SKU SFT-2201 · Sold by External
OpenR1-Math-220k
Product specifications
| SKU | SFT-2201 |
|---|---|
| Data type | SFT dataset |
| Volume | 220k problems |
| Size on disk | Not published by source |
| Format | Parquet (Hugging Face) |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | Apache-2.0 |
OpenR1-Math-220k is the Open-R1 project's math reasoning SFT corpus: ~220k problems drawn from NuminaMath-1.5, each answered with two to four step-by-step reasoning traces generated by DeepSeek-R1 and filtered with Math-Verify plus a Llama-3.3-70B judge. Solutions are pure text chain-of-thought (final answer in \boxed{}); there is no tool-integrated or program-of-thought code. It ships 'default' (~94k, best SFT performance) and 'extended' (~131k, includes cn_k12) configs; the larger, unfiltered OpenR1-Math-Raw (~500k+ generations) is a separate companion dataset.