off-Policy Evaluation", "authors": [ "Hossein Goli", "Michael Gimelfarb", "Nathan Samuel de Lara", "Haruki Nishimura", "Masha Itkina", "Florian Shkurti" ], "affiliation": "", "summary": "Off-policy evaluation (OPE) estimates the performance of a target policy using offline data collected from a behavior policy, and is crucial in domains such as robotics or healthcare where direct inte", "links": { "openreview": "https://openreview.net/forum?id=AghtKxDf7f", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-619", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Neural Atlas Graphs for Dynamic Scene Decomposition and Editing", "authors": [ "Jan Philipp Schneider", "Pratik Singh Bisht", "Ilya Chugunov", "Andreas Kolb", "Michael Moeller", "Felix Heide" ], "affiliation": "", "summary": "Learning editable high-resolution scene representations for dynamic scenes is an open problem with applications across the domains from autonomous driving to creative editing - the most successful app", "links": { "openreview": "https://openreview.net/forum?id=pkuVonMwhT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-620", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "LeMiCa: Lexicographic Minimax Path Caching for Efficient Diffusion-Based Video Generation", "authors": [ "Huanlin Gao", "Ping Chen", "Fuyuan Shi", "Chao Tan", "Zhaoxiang Liu", "Fang Zhao", "Kai Wang", "Shiguo Lian" ], "affiliation": "", "summary": "We present LeMiCa, a training-free and efficient acceleration framework for diffusion-based video generation", "links": { "openreview": "https://openreview.net/forum?id=QIXdI207nq", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-621", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Robust SuperAlignment: Weak-to-Strong Robustness Generalization for Vision-Language Models", "authors": [ "Junhao Dong", "Cong Zhang", "Xinghua Qu", "Zejun MA", "Piotr Koniusz", "Yew-Soon Ong" ], "affiliation": "", "summary": "Numerous well-established studies have demonstrated the superhuman capabilities of modern Vision-Language Models (VLMs) across a wide range of tasks", "links": { "openreview": "https://openreview.net/forum?id=rOR5IZcwJx", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-622", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Thought Communication in Multiagent Collaboration", "authors": [ "Yujia Zheng", "Zhuokai Zhao", "Zijian Li", "Yaqi Xie", "Mingze Gao", "Lizhu Zhang", "Kun Zhang" ], "affiliation": "", "summary": "Natural language has long enabled human cooperation, but its lossy, ambiguous, and indirect nature limits the potential of collective intelligence", "links": { "openreview": "https://openreview.net/forum?id=tq9lyV9Cml", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-623", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Polyline Path Masked Attention for Vision Transformer", "authors": [ "Zhongchen Zhao", "Chaodong Xiao", "Hui LIN", "Qi Xie", "Lei Zhang", "Deyu Meng" ], "affiliation": "", "summary": "Global dependency modeling and spatial position modeling are two core issues of the foundational architecture design in current deep learning frameworks", "links": { "openreview": "https://openreview.net/forum?id=EkoAKNikAj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-624", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PLMTrajRec: A Scalable and Generalizable Trajectory Recovery Method with Pre-trained Language Models", "authors": [ "Tonglong Wei", "Yan Lin", "Youfang Lin", "Shengnan Guo", "Jilin Hu", "Haitao Yuan", "Gao Cong", "Huaiyu Wan" ], "affiliation": "", "summary": "Spatiotemporal trajectory data is crucial for various traffic-related applications", "links": { "openreview": "https://openreview.net/forum?id=SJPq1xBPHX", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-625", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Optimization Inspired Few-Shot Adaptation for Large Language Models", "authors": [ "Boyan Gao", "Xin Wang", "Yibo Yang", "David A. Clifton" ], "affiliation": "", "summary": "Large Language Models (LLMs) have demonstrated remarkable performance in real-world applications", "links": { "openreview": "https://openreview.net/forum?id=rZ2nSt1X58", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-626", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Scent of Knowledge: Optimizing Search-Enhanced Reasoning with Information Foraging", "authors": [ "Hongjin Qian", "Zheng Liu" ], "affiliation": "", "summary": "Augmenting large language models (LLMs) with external retrieval has become a standard method to address their inherent knowledge cutoff limitations", "links": { "openreview": "https://openreview.net/forum?id=26kUrQm4zw", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-627", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Taccel: Scaling Up Vision-based Tactile Robotics via High-performance GPU Simulation", "authors": [ "Yuyang Li", "Wenxin Du", "Chang Yu", "Puhao Li", "Zihang Zhao", "Tengyu