Gradient Cuff: Detecting Jailbreak Attacks on Large Language Models by Exploring Refusal Loss Landscapes
Authors:
- Xiaomeng HuXHXiaomeng Hu
- Pin-Yu ChenPCPin-Yu Chen
- Tsung-Yi HoTHTsung-Yi Ho
- 1 January 2024
- conference paper
- Published by Curran Associates, Inc.
- p. 126265-126296