Text-centric image editing on TextEdit MiniSet-500 (Real)
77Overall Accuracy (OA)InternVL-U
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| InternVL-U# Params=2B+1.7B2026.03 | 77 | 74 | 70 | 71 | 71 | 76 | 5.79 | |
| Qwen-Image-Edit# Params=20B2026.03 | 76 | 69 | 67 | 67 | 70 | 75 | 5.81 | |
| Nano Banana Pro2026.03 | 76 | 71 | 69 | 70 | 70 | 75 | 5.86 | |
| GPT-Image-1.52026.03 | 72 | 68 | 66 | 67 | 67 | 75 | 5.85 | |
| BAGEL# Params=7B+7B2026.03 | 61 | 59 | 52 | 54 | 54 | 74 | 5.79 | |
| Ovis-U1# Params=2.4B+1.2B2026.03 | 37 | 34 | 32 | 32 | 33 | 72 | 5.39 | |
| Lumina-DiMOO# Params=8B2026.03 | 20 | 22 | 18 | 19 | 19 | 70 | 5.58 |