ThinkRL-Edit: Thinking in Reinforcement Learning for Reasoning-Centric Image Editing

Hengjia Li, Liming Jiang^✝, Yizhi Song, Qing Yan, Hao Kang, Zichuan Liu,
Xin Lu, Boxi Wu^§, Deng Cai

✝Project Lead   §Corresponding Author

Although unified multimodal generative models such as Qwen-Edit have substantially improved editing quality, their underlying reasoning remains underexplored, especially for reasoning-centric editing. In contrast, our method delivers accurate edits with deep reasoning, achieving strong consistency and high perceptual quality across diverse reasoning-driven editing scenarios. See our 🌐 Project Page for more details and results.

🚀 Getting Started

Installation

git clone https://github.com/EchoPluto/ThinkRL-Edit.git
cd ThinkRL
conda create -n thinkrl python=3.10.16
pip install -e .

Model Download

Download our model from Huggingface:

Models	Download Links	Description
Editing Model	🤗 Huggingface	Reasoning-Centric Editing model

Run the Code

You can run an example using the following command:

torchrun --nproc_per_node 8 --master_port 60001 infer_qwen.py

❤️ Ackowledgement

Our implementation is based on Flow-GRPO. Thanks for the great open-source work!

📌 Citation

If any part of our paper or code is helpful to your research, please consider citing our work 📝 and give us a star ⭐. Thanks for your support!

@article{li2026thinkrl,
  title={ThinkRL-Edit: Thinking in Reinforcement Learning for Reasoning-Centric Image Editing},
  author={Li, Hengjia and Jiang, Liming and Yan, Qing and Song, Yizhi and Kang, Hao and Liu, Zichuan and Lu, Xin and Wu, Boxi and Cai, Deng},
  journal={arXiv preprint arXiv:2601.03467},
  year={2026}
}