Anirudhan Badrinath

Hybrid Preference Optimization: Augmenting Direct Preference Optimization with Auxiliary Objectives

Discovering High-Quality Chess Puzzles Through One Billion Plays with Offline Reinforcement Learning