UniRef-Image-Edit: Towards Scalable and Consistent Multi-Reference Image Editing
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908835398025216 |
|---|---|
| author | Wei, Hongyang Wen, Bin Long, Yancheng Yang, Yankai Hu, Yuhang Zhang, Tianke Chen, Wei Fan, Haonan Jiang, Kaiyu Chen, Jiankang Liu, Changyi Tang, Kaiyu Ding, Haojie Yang, Xiao Sun, Jia Wang, Huaiqing Yang, Zhenyu Wei, Xinyu He, Xianglong Li, Yangguang Yang, Fan Gao, Tingting Zhang, Lei Zhou, Guorui Li, Han |
| author_facet | Wei, Hongyang Wen, Bin Long, Yancheng Yang, Yankai Hu, Yuhang Zhang, Tianke Chen, Wei Fan, Haonan Jiang, Kaiyu Chen, Jiankang Liu, Changyi Tang, Kaiyu Ding, Haojie Yang, Xiao Sun, Jia Wang, Huaiqing Yang, Zhenyu Wei, Xinyu He, Xianglong Li, Yangguang Yang, Fan Gao, Tingting Zhang, Lei Zhou, Guorui Li, Han |
| contents | We present UniRef-Image-Edit, a high-performance multi-modal generation system that unifies single-image editing and multi-image composition within a single framework. Existing diffusion-based editing methods often struggle to maintain consistency across multiple conditions due to limited interaction between reference inputs. To address this, we introduce Sequence-Extended Latent Fusion (SELF), a unified input representation that dynamically serializes multiple reference images into a coherent latent sequence. During a dedicated training stage, all reference images are jointly constrained to fit within a fixed-length sequence under a global pixel-budget constraint. Building upon SELF, we propose a two-stage training framework comprising supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we jointly train on single-image editing and multi-image composition tasks to establish a robust generative prior. We adopt a progressive sequence length training strategy, in which all input images are initially resized to a total pixel budget of $1024^2$, and are then gradually increased to $1536^2$ and $2048^2$ to improve visual fidelity and cross-reference consistency. This gradual relaxation of compression enables the model to incrementally capture finer visual details while maintaining stable alignment across references. For the RL stage, we introduce Multi-Source GRPO (MSGRPO), to our knowledge the first reinforcement learning framework tailored for multi-reference image generation. MSGRPO optimizes the model to reconcile conflicting visual constraints, significantly enhancing compositional consistency. We will open-source the code, models, training data, and reward data for community research purposes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_14186 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | UniRef-Image-Edit: Towards Scalable and Consistent Multi-Reference Image Editing Wei, Hongyang Wen, Bin Long, Yancheng Yang, Yankai Hu, Yuhang Zhang, Tianke Chen, Wei Fan, Haonan Jiang, Kaiyu Chen, Jiankang Liu, Changyi Tang, Kaiyu Ding, Haojie Yang, Xiao Sun, Jia Wang, Huaiqing Yang, Zhenyu Wei, Xinyu He, Xianglong Li, Yangguang Yang, Fan Gao, Tingting Zhang, Lei Zhou, Guorui Li, Han Computer Vision and Pattern Recognition We present UniRef-Image-Edit, a high-performance multi-modal generation system that unifies single-image editing and multi-image composition within a single framework. Existing diffusion-based editing methods often struggle to maintain consistency across multiple conditions due to limited interaction between reference inputs. To address this, we introduce Sequence-Extended Latent Fusion (SELF), a unified input representation that dynamically serializes multiple reference images into a coherent latent sequence. During a dedicated training stage, all reference images are jointly constrained to fit within a fixed-length sequence under a global pixel-budget constraint. Building upon SELF, we propose a two-stage training framework comprising supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we jointly train on single-image editing and multi-image composition tasks to establish a robust generative prior. We adopt a progressive sequence length training strategy, in which all input images are initially resized to a total pixel budget of $1024^2$, and are then gradually increased to $1536^2$ and $2048^2$ to improve visual fidelity and cross-reference consistency. This gradual relaxation of compression enables the model to incrementally capture finer visual details while maintaining stable alignment across references. For the RL stage, we introduce Multi-Source GRPO (MSGRPO), to our knowledge the first reinforcement learning framework tailored for multi-reference image generation. MSGRPO optimizes the model to reconcile conflicting visual constraints, significantly enhancing compositional consistency. We will open-source the code, models, training data, and reward data for community research purposes. |
| title | UniRef-Image-Edit: Towards Scalable and Consistent Multi-Reference Image Editing |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.14186 |