Mitigating Reward Hacking in RLHF via Bayesian Non-negative Reward Modeling | Signal Canvas

Mitigating Reward Hacking in RLHF via Bayesian Non-negative Reward Modeling | Signal Canvas | ScienceToStartup