CARE-Edit: Condition-Aware Routing of Experts for Contextual Image Editing (CVPR 2026)

🔥 Please star CARE-Edit ⭐ and share it. Thanks! 🔥

Yucheng Wang*, Zedong Wang*, Yuetong Wu, Yue Ma, Dan Xu†

The Hong Kong University of Science and Technology (HKUST)

---

🚩 Updates

• ☑ Our paper is now available on arXiv.
• ☑ CARE-Edit is accepted by CVPR 2026. Codes will be released soon.

💡 Motivation

Existing unified diffusion editors suffer from task interference and cannot dynamically handle conflicting conditions, leading to color bleeding, identity drift, and unpredictable behavior. We propose CARE-Edit - a unified editor which routes diffusion tokens to four specialized experts via a lightweight condition-aware router.

<p align="center"> <img src="assets/abstract.png" alt="Motivation" width="100%"> </p>

🔧 Framework

<p align="center"> <img src="assets/out.png" alt="CARE-Edit Framework" width="100%"> </p>

CARE-Edit introduces condition-aware specialized experts within the frozen DiT backbone. Given multimodal conditions, inputs are tokenized and projected to heterogeneous expert branches. The router assigns confidence scores and selects the Top-K experts to process each token. Expert outputs are normalized, modulated, and fused through the Latent Mixture module, yielding denoised representations refined by Mask Repaint module.

🎨 Results

Contextual Image Editing

<p align="center"> <img src="assets/comp1.png" alt="Comparison 1" width="100%"> </p>

Qualitative Comparisons

<p align="center"> <img src="assets/comp2.png" alt="Comparison 2" width="100%"> </p>

🛠️ Getting Started

Code coming soon! Stay tuned for the full release.

<!-- ### Installation ```bash git clone https://github.com/xxx/CARE-Edit.git cd CARE-Edit pip install -r requirements.txt ``` ### Usage ```bash python run.py --config configs/default.yaml ``` -->

📜 BibTeX

If CARE-Edit is helpful for your research, please cite:

@inproceedings{wang2026careedit,
  title={CARE-Edit: Condition-Aware Routing of Experts for Contextual Image Editing},
  author={Yucheng Wang and Zedong Wang and Yuetong Wu and Yue Ma and Dan Xu},
  booktitle={The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2026}
}

📧 Contact

If you have any questions, please email ywangls@connect.ust.hk.

📜 Sincere Acknowledgement

Appreciate the following works for their great contributions:

• UNO: Serves as the inspiration for our project.
• OmniControl: Foundational conditioning approaches that motivate our routing design.