RuleAdapter: Dynamic Rules for training Safety Reward Models in RLHF.

Xiaomin Li, Mingye Gao, Zhiwei Zhang, Jingxuan Fan, Weiyu Li

Browse the full ICML paper archive.