On the Generalization of SFT: A Reinforcement Learning Perspective with Reward Rectification

OffRLLRMRe-assign community