UniBreak: A Unified Evolutionary Token-Level Jailbreaking Framework for Large Language Models

Abstract

Large language models (LLMs) demonstrate promising capabilities in natural language understanding and reasoning with enormous parameter spaces and vast amount of training data. These attributes have facilitated their deployment into diverse application domains. However, the underlying parameters implicitly assume decision-making boundaries, resulting in significant decision space regions not covered by training data. It makes them susceptible to adversarial manipulations through carefully crafte

Categories

Framework mappings

Suggested from the entry's categories.

@article{you2026unibreak,
  title = {{UniBreak: A Unified Evolutionary Token-Level Jailbreaking Framework for Large Language Models}},
  author = {Shen You and Wei Jiang and Hefei Mei and Danei Gong and Zhongshen Li and Jixang Yu and Jun-Kai Ji and Qiuzhen Lin and Xiangtao Li and Ka-chun Wong},
  year = {2026},
  month = aug,
  journal = {IEEE Transactions on Evolutionary Computation},
  doi = {10.1109/TEVC.2026.3656951},
  url = {https://www.semanticscholar.org/paper/9b896fafa251bd1edf23469a7b625b9dbf768770}
}