Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach

NeurIPSSpotlight2025

Authors: Jonas Geiping, Sean Michael McLeish, Neel Jain, John Kirchenbauer, Siddharth Singh, Brian R. Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein
Venue: NeurIPS 2025
Track: Spotlight

TL;DR

We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space…

Opening excerpt from the authors’ abstract. source

Read the paper

OpenReview

Topics

language model reasoning

← All NeurIPS 2025 Spotlight papers · Browse the whole archive