Liu", "Chenfanfu Jiang", "Yixin Zhu", "Siyuan Huang" ], "affiliation": "", "summary": "Tactile sensing is crucial for achieving human-level robotic capabilities in manipulation tasks", "links": { "openreview": "https://openreview.net/forum?id=PtGMadeONU", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-628", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Toward Relative Positional Encoding in Spiking Transformers", "authors": [ "Changze Lv", "Yansen Wang", "Dongqi Han", "Yifei Shen", "Xiaoqing Zheng", "Xuanjing Huang", "Dongsheng Li" ], "affiliation": "", "summary": "Spiking neural networks (SNNs) are bio-inspired networks that mimic how neurons in the brain communicate through discrete spikes, which have great potential in various tasks due to their energy effici", "links": { "openreview": "https://openreview.net/forum?id=MDWJlTWZHH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-629", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "TransMLA: Migrating GQA Models to MLA with Full DeepSeek Compatibility and Speedup", "authors": [ "Fanxu Meng", "Pingzhi Tang", "Zengwei Yao", "Xing Sun", "Muhan Zhang" ], "affiliation": "", "summary": "Modern large-language models often face communication bottlenecks on current hardware rather than computational limitations", "links": { "openreview": "https://openreview.net/forum?id=TcVCu2PKb9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-630", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Understanding the Mechanisms of Classifier-Free Guidance", "authors": [ "Xiang Li", "Rongrong Wang", "Qing Qu" ], "affiliation": "", "summary": "Classifier-free guidance (CFG) is a core technique powering state-of-the-art image generation systems, yet its underlying mechanisms remain poorly understood", "links": { "openreview": "https://openreview.net/forum?id=bRAm7A02Qm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-631", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PhysX-3D: Physical-Grounded 3D Asset Generation", "authors": [ "Ziang Cao", "Zhaoxi Chen", "Liang Pan", "Ziwei Liu" ], "affiliation": "", "summary": "3D modeling is moving from virtual to physical", "links": { "openreview": "https://openreview.net/forum?id=hLJLP3CmHR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-632", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Towards Physics-informed Spatial Intelligence with Human Priors: An Autonomous Driving Pilot Study", "authors": [ "Guanlin Wu", "Boyan Su", "Yang Zhao", "Pu Wang", "Yichen Lin", "Hao Frank Yang" ], "affiliation": "", "summary": "How to integrate and verify spatial intelligence in foundation models remains an open challenge", "links": { "openreview": "https://openreview.net/forum?id=pEUBqS8nTk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-633", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "StreamForest: Efficient Online Video Understanding with Persistent Event Memory", "authors": [ "Xiangyu Zeng", "Kefan Qiu", "Qingyu Zhang", "Xinhao Li", "Jing Wang", "Jiaxin Li", "Ziang Yan", "Kun Tian", "Meng Tian", "Xinhai Zhao", "Yi Wang", "Limin Wang" ], "affiliation": "", "summary": "Multimodal Large Language Models (MLLMs) have recently achieved remarkable progress in video understanding", "links": { "openreview": "https://openreview.net/forum?id=9loSPaBwGO", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-634", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Vanish into Thin Air: Cross-prompt Universal Adversarial Attacks for SAM2", "authors": [ "Ziqi Zhou", "Yifan Hu", "Yufei Song", "Zijing Li", "Shengshan Hu", "Leo Yu Zhang", "Dezhong Yao", "Long Zheng", "Hai Jin" ], "affiliation": "", "summary": "Recent studies reveal the vulnerability of the image segmentation foundation model SAM to adversarial examples", "links": { "openreview": "https://openreview.net/forum?id=Ll29PmM3UH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-635", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Improving Evolutionary Multi-View Classification via Eliminating Individual Fitness Bias", "authors": [ "Xinyan Liang", "ShuaiLi", "Qian Guo", "Yuhua Qian", "Bingbing Jiang", "Tingjin Luo", "Liang Du" ], "affiliation": "", "summary": "Evolutionary multi-view classification (EMVC) methods have gained wide recognition due to their adaptive mechanisms", "links": { "openreview": "https://openreview.net/forum?id=xgTxQe3CNl", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-636", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Right Question is Already Half the Answer: Fully Unsupervised LLM Reasoning Incentivization", "authors": [ "Qingyang Zhang", "Haitao Wu", "Changqing Zhang", "Peilin Zhao", "Yatao Bian" ], "affiliation": "", "summary": "Existing methods to enhance the reasoning capability of large language models predominantly rely on supervised fine-tuning (SFT) followed by reinforcement learning (RL) on reasoning-specific