Reproduce “MMTutorBench: The First Multimodal Benchmark for AI Math Tutoring” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tengchao Yang, Sichen Guo, Mengzhao Jia, Jiaming Su, Yuanyang Liu, Zhihan Zhang, Meng Jiang
The authors introduce MMTutorBench, a benchmark for evaluating stepwise math tutoring from handwritten work and student questions. It contains 770 problems. In image-only, zero-shot evaluation, Gemini-2.5-Pro scores 4.69, compared with 3.40 for the top open-source model; human experts score 5.85 on a 66-sample subset, not a matched full-benchmark comparison. The benchmark tests guidance, not learning gains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.