@article{harvehackingawarerewardheadvectorediting, title = {HARVE: Hacking-Aware Reward-Head Vector Editing for Robust Reward Models}, author = {Shuang Liu and Yuxuan Bo and Qiuyang Zhao and Caiyue Huang and Xiaorong Chen and Yanguang Liu and Mengnan Du}, year = {2026}, eprint = {2606.03131}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2606.03131}, }