MathEDU: Towards Adaptive Feedback for Student Mathematical Problem-Solving

  • 2025-05-23 16:59:39
  • Wei-Ling Hsu, Yu-Chien Tang, An-Zi Yen
  • 0

Abstract

Online learning enhances educational accessibility, offering students theflexibility to learn anytime, anywhere. However, a key limitation is the lackof immediate, personalized feedback, particularly in helping students correcterrors in math problem-solving. Several studies have investigated theapplications of large language models (LLMs) in educational contexts. In thispaper, we explore the capabilities of LLMs to assess students' mathproblem-solving processes and provide adaptive feedback. The MathEDU dataset isintroduced, comprising authentic student solutions annotated with teacherfeedback. We evaluate the model's ability to support personalized learning intwo scenarios: one where the model has access to students' prior answerhistories, and another simulating a cold-start context. Experimental resultsshow that the fine-tuned model performs well in identifying correctness.However, the model still faces challenges in generating detailed feedback forpedagogical purposes.

 

Quick Read (beta)

loading the full paper ...