Safety Guardrails of Large Language Models Are Vulnerable to Value-Driven Adversarial Prompting.

Xiaohao Luo, Ying Wei, Zhijun Li

Browse the full ACL paper archive.