BriefGPT.xyz
May, 2024
语义引导的通用目标劫持LLM的提示组织
Semantic-guided Prompt Organization for Universal Goal Hijacking against LLMs
HTML
PDF
Yihao Huang, Chong Wang, Xiaojun Jia, Qing Guo, Felix Juefei-Xu...
TL;DR
通过语义导向的提示处理策略，我们提出了一种称为POUGH的通用目标劫持方法，用于评估大型语言模型的可信度，并在四个流行的语言模型和十种目标响应类型上进行了实验证实其有效性。
Abstract
With the rising popularity of
large language models
(LLMs), assessing their
trustworthiness
through security tasks has gained critical importance. Regarding the new task of
→