Pseudo Ground-Truth Forget Corpus
收藏arXiv2025-09-30 收录
下载链接:
http://rwku-bench.github.io
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了从维基百科页面获取的每个遗忘目标的文本描述,形成了一个语料库。此外,该语料库被用作与合成遗忘语料库进行比较,以评估不同的遗忘学习方法。这项任务旨在从大型语言模型中遗忘知识。
This dataset comprises textual descriptions of each unlearning target extracted from Wikipedia pages, constituting a dedicated corpus. Furthermore, this corpus is employed for comparative evaluation against a synthetic unlearning corpus to assess diverse unlearning methods. This task focuses on achieving knowledge unlearning from large language models.
提供机构:
RWKU Benchmark Team