data", "links": { "openreview": "https://openreview.net/forum?id=k8Mim6RI5O", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-637", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MDReID: Modality-Decoupled Learning for Any-to-Any Multi-Modal Object Re-Identification", "authors": [ "Yingying Feng", "Jie Li", "Jie Hu", "Yukang Zhang", "Lei Tan", "Jiayi Ji" ], "affiliation": "", "summary": "The challenge of inconsistent modalities in real-world applications presents significant obstacles to effective object re-identification (ReID)", "links": { "openreview": "https://openreview.net/forum?id=7jg26Fd1ra", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-638", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "RobustMerge: Parameter-Efficient Model Merging for MLLMs with Direction Robustness", "authors": [ "Fanhu Zeng", "Haiyang Guo", "Fei Zhu", "Li Shen", "Hao Tang" ], "affiliation": "", "summary": "Fine-tuning pre-trained models with custom data leads to numerous expert models on specific tasks", "links": { "openreview": "https://openreview.net/forum?id=U7RZ9cC73S", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-639", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Stable Part Diffusion 4D: Multi-View RGB and Kinematic Parts Video Generation", "authors": [ "Hao Zhang", "Chun-Han Yao", "Simon Donné", "Narendra Ahuja", "Varun Jampani" ], "affiliation": "", "summary": "We present Stable Part Diffusion 4D (SP4D), a framework for generating paired RGB and kinematic part videos from monocular inputs", "links": { "openreview": "https://openreview.net/forum?id=I9F53Qlwur", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-640", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OnlineSplatter: Pose-Free Online 3D Reconstruction for Free-Moving Objects", "authors": [ "Mark He Huang", "Lin Geng Foo", "Christian Theobalt", "Ying Sun", "De Wen Soh" ], "affiliation": "", "summary": "Free-moving object reconstruction from monocular video remains challenging, particularly without reliable pose or depth cues and under arbitrary object motion", "links": { "openreview": "https://openreview.net/forum?id=Y9AdTCCEgI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-641", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Rectified Point Flow: Generic Point Cloud Pose Estimation", "authors": [ "TAO SUN", "Liyuan Zhu", "Shengyu Huang", "Shuran Song", "Iro Armeni" ], "affiliation": "", "summary": "We present Rectified Point Flow, a unified parameterization that formulates pairwise point cloud registration and multi-part shape assembly as a single conditional generative problem", "links": { "openreview": "https://openreview.net/forum?id=bNTezDPlFH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-642", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Cue3D: Quantifying the Role of Image Cues in Single-Image 3D Generation", "authors": [ "Xiang Li", "Zirui Wang", "Zixuan Huang", "James Matthew Rehg" ], "affiliation": "", "summary": "Humans and traditional computer vision methods rely on a diverse set of monocular cues to infer 3D structure from a single image, such as shading, texture, silhouette, etc", "links": { "openreview": "https://openreview.net/forum?id=8a9bAZFeIu", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-643", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SceneDesigner: Controllable Multi-Object Image Generation with 9-DoF Pose Manipulation", "authors": [ "Zhenyuan Qin", "Xincheng Shuai", "Henghui Ding" ], "affiliation": "", "summary": "Controllable image generation has attracted increasing attention in recent years, enabling users to manipulate visual content such as identity and style", "links": { "openreview": "https://openreview.net/forum?id=yFasd68NyI", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-644", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "StelLA: Subspace Learning in Low-rank Adaptation using Stiefel Manifold", "authors": [ "Zhizhong Li", "Sina Sajadmanesh", "Jingtao Li", "Lingjuan Lyu" ], "affiliation": "", "summary": "Low-rank adaptation (LoRA) has been widely adopted as a parameter-efficient technique for fine-tuning large-scale pre-trained models", "links": { "openreview": "https://openreview.net/forum?id=55Lv1unlUL", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-645", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Non-Clairvoyant Scheduling with Progress Bars", "authors": [ "Ziyad Benomar", "Romain Cosson", "Alexander Lindermayr", "Jens Schlöter" ], "affiliation": "", "summary": "In non-clairvoyant scheduling, the goal is to minimize the total job completion time without prior knowledge of individual job processing times", "links": { "openreview": "https://openreview.net/forum?id=gYbreatcV1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-646", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Jacobian-Based Interpretation of Nonlinear Neural Encoding Model", "authors": [ "Xiaohui Gao", "Haoran Yang", "Yue Cheng", "Mengfei Zuo", "Yiheng Liu", "Peiyang