This paper introduces the EdLight Research Portal (ERP), a publicly available repository of expert annotated handwritten math student work. ERP addresses the gap between formative assessment practices in real mathematics classrooms and the digital datasets commonly used in AI-powered education research. The dataset captures authentic paper-based reasoning from 5th to 9th grade students and includes standardized rubric scores, misconception tags, and contextual metadata generated by trained math educators. ERP enables researchers to study student thinking at scale, examine patterns in mathematical misconceptions, and benchmark machine learning models on tasks that reflect real classroom data. The portal provides a foundation for advancing transparent, pedagogically grounded research in AI applications for mathematics education.
Paper
The full text of this publication is not hosted on 44B due to licensing.
Read it at OpenAlex