metadata
license: cc-by-nc-4.0
size_categories:
- 10K<n<100K
language:
- en
English multi document Q&A data created using RedPajamaCommonCrawl data as reference text
Each sample contains one reference document, 199 irrelevant documents, and a Q-A pair based on the reference document. It can be used to train models to extract the target information from a large number of documents.
以RedPajamaCommonCrawl数据为参考文本，制作的英文多文档问答数据
每个样本包含 一个参考文档、199个无关文档、一个基于参考文档的问答对。可以训练模型从大量文档中抽取关键信息的能力。
dataset size: about 10k