Li", "Xintao Hu" ], "affiliation": "", "summary": "In recent years, the alignment between artificial neural network (ANN) embeddings and blood oxygenation level dependent (BOLD) responses in functional magnetic resonance imaging (fMRI) via neural enco", "links": { "openreview": "https://openreview.net/forum?id=P37GIj4wB7", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-647", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mesh-RFT: Enhancing Mesh Generation via Fine-grained Reinforcement Fine-Tuning", "authors": [ "Jian Liu", "Jing Xu", "Song Guo", "Jing Li", "Guojingfeng", "Jiaao Yu", "Haohan Weng", "Biwen Lei", "Xianghui Yang", "Zhuo Chen", "Fangqi Zhu", "Tao Han", "Chunchao Guo" ], "affiliation": "", "summary": "Existing pretrained models for 3D mesh generation often suffer from data biases and produce low-quality results, while global reinforcement learning (RL) methods rely on object-level rewards that stru", "links": { "openreview": "https://openreview.net/forum?id=te2RsWcyQp", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-648", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "VisualQuality-R1: Reasoning-Induced Image Quality Assessment via Reinforcement Learning to Rank", "authors": [ "Tianhe Wu", "Jian Zou", "Jie Liang", "Lei Zhang", "Kede Ma" ], "affiliation": "", "summary": "DeepSeek-R1 has demonstrated remarkable effectiveness in incentivizing reasoning and generalization capabilities of large language models (LLMs) through reinforcement learning", "links": { "openreview": "https://openreview.net/forum?id=uL7lCOHtiZ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-649", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GenColor: Generative and Expressive Color Enhancement with Pixel-Perfect Texture Preservation", "authors": [ "Yi Dong", "Yuxi Wang", "Xianhui Lin", "Wenqi Ouyang", "Zhiqi Shen", "Peiran Ren", "Ruoxi Fan", "Rynson W. H. Lau" ], "affiliation": "", "summary": "Color enhancement is a crucial yet challenging task in digital photography", "links": { "openreview": "https://openreview.net/forum?id=n8AvXKcCeR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-650", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "G-Memory: Tracing Hierarchical Memory for Multi-Agent Systems", "authors": [ "Guibin Zhang", "Muxin Fu", "Kun Wang", "Guancheng Wan", "Miao Yu", "Shuicheng YAN" ], "affiliation": "", "summary": "Large language model (LLM)-powered multi-agent systems (MAS) have demonstrated cognitive and execution capabilities that far exceed those of single LLM agents, yet their capacity for self-evolution re", "links": { "openreview": "https://openreview.net/forum?id=mmIAp3cVS0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-651", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving", "authors": [ "Shuang Zeng", "Xinyuan Chang", "Mengwei Xie", "Xinran Liu", "Yifan Bai", "Zheng Pan", "Mu Xu", "Xing Wei" ], "affiliation": "", "summary": "Vision–Language–Action (VLA) models are increasingly used for end-to-end driving due to their world knowledge and reasoning ability", "links": { "openreview": "https://openreview.net/forum?id=fCirUh6FRb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-652", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "E2Former: An Efficient and Equivariant Transformer with Linear-Scaling Tensor Products", "authors": [ "Yunyang Li", "Lin Huang", "Zhihao Ding", "Xinran Wei", "Chu Wang", "Han Yang", "Zun Wang", "Chang Liu", "Yu Shi", "Peiran Jin", "Tao Qin", "Mark Gerstein", "Jia Zhang" ], "affiliation": "", "summary": "Equivariant Graph Neural Networks (EGNNs) have demonstrated significant success in modeling microscale systems, including those in chemistry, biology and materials science", "links": { "openreview": "https://openreview.net/forum?id=ls5L4IMEwt", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-653", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Adaptive Defense against Harmful Fine-Tuning for Large Language Models via Bayesian Data Scheduler", "authors": [ "Zixuan Hu", "Li Shen", "Zhenyi Wang", "Yongxian Wei", "Dacheng Tao" ], "affiliation": "", "summary": "Harmful fine-tuning poses critical safety risks to fine-tuning-as-a-service for large language models", "links": { "openreview": "https://openreview.net/forum?id=sm2e1SnMK4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-654", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Neptune-X: Active X-to-Maritime Generation for Universal Maritime Object Detection", "authors": [ "Yu Guo", "Shengfeng He", "Yuxu Lu", "Haonan An", "Yihang Tao", "Huilin Zhu", "Jingxian Liu", "Yuguang Fang" ], "affiliation": "", "summary": "Maritime object detection is essential for navigation safety, surveillance, and autonomous operations, yet constrained by two key challenges: the scarcity of annotated maritime data and poor generaliz", "links": { "openreview": "https://openreview.net/forum?id=oIpRvQkrH9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-655", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "AuroRA: