GuardrailsAI/detect-jailbreak
收藏Hugging Face2025-02-12 更新2025-02-15 收录
下载链接:
https://hf-mirror.com/datasets/GuardrailsAI/detect-jailbreak
下载链接
链接失效反馈官方服务:
资源简介:
本数据集是一个评估大型语言模型(LLM)漏洞和防御有效性的评价数据集,它结合了三个现有的数据集(Verazuo数据集、ZHX123基准和WMDP数据集)并进行了自定义增强。数据集包含了可能令人不适的不安全模型响应和用户查询内容。
This dataset is an evaluation dataset for assessing vulnerabilities and defense effectiveness of Large Language Models (LLMs). It combines three existing datasets (the Verazuo dataset, the ZHX123 benchmark, and the Weapons of Mass Destruction Proxy (WMDP) dataset) with custom augmentations. The dataset includes potentially disturbing unsafe model responses and user queries.
提供机构:
GuardrailsAI



