MultiMath-300K
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/pengshuai-rin/MultiMath
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为MultiMath-300K，包含298,670个数学问题，其中训练集有290,227个问题，验证集有8,443个问题，覆盖了从幼儿园到高中12年级的所有教育水平。每个问题都附有图像描述和分步解答。此外，数据集还包含了视觉与语言对齐数据以及逐步推理解决方案，确保了数据的完整性、多模态性和清晰性。该数据集的规模为298,670个问题，任务重点在于多模态数学推理。
The dataset, named MultiMath-300K, consists of 298,670 mathematical problems, among which 290,227 are allocated to the training set and 8,443 to the validation set. It covers all educational levels ranging from kindergarten to 12th grade. Each problem is paired with image descriptions and step-by-step solutions. Furthermore, the dataset contains visual-language alignment data and step-by-step reasoning solutions, which guarantee the data's completeness, multimodality and clarity. With a total of 298,670 problems, this dataset focuses on the task of multimodal mathematical reasoning.
