Revealing a Hidden Class of Task-in-Prompt Adversarial Attacks on LLMs

2 points | by aghuang 10 hours ago