GEditBench v2: A Human-Aligned Benchmark for General 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_ | 1866908921284788224 |
|---|---|
| author | Jiang, Zhangqi Sun, Zheng Zeng, Xianfang Yang, Yufeng Zhang, Xuanyang Wu, Yongliang Cheng, Wei Yu, Gang Yang, Xu Wen, Bihan |
| author_facet | Jiang, Zhangqi Sun, Zheng Zeng, Xianfang Yang, Yufeng Zhang, Xuanyang Wu, Yongliang Cheng, Wei Yu, Gang Yang, Xu Wen, Bihan |
| contents | Recent advances in image editing have enabled models to handle complex instructions with impressive realism. However, existing evaluation frameworks lag behind: current benchmarks suffer from narrow task coverage, while standard metrics fail to adequately capture visual consistency, i.e., the preservation of identity, structure and semantic coherence between edited and original images. To address these limitations, we introduce GEditBench v2, a comprehensive benchmark with 1,200 real-world user queries spanning 23 tasks, including a dedicated open-set category for unconstrained, out-of-distribution editing instructions beyond predefined tasks. Furthermore, we propose PVC-Judge, an open-source pairwise assessment model for visual consistency, trained via two novel region-decoupled preference data synthesis pipelines. Besides, we construct VCReward-Bench using expert-annotated preference pairs to assess the alignment of PVC-Judge with human judgments on visual consistency evaluation. Experiments show that our PVC-Judge achieves state-of-the-art evaluation performance among open-source models and even surpasses GPT-5.1 on average. Finally, by benchmarking 16 frontier editing models, we show that GEditBench v2 enables more human-aligned evaluation, revealing critical limitations of current models, and providing a reliable foundation for advancing precise image editing. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_28547 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | GEditBench v2: A Human-Aligned Benchmark for General Image Editing Jiang, Zhangqi Sun, Zheng Zeng, Xianfang Yang, Yufeng Zhang, Xuanyang Wu, Yongliang Cheng, Wei Yu, Gang Yang, Xu Wen, Bihan Computer Vision and Pattern Recognition Recent advances in image editing have enabled models to handle complex instructions with impressive realism. However, existing evaluation frameworks lag behind: current benchmarks suffer from narrow task coverage, while standard metrics fail to adequately capture visual consistency, i.e., the preservation of identity, structure and semantic coherence between edited and original images. To address these limitations, we introduce GEditBench v2, a comprehensive benchmark with 1,200 real-world user queries spanning 23 tasks, including a dedicated open-set category for unconstrained, out-of-distribution editing instructions beyond predefined tasks. Furthermore, we propose PVC-Judge, an open-source pairwise assessment model for visual consistency, trained via two novel region-decoupled preference data synthesis pipelines. Besides, we construct VCReward-Bench using expert-annotated preference pairs to assess the alignment of PVC-Judge with human judgments on visual consistency evaluation. Experiments show that our PVC-Judge achieves state-of-the-art evaluation performance among open-source models and even surpasses GPT-5.1 on average. Finally, by benchmarking 16 frontier editing models, we show that GEditBench v2 enables more human-aligned evaluation, revealing critical limitations of current models, and providing a reliable foundation for advancing precise image editing. |
| title | GEditBench v2: A Human-Aligned Benchmark for General Image Editing |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.28547 |