QA tasks
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/diana
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8个问答数据集，覆盖了3种格式：抽取式问答、抽象式问答和多项选择式问答。此外，模型在8个任务上进行训练，并在总共11个任务上进行测试，其中包括3个未见过的任务。规模上，涵盖了8个已见任务和3个未见任务。这一评估任务旨在评估终身学习的能力。
This dataset comprises 8 question answering (QA) datasets, covering three distinct formats: extractive QA, abstractive QA, and multiple-choice QA. Furthermore, the model is trained on 8 tasks and evaluated across a total of 11 tasks, including 3 unseen ones. In terms of scale, it covers 8 seen tasks and 3 unseen tasks. This evaluation benchmark is designed to assess the lifelong learning capabilities of models.
