4 ms·Revealing a Hidden Class of Task-in-Prompt Adversarial Attacks on LLMs2 points by aghuang 28d agolearning_to_shi 28d ago[dead]