Gradient Cuff: Detecting Jailbreak Attacks on Large Language Models by Exploring Refusal Loss Landscapes

Authors:
  • XHXiaomeng Hu
    Xiaomeng Hu
  • PCPin-Yu Chen
    Pin-Yu Chen
  • THTsung-Yi Ho
    Tsung-Yi Ho

Abstract
The world's premier source for conference proceedings, offering Print-on-Demand, DOI, and Content Hosting services.