@article{whenhumanpreferencesflipaninstancedepend, title = {When Human Preferences Flip: An Instance-Dependent Robust Loss for RLHF}, author = {Yifan Xu and Xichen Ye and Yifan Chen and Qiaosheng Zhang}, year = {2025}, eprint = {2512.00709}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2512.00709}, }