Fineweb-Instruct

We convert the pre-training corpus from Fineweb-Edu (https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu) to instruction following format. We select a subset with quality filter and then use GPT-4 to extract instruction-following pairs. The dataset contains roughly 16M instruction pairs. The basic concept is similar to MAmmoTH2 (https://arxiv.org/abs/2405.03548).


## Citation
If you use dataset useful, please cite the following paper:

```
@article{yue2024mammoth2,
title={MAmmoTH2: Scaling Instructions from the Web},
author={Yue, Xiang and Zheng, Tuney and Zhang, Ge and Chen, Wenhu},
journal={arXiv preprint arXiv:2405.03548},
year={2024}
}
```

我们将来自Fineweb-Edu（https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu）的预训练语料库转换为指令遵循格式。我们通过质量过滤选取了一个子集，随后使用GPT-4提取指令遵循样本对。本数据集共包含约1600万条指令样本对，其核心设计理念与MAmmoTH2（https://arxiv.org/abs/2405.03548）相近。


## 引用
若您使用本数据集，请引用如下论文：

@article{yue2024mammoth2,
title={MAmmoTH2：从网络中扩展指令},
author={Yue, Xiang and Zheng, Tuney and Zhang, Ge and Chen, Wenhu},
journal={arXiv预印本 arXiv:2405.03548},
year={2024}
}