Before the Model Learns the Bug:Fuzzing RLVR Verifiers
Abstract
Reinforcement learning with verifiable rewards can exploit buggy verifiers, so a fuzzing framework measures verifier errors through adversarial completions and paired decision metrics.
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 generates adversarial completions, compares buggy and stricter reference verifiers, logs paired decisions, and reports false-positive, false-negative, disagreement, exploit, and uncertainty metrics.
Get this paper in your agent:
hf papers read 2606.01066 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 1
Collections including this paper 0
No Collection including this paper