AI Hub
All papers
Reinforcement LearningTraining

Direct Preference Optimization: Your Language Model is Secretly a Reward Model

Stanford·May 29, 2023

Rafael Rafailov, Archit Sharma, Eric Mitchell

View on arXiv

TL;DR

Derives a simple supervised loss that optimizes a model directly on preference data, matching RLHF without a separate reward model or RL loop.

Why it matters

DPO made preference tuning simpler and more stable, and quickly became a default alignment method for open models.

Related terms