Countering Reward Over-optimization in LLM with Demonstration-Guided Reinforcement Learning

Re-assign community