论文阅读：arxiv 2025 Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language

总目录 大模型相关研究：https://blog.youkuaiyun.com/WhiffeYF/article/details/142132328

Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models

https://arxiv.org/pdf/2511.15304

https://www.doubao.com/chat/32148934692605954

AI安全新漏洞：一首诗就能攻破顶级大模型？

对抗性诗歌与 AI 单轮越狱的降咒实验

论文翻译：