Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents.

Qiusi Zhan, Richard Fang, Henil Shalin Panchal, Daniel Kang

Browse the full NAACL paper archive.