Bayesian Reward Models for LLM Alignment

arXiv:2402.13210v1 Announce Type: new
Abstract: To ensure that large language model (LLM) responses are helpful and non-toxic, we usually fine-tune a reward model on human preference data. We then select policy responses with high rewards (best-of-n sampling) or further optimize the policy to produce responses with high rewards (reinforcement learning from human feedback). However, this process is vulnerable to reward overoptimization or hacking, in which the responses selected have high rewards due to errors in the reward model rather than …

abstract alignment arxiv bayesian cs.lg data feedback human human feedback language language model large language large language model llm policy reinforcement reinforcement learning responses reward model sampling type