FineMath
Hugging FaceClassifier-filtered mathematical web corpus from Hugging Face, extracted from Common Crawl and released at graded quality tiers holding tens of billions of tokens. It is a documented pretraining component of SmolLM3, Apertus, Marin and the Dolma 3 pool behind OLMo 3.
Verified 2026-08-13 via the HuggingFaceTB/finemath dataset card on Hugging Face.
Openness
5 high confidence- license
- ODC-BY
- gated
- false
- dataset_card
- present
ODC-BY, ungated, full dataset card.
- https://huggingface.co/datasets/HuggingFaceTB/finemath recorded 2026-08-13
ODC-BY license, ungated, dataset card
Adoption
3 high confidence29,279 downloads in the trailing 30 days for HuggingFaceTB/finemath, which bands at 10K-100K, level 3 on the dataset adoption scale.
- https://huggingface.co/api/datasets/HuggingFaceTB/finemath recorded 2026-08-13
29,279 downloads in the trailing 30 days for HuggingFaceTB/finemath
Capability
5 high confidenceLeads controlled math ablations (FineMath-4+ ~2x GSM8K and ~6x MATH over InfiMM-WebMath, beats OpenWebMath); current 2025 default in SmolLM3, Apertus, Marin, and OLMo 3.
- https://huggingface.co/datasets/HuggingFaceTB/finemath recorded 2026-08-13
FineMath-4+ achieves ~2x GSM8K and ~6x MATH over InfiMM-WebMath and beats OpenWebMath
- https://arxiv.org/html/2502.02737v1 recorded 2026-08-13
SmolLM2 paper details the FineMath ablation vs OpenWebMath/InfiMM-WebMath
Verified 2026-08-13