Magpie-100K-Generator-Zoo

### About This Dataset

This dataset is used by the paper ["Stronger Models are NOT Stronger Teachers for Instruction Tuning"](https://huggingface.co/papers/2411.07133).

To create this dataset, instructions are sampled from [Magpie-Air](https://huggingface.co/datasets/Magpie-Align/Llama-3-Magpie-Air-3M-v0.1). Responses are generated using 19 different response generators.

You can build a DPO dataset based on reward values we provided.

**Questions?** Contact [Zhangchen](https://www.zhangchenxu.com) by email.

### Citation
```
@article{xu2024stronger,
title={Stronger Models are NOT Stronger Teachers for Instruction Tuning},
author={Xu, Zhangchen and Jiang, Fengqing and Niu, Luyao and Lin, Bill Yuchen and Poovendran, Radha},
journal={arXiv preprint arXiv:2411.07133},
year={2024}
}
```

### 数据集简介
本数据集已被论文《更强的模型并非指令微调的更强教师》（https://huggingface.co/papers/2411.07133）所使用。
为构建本数据集，我们从Magpie-Air数据集（https://huggingface.co/datasets/Magpie-Align/Llama-3-Magpie-Air-3M-v0.1）中采样获取指令样本，回复内容由19种不同的回复生成器生成。
您可依据我们提供的奖励值构建直接偏好优化（DPO）数据集。
**如有疑问？** 可通过邮件联系张晨（https://www.zhangchenxu.com）。

### 引用格式

@article{xu2024stronger,
title={Stronger Models are NOT Stronger Teachers for Instruction Tuning},
author={Xu, Zhangchen and Jiang, Fengqing and Niu, Luyao and Lin, Bill Yuchen and Poovendran, Radha},
journal={arXiv preprint arXiv:2411.07133},
year={2024}
}