Jailbreak Antidote: Runtime Safety-Utility Balance via Sparse Representation Adjustment in Large Language Models

@misc{he2025jailbreakantidoteruntime,
      title={Jailbreak Antidote: Runtime Safety-Utility Balance via Sparse Representation Adjustment in Large Language Models}, 
      author={Xiang He and Yi Zeng and Guobin Shen and Dongcheng Zhao and Yiting Dong},
      year={2025},
      eprint={2410.02298},
      archivePrefix={arXiv},
      primaryClass={cs.CR},
      url={https://arxiv.org/abs/2410.02298}, 
}