Skip to content
arXiv cs.LG · Papers

When RLHF Fails: A Mechanistic Taxonomy of Reward Hacking, Collapse, and Evaluator Gaming

arXiv:2606.03238v2 Announce Type: replace Abstract: RLHF evaluation should track how failures emerge, where they localize, and which warning signals appear before external quality degrades. We study this problem with a compact RLHF pipeline built for this paper, including PPO, DPO, uncertainty-penalized PPO (UP-PPO), r