Fetching the paper…

Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents · Around