Dan-Qi Chen

Where RL Breaks: Localizing and Repairing Failures via Targeted Continued Pretraining

Reinforcement learning with verifiable rewards (RLVR) is widely used to improve reasoning in language models, yet growing evidence suggests that it mostly sharpens what the base model can already do. This raises a practical question: when RL fails, what is the base model missing? We look inside the RL policy's own roll...