eth-nlped/stepverify

这是一个包含1000个学生数学解题步骤错误和注释的数据集，旨在帮助大型语言模型检测和针对学生错误提供反馈。数据集由教师注释，包括问题、主题、参考解决方案、错误解决方案、错误步骤、错误类别和描述、对话历史以及学生正确响应。

This dataset contains 1,000 student math problem-solving step errors and annotations, designed to assist large language models in detecting and providing feedback targeted at student errors. The dataset is annotated by teachers and includes the problem, topic, reference solution, incorrect solution, error step, error category, error description, dialogue history, and student correct response.