1 paper · 1 filter
Lynn Delcon, Andres Algaba, Vincent Ginis
Perturbation techniques that turn unsuccessful jailbreak prompts into successful ones are continuously evolving, constituting a major security threat to LLM safety. In this paper,…