Mechanistic-Anomaly-Detection/llama3-sandwich-backdoor-dataset
收藏Hugging Face2024-10-23 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/Mechanistic-Anomaly-Detection/llama3-sandwich-backdoor-dataset
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含两个主要特征：prompt和completion，均为字符串类型。数据集被分为多个分割，包括正常良性训练集、正常有害训练集、后门训练集、正常良性测试集、正常有害测试集和后门测试集。每个分割都有对应的文件大小和示例数量。数据集的下载大小为97658421字节，总大小为171893732.6372876字节。
This dataset is primarily used for training and testing models, containing normal and harmful prompts along with their corresponding completions. The dataset is divided into training and test sets, used for training and testing in normal, harmful, and backdoored scenarios.
提供机构：
Mechanistic-Anomaly-Detection
