Saiyong Yang

Thinking-Free Policy Initialization Makes Distilled Reasoning Models More Effective and Efficient Reasoners

LaSeR: Reinforcement Learning with Last-Token Self-Rewarding