Wish-QA, Wish-Summarization, Wish-IE

Genie数据集是由IBM以色列研究实验室和耶路撒冷希伯来大学联合开发的，旨在解决内容相关生成任务中高质量数据缺乏的问题。该数据集通过自动化的三步流程（内容准备、生成和过滤）创建，包括了长形式问答、摘要和信息提取等多个任务的数据。数据集内容丰富，覆盖多种主题，通过人工评估显示其生成的数据自然、高质量且忠实于原始内容。此外，该数据集在医学领域的应用表明，其生成的数据与人类生成的数据相比具有竞争力，甚至在某些方面表现更优。

The Genie dataset was jointly developed by IBM Research Israel and the Hebrew University of Jerusalem to address the shortage of high-quality data for content-related generation tasks. Constructed via an automated three-step workflow (content preparation, generation, and filtering), the dataset encompasses data for multiple tasks including long-form question answering, summarization, and information extraction. Featuring rich content spanning diverse topics, the dataset’s generated samples have been verified via human evaluation to be natural, high-quality, and faithful to the original content. Furthermore, applications of the dataset in the medical field have demonstrated that its generated data is competitive with, and even outperforms, human-generated data in certain aspects.