Masked Generative Transformer Is What You Need for Image Editing
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909033519120384 |
|---|---|
| author | Chow, Wei Li, Linfeng Sun, Xian Kong, Lingdong Li, Zefeng Xu, Qi Song, Hang Ye, Tian Wang, Xian Bai, Jinbin Xu, Shilin Li, Xiangtai Pan, Junting Liu, Shaoteng Zhou, Ran Yang, Tianshu Liu, Songhua |
| author_facet | Chow, Wei Li, Linfeng Sun, Xian Kong, Lingdong Li, Zefeng Xu, Qi Song, Hang Ye, Tian Wang, Xian Bai, Jinbin Xu, Shilin Li, Xiangtai Pan, Junting Liu, Shaoteng Zhou, Ran Yang, Tianshu Liu, Songhua |
| contents | Diffusion models dominate image editing, yet their global denoising mechanism entangles edited regions with surrounding context, causing modifications to propagate into areas that should remain intact. We propose a fundamentally different approach by leveraging Masked Generative Transformers (MGTs), whose localized token-prediction paradigm naturally confines changes to intended regions. We present EditMGT, an MGT-based editing framework that is the first of its kind. Our approach employs multi-layer attention consolidation to aggregate cross-attention maps into precise edit localization signals, and region-hold sampling to explicitly prevent token flipping in non-target areas. To support training, we construct CrispEdit-2M, a 2M-sample high-resolution (>1024) editing dataset spanning seven categories. With only 960M parameters, EditMGT achieves state-of-the-art image similarity on multiple benchmarks while delivering 6x faster editing, demonstrating that MGTs offer a compelling alternative to diffusion-based editing. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_10859 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Masked Generative Transformer Is What You Need for Image Editing Chow, Wei Li, Linfeng Sun, Xian Kong, Lingdong Li, Zefeng Xu, Qi Song, Hang Ye, Tian Wang, Xian Bai, Jinbin Xu, Shilin Li, Xiangtai Pan, Junting Liu, Shaoteng Zhou, Ran Yang, Tianshu Liu, Songhua Computer Vision and Pattern Recognition Machine Learning Diffusion models dominate image editing, yet their global denoising mechanism entangles edited regions with surrounding context, causing modifications to propagate into areas that should remain intact. We propose a fundamentally different approach by leveraging Masked Generative Transformers (MGTs), whose localized token-prediction paradigm naturally confines changes to intended regions. We present EditMGT, an MGT-based editing framework that is the first of its kind. Our approach employs multi-layer attention consolidation to aggregate cross-attention maps into precise edit localization signals, and region-hold sampling to explicitly prevent token flipping in non-target areas. To support training, we construct CrispEdit-2M, a 2M-sample high-resolution (>1024) editing dataset spanning seven categories. With only 960M parameters, EditMGT achieves state-of-the-art image similarity on multiple benchmarks while delivering 6x faster editing, demonstrating that MGTs offer a compelling alternative to diffusion-based editing. |
| title | Masked Generative Transformer Is What You Need for Image Editing |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2605.10859 |