README.md

---
task_categories: ['text-generation', 'text-classification']
language: ['zh']
license: apache-2.0
tags: ['legal', 'chinese', 'court-view-generation', 'charge-prediction']
---

# CCVG: Chinese Court View Generation Dataset

[中文](README_zh.md) | 🤗 [huggingface](https://huggingface.co/datasets/TIM0927/CCVG) | 🤖 [modelscope](https://www.modelscope.cn/datasets/ZhitianHou/CCVG) | 📄 [Arxiv](https://arxiv.org/abs/2510.09297) | 💻 [GitHub](https://github.com/ZhitianHou/ShiZhi)

**Now the anonymous test data has been released.**

For privacy protection reasons, the original training data has been removed pending de-identification and anonymization. Only some examples of training data are avaiable.


**CCVG** is a curated Chinese dataset designed for **Criminal Court View Generation (CVG)** and **charge prediction** tasks.  
It contains criminal case documents with **fact descriptions** and **court views**, supporting research in legal AI and natural language generation.

---

## Dataset Overview

- **Language:** Chinese  
- **Domain:** Legal / Criminal law  
- **Task:** Court View Generation & Charge Prediction  
- **Size:** 110K+ criminal cases  
- **Time Span:** 1985–2021  

Each case contains:  
1. **system**: The system prompt we used in Training.
2. **query:** The "Fact" section summarying the case fact in legal case documents.  
3. **response:** The "Court View" section explaining the judgment in legal case documents.  
4. **charge:** The criminal charge corresponding to the case.  



## Citation

If you find this project helpful, please consider citing our paper:

```bibtex
@misc{hou2025shizhichineselightweightlarge,
      title={ShiZhi: A Chinese Lightweight Large Language Model for Court View Generation}, 
      author={Zhitian Hou and Kun Zeng},
      year={2025},
      eprint={2510.09297},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2510.09297}, 
}
```