Image Editing on TV-Edit-Bench
91.34DS_tgt_globalTV-Edit-Kontext
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TV-Edit-KontextInput=I+T2026.06 | 91.34 | 95.14 | 16.96 | 4.84 | 5.08 | 86 | 99 | |
| TV-Edit-QwenInput=I+T2026.06 | 91.34 | 94.9 | 16.72 | 4.21 | 4.62 | 93 | 100 | |
| NanoBananaProInput=I2026.06 | 90.96 | 94.32 | 20.72 | 12.01 | 11.95 | 89 | 100 | |
| CLIP-DragInput=M+P+T2026.06 | 87.41 | 93.01 | 23.12 | 13.83 | 13.63 | 71 | 95 | |
| FLUX-KontextInput=I2026.06 | 87.4 | 93.02 | 27.28 | 16.26 | 16.13 | 82 | 98 | |
| GoodDragInput=M+P+T2026.06 | 87.24 | 93.05 | 22.34 | 6.3 | 6.48 | 75 | 96 | |
| GeoDragInput=M+P+T2026.06 | 87.1 | 90.59 | 29.04 | 7.75 | 7.76 | 79 | 94 | |
| Qwen-Image-EditInput=I2026.06 | 85.75 | 91.91 | 31.23 | 13.79 | 13.8 | 86 | 97 | |
| LongCat-Image-EditInput=I2026.06 | 85.5 | 91.85 | 33.91 | 12.47 | 12.29 | 93 | 99 | |
| MotionEditInput=I2026.06 | 84.59 | 91.66 | 29.43 | 12.28 | 11.73 | 92 | 97 | |
| DragLoRAInput=M+P+T2026.06 | 84.44 | 92.07 | 23.8 | 6.38 | 6.71 | 77 | 94 | |
| DragDiffusionInput=M+P+T2026.06 | 82.9 | 92.88 | 24.9 | 10.44 | 10.48 | 72 | 98 | |
| LightningDragInput=M+T2026.06 | 82.54 | 91.55 | 26.33 | 6.49 | 6.89 | 79 | 93 |