reward regularization

language model training (44) instruction following (372) reinforcement learning from human feedback (261) credit assignment (95) token-level reward (8) preference optimization (273)