论文链接： https://arxiv.org/pdf/2309.01446.pdf Open sesame! universal black box jailbreaking of large language models Abstract1 Introduction2 Previous Work3 Threat Model4 Our Method4.1 Genetic Algorith