rStar-Critique-Data

# rStar-Critique-Data

This repository contains `rStar-Critique-Data`, the dataset used for the paper [Critique-Coder: Enhancing Coder Models by Critique Reinforcement Learning](https://huggingface.co/papers/2509.22824).

This dataset is integral to the Critique Reinforcement Learning (CRL) paradigm, which enhances coder models by explicitly training them to generate critiques for (question, solution) pairs, as described in the accompanying paper.

Data Construction Pipeline is shown:


## Paper
[Critique-Coder: Enhancing Coder Models by Critique Reinforcement Learning](https://huggingface.co/papers/2509.22824)

## Project Page
https://tiger-ai-lab.github.io/Critique-Coder

## Code
https://github.com/TIGER-AI-Lab/Critique-Coder

## Sample Usage

You can download this dataset using the Hugging Face CLI:

```bash
hf download Critique-Coder/rStar-Critique-Data --local-dir ./data/critique-coder-dataset --repo dataset
```

## Citation
```
@article{ruan2025critiquecoder,
title={Critique-Coder: Enhancing Coder Models by Critique Reinforcement Learning},
author={Ruan, Chi and Jiang, Dongfu and Wang, Yubo and Chen, Wenhu},
journal={ArXiv},
year={2025},
volume={2509.22824}
}
```

# rStar-Critique-Data
本仓库收录`rStar-Critique-Data`数据集，该数据集支撑了论文《Critique-Coder：基于批判强化学习提升代码大模型性能》（https://huggingface.co/papers/2509.22824）的研究工作。

该数据集是批判强化学习（Critique Reinforcement Learning, CRL）范式的核心组成部分；如配套论文所述，该范式通过显式训练代码大模型（coder models）为（问题、解决方案）对生成批判评价，以此实现代码模型性能的优化提升。

数据集构建流程如下：

## 论文
[Critique-Coder：基于批判强化学习提升代码大模型性能](https://huggingface.co/papers/2509.22824)

## 项目主页
https://tiger-ai-lab.github.io/Critique-Coder

## 代码仓库
https://github.com/TIGER-AI-Lab/Critique-Coder

## 使用示例
你可以通过Hugging Face命令行工具（CLI）下载该数据集：
bash
hf download Critique-Coder/rStar-Critique-Data --local-dir ./data/critique-coder-dataset --repo dataset

## 引用格式

@article{ruan2025critiquecoder,
title={Critique-Coder: Enhancing Coder Models by Critique Reinforcement Learning},
author={Ruan, Chi and Jiang, Dongfu and Wang, Yubo and Chen, Wenhu},
journal={ArXiv},
year={2025},
volume={2509.22824}
}