Fetching the paper…

InfoRM: Mitigating Reward Hacking in RLHF via Information-Theoretic Reward Modeling · Around