@article{guardroleplayingtogeneratenatural,
  title = {GUARD: Role-playing to Generate Natural-language Jailbreakings to Test Guideline Adherence of Large Language Models},
  author = {Haibo Jin and Ruoxi Chen and Andy Zhou and Yang Zhang and Haohan Wang},
  year = {2024},
  eprint = {2402.03299},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2402.03299v4},
}