DocBank
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/doc-analysis/docbank
下载链接
链接失效反馈官方服务：
资源简介：
该数据集由 500,000 份文档组成，每份文档具有独特的版式，并且每份文档仅包含一页。该数据集已被充分利用，以提升训练效率。从规模上看，它涵盖了 500,000 份文档，其主要任务集中在文档理解上。
This dataset consists of 500,000 documents, each with a unique layout and containing only one page. It has been fully leveraged to enhance training efficiency. In terms of scale, it covers 500,000 documents, and its core tasks focus on document understanding.
提供机构：
Open-source
