SKU SFT-2202 · Sold by External
NuminaMath-1.5
Product specifications
| SKU | SFT-2202 |
|---|---|
| Data type | SFT dataset |
| Volume | 896k problems |
| Size on disk | 531 MB |
| Format | Parquet (Hugging Face) |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | Apache-2.0 |
NuminaMath-1.5 is Project Numina / AI-MO's flagship competition-math SFT corpus of ~896k problem-solution pairs, sourced from Chinese high-school exercises up to international olympiad problems and formatted as chain-of-thought (CoT) solutions. It is the improved, cleaned successor to the earlier NuminaMath-CoT (~859k pairs); both are text CoT. Metadata flags problem_type, question_type (proof/MCQ/math-word-problem), source and a synthetic indicator. A separate NuminaMath-TIR (~70k) provides tool-integrated (code) reasoning and is not part of this listing.