SKU SFT-2203 · Sold by External
MetaMathQA
Product specifications
| SKU | SFT-2203 |
|---|---|
| Data type | SFT dataset |
| Volume | 395k QA pairs |
| Size on disk | 396 MB |
| Format | Parquet (Hugging Face) |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | MIT |
MetaMathQA is a 395k-row math SFT dataset that bootstraps new questions from the GSM8K and MATH training sets using eight augmentation strategies (answer augmentation, rephrasing, self-verification, and FOBAR backward reasoning). All items are augmented from training splits only (no test leakage). Solutions are text reasoning responses; there is no code or tool-integrated reasoning.