Before the Model Learns the Bug:Fuzzing RLVR Verifiers 事件

Before the Model Learns the Bug:Fuzzing RLVR Verifiers arXiv:2606.01066v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) replaces human preference labels with executable reward functions such as math answer checkers, JSON tool-call validators, and code unit-test harnesses. That makes the reward partly a software artifact: if the verifier is wrong, optimization can learn the bug. We study this failure mode with a lightweight verifier-fuzzing framework that gen