New Rubric Assesses VLM Adaptivity in Math Education
A research paper on arXiv proposes a learner model-based rubric to evaluate whether vision language models (VLMs) can adapt to different learner profiles in mathematics education. The rubric, drawing on the adaptive learning framework by Shute and Towle (2018), formalizes adaptivity assessment into three aspects: cognitive, motivational, and unspecified others. The study addresses the gap that current VLMs lack a systematic evaluation framework for adaptivity in math tutoring tasks. The paper is identified as arXiv:2605.16011, announced as a cross submission. It highlights the growing use of VLMs as personalized learning aids but questions their ability to adapt instruction based on individual learning performance.
Key facts
- Paper arXiv:2605.16011 proposes a rubric for VLM adaptivity in math education.
- Rubric based on adaptive learning framework by Shute and Towle (2018).
- Adaptivity assessment covers cognitive and motivational aspects.
- VLMs are increasingly used as personalized learning aids in mathematics.
- Current VLMs lack systematic evaluation for adaptivity to learner profiles.
- The study aims to address the gap in evaluating VLM adaptivity.
- Announcement type is cross.
- The paper is from arXiv.
Entities
Institutions
- arXiv