Breaking Low-Rank Bottleneck of LoRA with Nonlinear Mapping", "authors": [ "Haonan Dong", "Wenhao Zhu", "Guojie Song", "Liang Wang" ], "affiliation": "", "summary": "Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method validated across NLP and CV domains", "links": { "openreview": "https://openreview.net/forum?id=2hgHyoyVWj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-656", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Puppeteer: Rig and Animate Your 3D Models", "authors": [ "Chaoyue Song", "Xiu Li", "Fan Yang", "Zhongcong Xu", "Jiacheng Wei", "Fayao Liu", "Jiashi Feng", "Guosheng Lin", "Jianfeng Zhang" ], "affiliation": "", "summary": "Modern interactive applications increasingly demand dynamic 3D content, yet the transformation of static 3D models into animated assets constitutes a significant bottleneck in content creation pipelin", "links": { "openreview": "https://openreview.net/forum?id=Wbc3PutCyQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-657", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Talk2Event: Grounded Understanding of Dynamic Scenes from Event Cameras", "authors": [ "Lingdong Kong", "Dongyue Lu", "Alan Liang", "Rong Li", "Yuhao Dong", "Tianshuai Hu", "Lai Xing Ng", "Wei Tsang Ooi", "Benoit R Cottereau" ], "affiliation": "", "summary": "Event cameras offer microsecond-level latency and robustness to motion blur, making them ideal for understanding dynamic environments", "links": { "openreview": "https://openreview.net/forum?id=fxERuSBpfQ", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-658", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing", "authors": [ "Chenxi Xie", "Minghan Li", "Shuai Li", "Yuhui Wu", "Qiaosi Yi", "Lei Zhang" ], "affiliation": "", "summary": "Leveraging the powerful generation capability of large-scale pretrained text-to-image models, training-free methods have demonstrated impressive image editing results", "links": { "openreview": "https://openreview.net/forum?id=JxBA9OJExP", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-659", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Alligat0R: Pre-Training through Covisibility Segmentation for Relative Camera Pose Regression", "authors": [ "Thibaut Loiseau", "Guillaume Bourmaud", "Vincent Lepetit" ], "affiliation": "", "summary": "Pre-training techniques have greatly advanced computer vision, with CroCo’s cross-view completion approach yielding impressive results in tasks like 3D reconstruction and pose regression", "links": { "openreview": "https://openreview.net/forum?id=yHJRI6rzaA", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-660", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DiCo: Revitalizing ConvNets for Scalable and Efficient Diffusion Modeling", "authors": [ "Yuang Ai", "Qihang Fan", "Xuefeng Hu", "Zhenheng Yang", "Ran He", "Huaibo Huang" ], "affiliation": "", "summary": "Diffusion Transformer (DiT), a promising diffusion model for visual generation, demonstrates impressive performance but incurs significant computational overhead", "links": { "openreview": "https://openreview.net/forum?id=UnslcaZSnb", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-661", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Enhancing Time Series Forecasting through Selective Representation Spaces: A Patch Perspective", "authors": [ "Xingjian Wu", "Xiangfei Qiu", "Hanyin Cheng", "Zhengyu Li", "Jilin Hu", "Chenjuan Guo", "Bin Yang" ], "affiliation": "", "summary": "Time Series Forecasting has made significant progress with the help of Patching technique, which partitions time series into multiple patches to effectively retain contextual semantic information into", "links": { "openreview": "https://openreview.net/forum?id=BirE0jYKt0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-662", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Injecting Frame-Event Complementary Fusion into Diffusion for Optical Flow in Challenging Scenes", "authors": [ "Haonan Wang", "Hanyu Zhou", "Haoyue Liu", "Luxin Yan" ], "affiliation": "", "summary": "Optical flow estimation has achieved promising results in conventional scenes but faces challenges in high-speed and low-light scenes, which suffer from motion blur and insufficient illumination", "links": { "openreview": "https://openreview.net/forum?id=oaWpRaZ4jj", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-663", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "OmniSync: Towards Universal Lip Synchronization via Diffusion Transformers", "authors": [ "Ziqiao Peng", "Jiwen Liu", "Haoxian Zhang", "Xiaoqiang Liu", "Songlin Tang", "Pengfei Wan", "Di ZHANG", "Hongyan Liu", "Jun He" ], "affiliation": "", "summary": "Lip synchronization is the task of aligning a speaker’s lip movements in video with corresponding speech audio, and it is essential for creating realistic, expressive video