@article{
ono2026mitigating,
title={Mitigating Reward Hacking in Reinforcement Learning from Human Feedback via Advantage Sign Robustness},
author={Shinnosuke Ono and Johannes Ackermann and Soichiro Nishimori and Takashi Ishida and Masashi Sugiyama},
journal={Transactions on Machine Learning Research},
issn={2835-8856},
year={2026},
url={https://openreview.net/forum?id=yu3lbJi5sh},
note={}
}