aiqbench analysis
This paper is about ML training data: a corpus built from 2,118 textbooks to supervise language models on mathematical reasoning, with structure like chapter order and proof-dependency links kept intact. The headline 16.7-point gain is self-reported on the authors' own held-out evaluations plus public benchmarks, so it's worth checking against independent replication and the released corpus once available. The more reusable idea is that data organization, not just volume, may matter for reasoning training.
Scaling Mathematical Proof Supervision from Structured Textbooks
arXiv:2610.40394v1 Announce Type: cross Abstract: Mathematical supervision for language models is typically flattened into isolated problem-solution pairs, discarding the curricular organization and evidential structure of the original sources. We introduce TOME-Math (Textbook-Oriented Mathematical Evidence), a large-scale corpus that converts 2,118 textbooks spanning the core areas of university-level mathematics subjects into approximately 380K structured proof-supervision records while preserving chapter order, proposition-proof links, dependency references, and source provenance. Leveraging these structural signals, we explore curriculum learning across multiple mathematical subjects and characterize scaling patterns with supervision volume and source breadth. At full scale, fine-tuning on TOME-Math raises the unweighted mean score across our held-out evaluations and established public benchmarks by 16.7 percentage points over the base model, while exceeding all three public instruction-data baselines evaluated in aggregate. Our findings highlight structured textbooks as both a scalable source of mathematical supervision and an empirical framework for studying the impact of data organization on mathematical reasoning. We will release the corpus to facilitate further research on structured mathematical supervision.
Abstract:Mathematical supervision for language models is typically flattened into isolated problem-solution pairs, discarding the curricular organization and evidential structure of the original sources. We introduce TOME-Math (Textbook-Oriented Mathematical Evidence), a large-scale corpus that converts 2,118 textbooks spanning the core areas of university-level mathematics subjects into approximately 380K structured proof-supervision records while preserving chapter order, proposition-proof links, dependency references, and source provenance. Leveraging these structural signals, we explore curriculum learning across multiple mathematical subjects and characterize scaling patterns with supervision volume and source breadth. At full scale, fine-tuning on TOME-Math raises the unweighted mean score across our held-out evaluations and established public benchmarks by 16.7 percentage points over the base model, while exceeding all three public instruction-data baselines evaluated in aggregate. Our findings highlight structured textbooks as both a scalable source of mathematical supervision and an empirical framework for studying the impact of data organization on mathematical reasoning. We will release the corpus to facilitate further research on structured mathematical supervision.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2610.40394 [cs.CL]
(or arXiv:2610.40394v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.40394 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 256 words
Related coverage
Multiclass Classification without Labels via Posterior Simplex Geometry
arXiv · Jul 29, 2026
Function-Level Execution Feedback for Code Preference Optimization
arXiv · Aug 27, 2026
DIAG: Diagnostic Iterative Alignment and Generation for Data-Efficient Mathematical Preference Distillation
arXiv · Aug 25, 2026
When and Where to Trust the Teacher: Unifying On-Policy Distillation and GRPO through Entropy-Calibrated Credit Assignment
arXiv · Sep 24, 2026