content", "links": { "openreview": "https://openreview.net/forum?id=9XCyUFsm1H", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-664", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "MesaTask: Towards Task-Driven Tabletop Scene Generation via 3D Spatial Reasoning", "authors": [ "Jinkun Hao", "Naifu Liang", "Zhen Luo", "Xudong XU", "Weipeng Zhong", "Ran Yi", "Yichen Jin", "Zhaoyang Lyu", "Feng Zheng", "Lizhuang Ma", "Jiangmiao Pang" ], "affiliation": "", "summary": "The ability of robots to interpret human instructions and execute manipulation tasks necessitates the availability of task-relevant tabletop scenes for training", "links": { "openreview": "https://openreview.net/forum?id=U88JlpY0vR", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-665", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion", "authors": [ "Xun Huang", "Zhengqi Li", "Guande He", "Mingyuan Zhou", "Eli Shechtman" ], "affiliation": "", "summary": "We introduce Self Forcing, a novel training paradigm for autoregressive video diffusion models", "links": { "openreview": "https://openreview.net/forum?id=mSiN7i0BYH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-666", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "DICEPTION: A Generalist Diffusion Model for Visual Perceptual Tasks", "authors": [ "Canyu Zhao", "Yanlong Sun", "Mingyu Liu", "Huanyi Zheng", "Muzhi Zhu", "Zhiyue Zhao", "Hao Chen", "Tong He", "Chunhua Shen" ], "affiliation": "", "summary": "This paper's primary objective is to develop a robust generalist perception model capable of addressing multiple tasks under constraints of computational resources and limited training data", "links": { "openreview": "https://openreview.net/forum?id=hQhAPGCtPo", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-667", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Variational Learning Finds Flatter Solutions at the Edge of Stability", "authors": [ "Avrajit Ghosh", "Bai Cong", "Rio Yokota", "Saiprasad Ravishankar", "Rongrong Wang", "Molei Tao", "Mohammad Emtiyaz Khan", "Thomas Möllenhoff" ], "affiliation": "", "summary": "Variational Learning (VL) has recently gained popularity for training deep neural networks", "links": { "openreview": "https://openreview.net/forum?id=nIFFMrDQ5w", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-668", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Seeing Sound, Hearing Sight: Uncovering Modality Bias and Conflict of AI models in Sound Localization", "authors": [ "Yanhao Jia", "Ji Xie", "S Jivaganesh", "Li Hao", "Xu Wu", "Mengmi Zhang" ], "affiliation": "", "summary": "Imagine hearing a dog bark and instinctively turning toward the sound—only to find a parked car, while a silent dog sits nearby", "links": { "openreview": "https://openreview.net/forum?id=PWdWmw9wh0", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-669", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ALINE: Joint Amortization for Bayesian Inference and Active Data Acquisition", "authors": [ "Daolang Huang", "Xinyi Wen", "Ayush Bharti", "Samuel Kaski", "Luigi Acerbi" ], "affiliation": "", "summary": "Many critical applications, from autonomous scientific discovery to personalized medicine, demand systems that can both strategically acquire the most informative data and instantaneously perform infe", "links": { "openreview": "https://openreview.net/forum?id=btm5Z5Vu8G", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-670", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Fully Autonomous Neuromorphic Navigation and Dynamic Obstacle Avoidance", "authors": [ "Xiaochen Shang", "Luo Pengwei", "Xinning Wang", "Jiayue Zhao", "Huilin Ge", "Bo Dong", "Xin Yang" ], "affiliation": "", "summary": "Unmanned aerial vehicles could accurately accomplish complex navigation and obstacle avoidance tasks under external control", "links": { "openreview": "https://openreview.net/forum?id=11fe8wKkmk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-671", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Enhancing LLM Watermark Resilience Against Both Scrubbing and Spoofing Attacks", "authors": [ "Huanming Shen", "Baizhou Huang", "Xiaojun Wan" ], "affiliation": "", "summary": "Watermarking is widely regarded as a promising defense against the misuse of large language models (LLMs); however, existing methods are fundamentally constrained by their vulnerability to scrubbing a", "links": { "openreview": "https://openreview.net/forum?id=RbdLnwEEjk", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-672", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Mulberry: Empowering MLLM with o1-like Reasoning and Reflection via Collective Monte Carlo Tree Search", "authors": [ "Huanjin Yao", "Jiaxing Huang", "Wenhao Wu", "Jingyi Zhang", "Yibo Wang", "Shunyu Liu", "Yingjie Wang", "YuXin Song", "Haocheng Feng", "Li Shen", "Dacheng Tao" ], "affiliation": "", "summary": "In this work, we aim to develop an MLLM that understands and solves questions by learning to create each