HumorReject: Decoupling LLM Safety from Refusal Prefix via A Little Humor

@misc{liu2025humorrejectdecouplingllm,
      title={HumorReject: Decoupling LLM Safety from Refusal Prefix via A Little Humor}, 
      author={Zhaoxiang Liu and Zihui Wu and Haichang Gao and Jiacheng Luo},
      year={2025},
      eprint={2501.13677},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2501.13677}, 
}