Reward Inside the Model: A Lightweight Hidden-State Reward Model for LLM's Best-of-N sampling

Re-assign community