大模型对齐与安全

Response Attack: Exploiting Contextual Priming to Jailbreak Large Language Models

响应攻击：利用上下文优先级来劫持大语言模型

Ziqi Miao, Lijun Li, Yuan Xiong, Zhenhua Liu, Pengyu Zhu, Jing Shao

机构 * Shanghai Artificial Intelligence Laboratory（上海人工智能实验室） ； Xi’an Jiaotong University（西安交通大学） ； Beijing University of Posts and Telecommunications（北京邮电大学） 纠错

专题命中 越狱攻击 ：jailbreak(title,abstract)；分类 cs.CL

AI总结 响应攻击通过利用对话中中间响应作为上下文优先级，有效劫持大语言模型生成违反政策的内容。

Comments 20 pages, 10 figures. Code and data available at https://github.com/Dtc7w3PQ/Response-Attack

Reason2Attack: Jailbreaking Text-to-Image Models via LLM Reasoning

通过LLM推理进行文本到图像模型的劫持：Reason2Attack

Chenyu Zhang, Lanjun Wang, Yiwen Ma, Wenhui Li, An-An Liu

专题命中 越狱攻击 ：safety(abstract)；分类 cs.AI

AI总结 Reason2Attack通过将劫持攻击融入LLM后训练过程，提升生成对抗性提示的推理能力，实现更高效的文本到图像模型攻击

Comments Noted that This paper includes model-generated content that may contain offensive or distressing material