WikiText-103 and AI4Privacy Fusion
收藏arXiv2025-09-30 收录
下载链接：
https://huggingface.co/datasets/ai4privacy/pii-masking-200k
下载链接
链接失效反馈官方服务：
资源简介：
该数据集融合了WikiText-103和AI4Privacy数据集，旨在实验性地提取隐私信息。它包含了将真实生活中的个人识别信息（PII）注入现有数据集中的样本，规模为在WikiText-103中注入了1,000个样本。该数据集的任务是评估针对提取的隐私信息，如地址、电话号码和密码等的攻击。
This dataset is a fusion of the WikiText-103 and AI4Privacy datasets, intended for experimental privacy information extraction. It includes samples created by injecting real-world personally identifiable information (PII) into the source datasets, with 1,000 such PII-injected samples prepared for the WikiText-103 portion. The task of this dataset is to evaluate attacks targeting extracted privacy information, such as addresses, telephone numbers, passwords, and other similar personal data.
提供机构：
Hugging Face
