Reproduce “Scaling Mathematical Proof Supervision from Structured Textbooks” 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
Youqing Fang, Zhouqi Hua, Yinhao Tang, Yuzhe Gu, Lingkai Kong, Bin Liu, Weiming Zhang, Kai Chen and 2 more
TOME-Math organizes mathematics textbook propositions, proofs, dependencies, and chapter order into Proof QA and Gap QA records, making implicit proof steps explicit for retrieval and training curricula. Fine-tuning Qwen3-30B-A3B-Base on the corpus reaches a 45.83% unweighted mean across six benchmarks, versus 29.17% for the base model; differing corpora and token budgets prevent attributing this gain to structure alone.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.