# MetaMathQA by MetaMath: AI Training Dataset
Source: https://savrn.com/datasets/metamathqa
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

## Dataset Card

By MetaMath, published under mit, revision aa4f34d3d2d3.

View the project page: https://meta-math.github.io/ see our paper at https://arxiv.org/abs/2309.12284

### Note

All MetaMathQA data are augmented from the training sets of GSM8K and MATH. None of the augmented data is from the testing set.

You can check the original_question in meta-math/MetaMathQA, each item is from the GSM8K or MATH train set.

### Model Details

MetaMath-Mistral-7B is fully fine-tuned on the MetaMathQA datasets and based on the powerful Mistral-7B model. It is glad to see using MetaMathQA datasets and changing the base model from llama-2-7B to Mistral-7b can boost the GSM8K performance from 66.5 to 77.7.

To fine-tune Mistral-7B, I would suggest using a smaller learning rate (usually 1/5 to 1/10 of the lr for LlaMa-2-7B) and staying other training args unchanged. More training details and scripts can be seen at [https://github.com/meta-math/MetaMath](https://github.com/meta-math/MetaMath).

### Installation

[Read the full dataset card (427 words)](https://savrn.com/datasets/metamathqa/card)

## Structure

### default 395,000 rows

| Split | Rows | Size |
| --- | --- | --- |
| train | 395,000 | 369.5 MB |

typestringquerystringoriginal_questionstringresponsestring

## Details

Repository

meta-math/MetaMathQA

Publisher

MetaMath

Task category

Not stated by the source

Tags

math, math-qa

Size category

Not stated by the source

Languages

Not stated by the source

Revision

aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18

Last updated

2023-12-21

## Files

3 files, 395.6 MB in total.

Data1 file · 395.6 MB

Documentation1 file · 4.4 KB

Repository1 file · 2.5 KB

Every file

| File | Type | Size | SHA-256 |
| --- | --- | --- | --- |
| MetaMathQA-395K.json | Data | 395.6 MB | fb39a5d8c05c |
| README.md | Documentation | 4.4 KB | — |
| .gitattributes | Repository | 2.5 KB | — |

## License and Download

License

mit

Access

No access gate

[Download from MetaMath](https://huggingface.co/datasets/meta-math/MetaMathQA)

Released by MetaMath through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit).

## Models Trained on This Dataset

- Trained on (disclosed)[Symbiotic-Beta](https://savrn.com/models/symbiotic-beta)

## MetaMath

[All models and datasets](https://savrn.com/model-publishers/meta-math)

## Explore More

- [The dataset directory](https://savrn.com/datasets)
- [The model directory](https://savrn.com/models)
- [Research papers](https://savrn.com/papers)

## Source

- Repository metadata, read 2026-10-19.
- [Hugging Face record](https://huggingface.co/datasets/meta-math/MetaMathQA)
- [How the hub is built](https://savrn.com/model-hub/methodology)