---
license: apache-2.0
---
# Video-Coder Training Data
This repository contains the training data for [Video-Coder: Reinforcing Agentic Coding for Video Generation](https://arxiv.org/abs/2611.09894).
[**🌐 Project Page**](https://cywang735.github.io/Video-Coder/) | [**📖 Paper**](https://arxiv.org/abs/2611.09894) | [**💻 Code**](https://github.com/CYWang735/Video-Coder)
# 👀 Intro
**Video-Coder** is a code-driven agent for video generation. It iteratively writes, executes, and refines Blender Python programs to construct dynamic 3D scenes and render draft videos, providing structural and spatial guidance for downstream video generators.
We design a dedicated data pipeline and curate two high-quality datasets, **Video-Coder-SFT-10k** and **Video-Coder-RL-4k**, which contain diverse prompts and ground-truth reference videos.
## Dataset Files
The Video-Coder training data consists of JSON files tailored for different training stages:
* `sft_data.json`: Used for Supervised Fine-Tuning (SFT) cold start.
* `rl_data.json`: Used for Reinforcement Learning (RL) training using GRPO.
## 🎥 Demo
## Citation
If you find this work helpful for your research, please consider citing:
```bibtex
@article{wang2026video,
title={Video-Coder: Reinforcing Agentic Coding for Video Generation},
author={Wang, Chaoyang and Gao, Sicheng and Hu, Junan and Li, Kunyang and Chen, Dongyang and Zhou, Yiyang and Yao, Huaxiu and Lee, Yong Jae and Timofte, Radu and Shang, Yuzhang},
journal={arXiv preprint arXiv:2611.09894},
year={2026}
}
```