Less is More: Improving LLM Alignment via Preference Data Selection
NeurIPSSpotlight2025
TL;DR
Direct Preference Optimization (DPO) has emerged as a promising approach for aligning large language models with human preferences…
Opening excerpt from the authors’ abstract. source
Read the paper
Topics
preference optimization large language model language model optimization alignment llm
← All NeurIPS 2025 Spotlight papers · Browse the whole archive