Standards-Targeted Educational Math dataset
收藏资源简介:
EDUMATH数据集是首个针对教育数学问题生成的标准对齐数据集,由教师标注,包含超过11,000个数学问题,用于训练大语言模型生成符合学生兴趣和能力水平的数学教育问题。该数据集由弗吉尼亚大学的研究人员开发,旨在解决教师在定制数学问题时面临的时间和资源限制问题。数据集内容涵盖了3至5年级的数学标准,并为每个问题提供了自然语言的解决方案。该数据集的创建过程包括对现有数学问题数据集的标注、LLM生成的数学问题的评估和筛选,以及最终的数据集构建。数据集的应用领域为K-12数学教育,旨在帮助学生更好地理解和掌握数学概念。
The EDUMATH dataset is the first standard-aligned dataset dedicated to educational math problem generation. Annotated by teachers, it contains over 11,000 math problems, and is designed to train large language models (LLMs) to generate math educational problems tailored to students' interests and proficiency levels. Developed by researchers at the University of Virginia, this dataset aims to address the time and resource constraints faced by teachers when creating customized math problems. Its content covers math curriculum standards for grades 3 through 5, and provides natural language solutions for each problem. The dataset creation process includes annotating existing math problem datasets, evaluating and filtering math problems generated by LLMs, and constructing the final dataset. The dataset is applied in K-12 math education, with the goal of helping students better understand and master mathematical concepts.




