Solus-PG/adversarial_vs_benign_reasoning_v1
收藏Hugging Face2025-11-09 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/Solus-PG/adversarial_vs_benign_reasoning_v1
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了模型在不同层的输出，以及与输出相关的多个特征，如标签、提示信息、响应文本等。标签分为对抗性和良性两种类型，等级分为良性、显式和细微三个等级。数据集分为训练集，提供了相关的配置信息。
The dataset contains model outputs at different layers, along with multiple features related to the outputs, such as labels, prompt information, response text, etc. The labels are divided into adversarial and benign types, and the levels are divided into three categories: benign, explicit, and subtle. The dataset is split into a training set and provides related configuration information.
提供机构：
Solus-PG
