OCR-Reasoning
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/SCUT-DLVCLab/OCR-Reasoning
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个全面性的基准测试，旨在系统地评估多模态大型语言模型在富含文本的图像推理任务上的表现。它包含了1,069个人工标注的示例，覆盖了6项核心推理能力和18个实际推理任务。此外，该基准测试不仅评估模型生成的最终答案，还评估其推理过程。该数据集的规模为1,069个示例，任务重点在于文本丰富的图像推理。
This dataset is a comprehensive benchmark designed to systematically evaluate the performance of multimodal large language models on text-rich visual reasoning tasks. It contains 1,069 manually annotated examples, covering 6 core reasoning capabilities and 18 real-world reasoning tasks. Furthermore, this benchmark not only evaluates the final answers generated by the models but also assesses their reasoning processes. The dataset consists of 1,069 examples, with the tasks focusing on text-rich visual reasoning.
