Steering Away from Refusal: A Black-box Jailbreak Method Based on First-Token Distribution

Shuangjie Fu, Du Su, Xin Chen, Fei Sun 0002, Huawei Shen, Xueqi Cheng. Steering Away from Refusal: A Black-box Jailbreak Method Based on First-Token Distribution. In Maria Liakata, Viviane P. Moreira, Jiajun Zhang 0001, David Jurgens, editors, Findings of the Association for Computational Linguistics, ACL 2026, San Diego, California, United States, July 2-7, 2026. pages 25969-25979, Association for Computational Linguistics, 2026. [doi]

@inproceedings{FuSCSSC26,
  title = {Steering Away from Refusal: A Black-box Jailbreak Method Based on First-Token Distribution},
  author = {Shuangjie Fu and Du Su and Xin Chen and Fei Sun 0002 and Huawei Shen and Xueqi Cheng},
  year = {2026},
  url = {https://aclanthology.org/2026.findings-acl.1294/},
  researchr = {https://researchr.org/publication/FuSCSSC26},
  cites = {0},
  citedby = {0},
  pages = {25969-25979},
  booktitle = {Findings of the Association for Computational Linguistics, ACL 2026, San Diego, California, United States, July 2-7, 2026},
  editor = {Maria Liakata and Viviane P. Moreira and Jiajun Zhang 0001 and David Jurgens},
  publisher = {Association for Computational Linguistics},
  isbn = {979-8-89176-395-1},
}