LLMDoc

该数据集是为了比较在文档领域中的指令理解性能而构建的，它通过人工评估形成了一个评估集合。此外，该数据集还旨在评估mPLUG-DocOwl在多种文档场景下的指令理解能力。该数据集包含了100个测试样本，任务聚焦于指令理解和文档理解。

This dataset is constructed for comparing instruction understanding performance within the document domain, and forms an evaluation set through human evaluation. Moreover, it is designed to evaluate the instruction understanding capability of mPLUG-DocOwl across diverse document scenarios. The dataset contains 100 test samples, with tasks focusing on both instruction understanding and document understanding.