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

Devs

InfoRM: Mitigating Reward Hacking in RLHF via Information-Theoretic Reward Modeling | Read Paper on Bytez