| | --- |
| | license: apache-2.0 |
| | datasets: |
| | - TIGER-Lab/VisCode-Multi-679K |
| | base_model: |
| | - Qwen/Qwen2.5-Coder-3B-Instruct |
| | library_name: transformers |
| | language: |
| | - en |
| | tags: |
| | - code |
| | --- |
| | |
| | # VisCoder2-3B |
| |
|
| | [🏠 Project Page](https://tiger-ai-lab.github.io/VisCoder2) | [📖 Paper](https://arxiv.org/abs/2510.23642) | [💻 GitHub](https://github.com/TIGER-AI-Lab/VisCoder2) | [🤗 VisCode2](https://hf.co/collections/TIGER-Lab/viscoder2) |
| |
|
| | **VisCoder2-3B** is a lightweight multi-language visualization coding model trained for **executable code generation, rendering, and iterative self-debugging**. |
| |
|
| | --- |
| |
|
| | ## 🧠 Model Description |
| |
|
| | **VisCoder2-3B** is trained on the **VisCode-Multi-679K** dataset, a large-scale instruction-tuning dataset for executable visualization tasks across **12 programming language**. It addresses a core challenge in multi-language visualization: generating code that not only executes successfully but also produces semantically consistent visual outputs by aligning natural-language instructions and rendering results. |
| |
|
| | --- |
| |
|
| | ## 📊 Main Results on VisPlotBench |
| |
|
| | We evaluate VisCoder2-3B on [**VisPlotBench**](https://huggingface.co/datasets/TIGER-Lab/VisPlotBench), which includes 888 executable visualization tasks spanning 8 languages, supporting both standard generation and multi-turn self-debugging. |
| |
|
| |  |
| |
|
| | > **VisCoder2-3B** shows consistent performance across multiple languages and achieves notable improvements under the multi-round self-debug setting. |
| | --- |
| |
|
| | ## 📁 Training Details |
| |
|
| | - **Base model**: Qwen2.5-Coder-3B-Instruct |
| | - **Framework**: [ms-swift](https://github.com/modelscope/swift) |
| | - **Tuning method**: Full-parameter supervised fine-tuning (SFT) |
| | - **Dataset**: [VisCode-Multi-679K](https://huggingface.co/datasets/TIGER-Lab/VisCode-Multi-679K) |
| |
|
| | --- |
| |
|
| | ## 📖 Citation |
| |
|
| | If you use VisCoder2-3B or related datasets in your research, please cite: |
| |
|
| | ```bibtex |
| | @article{ni2025viscoder2, |
| | title={VisCoder2: Building Multi-Language Visualization Coding Agents}, |
| | author={Ni, Yuansheng and Cai, Songcheng and Chen, Xiangchao and Liang, Jiarong and Lyu, Zhiheng and Deng, Jiaqi and Zou, Kai and Nie, Ping and Yuan, Fei and Yue, Xiang and others}, |
| | journal={arXiv preprint arXiv:2510.23642}, |
| | year={2025} |
| | } |
| | |
| | @article{ni2025viscoder, |
| | title={VisCoder: Fine-Tuning LLMs for Executable Python Visualization Code Generation}, |
| | author={Ni, Yuansheng and Nie, Ping and Zou, Kai and Yue, Xiang and Chen, Wenhu}, |
| | journal={arXiv preprint arXiv:2506.03930}, |
| | year={2025} |
| | } |
| | ``` |
| |
|
| | For evaluation scripts and more information, see our [GitHub repository](https://github.com/TIGER-AI-Lab/VisCoder2). |