intermediate step of the reasoning involved till the final answer", "links": { "openreview": "https://openreview.net/forum?id=lwOV2ACEK9", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-673", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ProtInvTree: Deliberate Protein Inverse Folding with Reward-guided Tree Search", "authors": [ "Mengdi Liu", "Xiaoxue Cheng", "Zhangyang Gao", "Hong Chang", "Cheng Tan", "Shiguang Shan", "Xilin Chen" ], "affiliation": "", "summary": "Designing protein sequences that fold into a target 3D structure—known as protein inverse folding—is a fundamental challenge in protein engineering", "links": { "openreview": "https://openreview.net/forum?id=2uKVyGq5zK", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-674", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "On the Value of Cross-Modal Misalignment in Multimodal Representation Learning", "authors": [ "Yichao Cai", "Yuhang Liu", "Erdun Gao", "Tianjiao Jiang", "Zhen Zhang", "Anton van den Hengel", "Javen Qinfeng Shi" ], "affiliation": "", "summary": "Multimodal representation learning, exemplified by multimodal contrastive learning (MMCL) using image-text pairs, aims to learn powerful representations by aligning cues across modalities", "links": { "openreview": "https://openreview.net/forum?id=3KtPujOw5z", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-675", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "A Unified Solution to Video Fusion: From Multi-Frame Learning to Benchmarking", "authors": [ "Zixiang Zhao", "Haowen Bai", "Bingxin Ke", "Yukun Cui", "Lilun Deng", "Yulun Zhang", "Kai Zhang", "Konrad Schindler" ], "affiliation": "", "summary": "The real world is dynamic, yet most image fusion methods process static frames independently, ignoring temporal correlations in videos and leading to flickering and temporal inconsistency", "links": { "openreview": "https://openreview.net/forum?id=nlQRra0OLH", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-676", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Multi-agent Markov Entanglement", "authors": [ "Shuze Chen", "Tianyi Peng" ], "affiliation": "", "summary": "Value decomposition has long been a fundamental technique in multi-agent reinforcement learning and dynamic programming", "links": { "openreview": "https://openreview.net/forum?id=IupCqXiiOE", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-677", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "SQS: Enhancing Sparse Perception Models via Query-based Splatting in Autonomous Driving", "authors": [ "Haiming Zhang", "Yiyao Zhu", "Wending Zhou", "Xu Yan", "Yingjie CAI", "Bingbing Liu", "Shuguang Cui", "Zhen Li" ], "affiliation": "", "summary": "Sparse Perception Models (SPMs) adopt a query-driven paradigm that forgoes explicit dense BEV or volumetric construction, enabling highly efficient computation and accelerated inference", "links": { "openreview": "https://openreview.net/forum?id=plpAecfkf4", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-678", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding", "authors": [ "Jang Hyun Cho", "Andrea Madotto", "Effrosyni Mavroudi", "Triantafyllos Afouras", "Tushar Nagarajan", "Muhammad Maaz", "Yale Song", "Tengyu Ma", "Shuming Hu", "Suyog Jain", "Miguel Martin", "Huiyu Wang", "Hanoona Abdul Rasheed", "Peize Sun", "Po-Yao Huang", "Daniel Bolya", "Nikhila Ravi", "Shashank Jain", "Tammy Stark", "Seungwhan Moon", "Babak Damavandi", "Vivian Lee", "Andrew Westbury", "Salman Khan", "Philipp Kraehenbuehl", "Piotr Dollar", "Lorenzo Torresani", "Kristen Grauman", "Christoph Feichtenhofer" ], "affiliation": "", "summary": "Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe", "links": { "openreview": "https://openreview.net/forum?id=5NkfjxMpWe", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-679", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models?", "authors": [ "Hyeong Kyu Choi", "Jerry Zhu", "Sharon Li" ], "affiliation": "", "summary": "Multi-Agent Debate (MAD) has emerged as a promising paradigm for improving the performance of large language models through collaborative reasoning", "links": { "openreview": "https://openreview.net/forum?id=iUjGNJzrF1", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-680", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "VoxDet: Rethinking 3D Semantic Scene Completion as Dense Object Detection", "authors": [ "Wuyang Li", "Zhu Yu", "Alexandre Alahi" ], "affiliation": "", "summary": "Semantic Scene Completion (SSC) aims to reconstruct the 3D geometry and semantics of the surrounding environment", "links": { "openreview": "https://openreview.net/forum?id=lMhNrt0Bnm", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-681", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Approximate Domain Unlearning for Vision-Language Models", "authors": [ "Kodai Kawamura", "Yuta Goto", "Rintaro Yanagi", "Hirokatsu Kataoka", "Go Irie" ], "affiliation": "", "summary": "Pre-trained