CMM-Math
收藏资源简介:
CMM-Math是由华东师范大学创建的中文多模态数学数据集,旨在评估和提升大型多模态模型在数学推理方面的能力。该数据集包含超过28,000个高质量样本,涵盖从小学到高中的12个年级,涉及多种题型和详细的解答。数据集的创建过程包括基础预训练、基础微调和数学微调三个阶段,确保了数据集的高质量和多样性。CMM-Math主要应用于教育领域,旨在解决数学推理任务中的多模态问题,提升模型的数学推理能力。
CMM-Math is a Chinese multimodal mathematics dataset created by East China Normal University, aiming to evaluate and enhance the mathematical reasoning capabilities of large multimodal models. This dataset contains over 28,000 high-quality samples, covering 12 grade levels from primary school to senior high school, involving various question types and detailed solutions. The construction process of the dataset includes three stages: basic pre-training, basic fine-tuning, and mathematical fine-tuning, which ensures the high quality and diversity of the dataset. CMM-Math is mainly applied in the field of education, aiming to solve multimodal problems in mathematical reasoning tasks and improve the mathematical reasoning abilities of models.




