Adversarial poetry as a universal single-turn jailbreak mechanism in LLMs
📰 Hacker News · capgre
Adversarial poetry as a universal single-turn jailbreak mechanism in LLMs. 189 comments, 384 points on Hacker News.
Adversarial poetry as a universal single-turn jailbreak mechanism in LLMs. 189 comments, 384 points on Hacker News.