Vision-Language Models (VLMs) exhibit strong generalization capabilities, enabling them to recognize a wide range of objects across diverse domains without additional training", "links": { "openreview": "https://openreview.net/forum?id=lv4zLWzOi2", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-682", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "BioCLIP 2: Emergent Properties from Scaling Hierarchical Contrastive Learning", "authors": [ "Jianyang Gu", "Samuel Stevens", "Elizabeth G Campolongo", "Matthew J Thompson", "Net Zhang", "Jiaman Wu", "Andrei Kopanev", "Zheda Mai", "Alexander E. White", "James Balhoff", "Wasila Dahdul", "Daniel Rubenstein", "Hilmar Lapp", "Tanya Berger-Wolf", "Wei-Lun Chao", "Yu Su" ], "affiliation": "", "summary": "Foundation models trained at scale exhibit remarkable emergent behaviors, learning new capabilities beyond their initial training objectives", "links": { "openreview": "https://openreview.net/forum?id=yPC9zmkQgG", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-683", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "GraphMaster: Automated Graph Synthesis via LLM Agents in Data-Limited Environments", "authors": [ "Enjun Du", "Xunkai Li", "Tian Jin", "Zhihan Zhang", "Rong-Hua Li", "Guoren Wang" ], "affiliation": "", "summary": "The era of foundation models has revolutionized AI research, yet Graph Foundation Models (GFMs) remain constrained by the scarcity of large-scale graph corpora", "links": { "openreview": "https://openreview.net/forum?id=h3dbocj7po", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-684", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Q-Insight: Understanding Image Quality via Visual Reinforcement Learning", "authors": [ "Weiqi Li", "Xuanyu Zhang", "Shijie Zhao", "Yabin ZHANG", "Junlin Li", "Li zhang", "Jian Zhang" ], "affiliation": "", "summary": "Image quality assessment (IQA) focuses on the perceptual visual quality of images, playing a crucial role in downstream tasks such as image reconstruction, compression, and generation", "links": { "openreview": "https://openreview.net/forum?id=Bds54EfR9x", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-685", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "What Makes a Reward Model a Good Teacher? An Optimization Perspective", "authors": [ "Noam Razin", "Zixuan Wang", "Hubert Strauss", "Stanley Wei", "Jason D. Lee", "Sanjeev Arora" ], "affiliation": "", "summary": "The success of Reinforcement Learning from Human Feedback (RLHF) critically depends on the quality of the reward model", "links": { "openreview": "https://openreview.net/forum?id=7pufO0SJAC", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-686", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "$\\Psi$-Sampler: Initial Particle Sampling for SMC-Based Inference-Time Reward Alignment in Score Models", "authors": [ "TaeHoon Yoon", "Yunhong Min", "Kyeongmin Yeo", "Minhyuk Sung" ], "affiliation": "", "summary": "We introduce $\\Psi$-Sampler, an SMC-based framework incorporating pCNL-based initial particle sampling for effective inference-time reward alignment with a score-based model", "links": { "openreview": "https://openreview.net/forum?id=slVqJAI5sT", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-687", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Do-PFN: In-Context Learning for Causal Effect Estimation", "authors": [ "Jake Robertson", "Arik Reuter", "Siyuan Guo", "Noah Hollmann", "Frank Hutter", "Bernhard Schölkopf" ], "affiliation": "", "summary": "Causal effect estimation is critical to a range of scientific disciplines", "links": { "openreview": "https://openreview.net/forum?id=OaNbl9b56B", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-688", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "ReSim: Reliable World Simulation for Autonomous Driving", "authors": [ "Jiazhi Yang", "Kashyap Chitta", "Shenyuan Gao", "Long Chen", "Yuqian Shao", "Xiaosong Jia", "Hongyang Li", "Andreas Geiger", "Xiangyu Yue", "Li Chen" ], "affiliation": "", "summary": "How can we reliably simulate future driving scenarios under a wide range of ego driving behaviors? Recent driving world models, developed exclusively on real-world driving data composed mainly of safe", "links": { "openreview": "https://openreview.net/forum?id=T0CiI4gDFB", "arxiv": "", "detail": "" } }, { "uid": "neurips-2025-spotlight-689", "conference": "NeurIPS", "year": 2025, "track": "Spotlight", "title": "Differentiable Sparsity via $D$-Gating: Simple and Versatile Structured Penalization", "authors": [ "Chris Kolb", "Laetitia Frost", "Bernd Bischl", "David Rügamer" ], "affiliation": "", "summary": "Structured sparsity regularization offers a principled way to compact neural networks, but its non-differentiability breaks compatibility with conventional stochastic gradient descent and requires eit", "links": { "openreview": "https://openreview.net/forum?id=8OGTkEJrmb", "arxiv": "", "detail": "" } } ] }