Harness-agnostic detection and immunization of reward hacking in self-evolving language models
This work introduces HackProbe, a monitor that attaches to an arbitrary self-evolving loop through two black-box hooks, with no access to weights or activations, and proves a detectability bound that converts a target error rate into an explicit probe-size budget, and delimit what probe rotation does and does not buy.
Rong-Xin Yang, Yang Liu, Shang Luo et al.
· 2 citations