collinear-ai/jailbreak-prompts-all-judgements-sdaia

该数据集包含了用于生成SDAIA报告的英文和阿拉伯文提示语的响应及安全性判断数据。数据集分为三个子集，其中report_1子集包含第一份报告的数据，report_2_7b和report_2_34b子集包含第二份报告（进行中）的数据。每个子集都详细记录了提示语的ID、来源、攻击类别、危害类别、子类别以及两个模型（allam_7b和llama2_7b，或allam_34b和falcon_h1_34b）对提示语的响应和安全性判断。

This dataset includes the responses and safety judgements in both English and Arabic of the prompts used for generating SDAIA reports. The dataset is divided into three subsets, with report_1 containing data for the first report, and report_2_7b and report_2_34b containing data for the second report (in progress). Each subset records detailed information such as prompt ID, source, attack category, harm category, subcategory, and the responses and safety judgements from two models (allam_7b and llama2_7b, or allam_34b and falcon_h1_34b) for the prompts.