ERRORRADAR
收藏资源简介:
ERRORRADAR数据集由松鼠AI和香港科技大学等机构联合创建,旨在评估多模态大语言模型在复杂数学推理中的错误检测能力。该数据集包含2500个高质量的多模态K-12数学问题,来源于真实的学生互动数据,经过严格的手动标注和丰富的元数据注释。数据集的创建过程包括从教育组织中收集问题,并通过专业注释者进行详细标注。ERRORRADAR主要应用于教育领域,旨在解决多模态数学推理中的错误检测问题,提升模型的复杂推理能力。
The ERRORRADAR dataset was jointly created by institutions including Squirrel AI and The Hong Kong University of Science and Technology, aiming to evaluate the error detection capabilities of multimodal large language models in complex mathematical reasoning. This dataset contains 2,500 high-quality multimodal K-12 mathematics problems sourced from real student interaction data, with rigorous manual annotations and rich metadata. The dataset construction process includes collecting problems from educational organizations and performing detailed annotations by professional annotators. ERRORRADAR is mainly applied in the education field, aiming to address the error detection challenge in multimodal mathematical reasoning and enhance the complex reasoning capabilities of models.




