@article{fightfirewithfiredefendingagainst, title = {Fight Fire with Fire: Defending Against Malicious RL Fine-Tuning via Reward Neutralization}, author = {Wenjun Cao}, year = {2025}, eprint = {2505.04578}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2505.04578v1}, }