no code implementations • 7 Apr 2024 • Zhilong Wang, Yebo Cao, Peng Liu
This paper proposes a new type of jailbreak attacks which can deceive both the LLMs and human (i. e., security analyst).
Language Modelling