Document Editing on MiLDEBench 1.0 (test)
25.46Instruction Following ScoreGPT-Image-1
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-Image-1Access=Closed-source2026.01 | 25.46 | 36.24 | 4.66 | 39.67 | 25.6 | — | |
| Nano BananaAccess=Closed-source2026.01 | 24.04 | 58.42 | 4.52 | 40.32 | 27.1 | — | |
| MiLDEAgent (Qwen2.5VL-7B + Flux)Backbone=Qwen2.5VL-7B, Base Model=Flux2026.01 | 20.71 | 93.24 | 4.19 | 36.75 | 25.9 | 80.46 | |
| BagelAccess=Open-source2026.01 | 14.23 | 48.59 | 3.54 | 13.49 | 10.8 | — | |
| BagelMechanism=w/ Thinking2026.01 | 13.6 | 60.91 | 3.65 | 14.51 | 11.23 | — | |
| MiLDEAgent (Qwen2.5VL-3B + Flux)Backbone=Qwen2.5VL-3B, Base Model=Flux2026.01 | 13.29 | 90.15 | 4.32 | 27.52 | 16.1 | 42.9 | |
| Flux1-KontextAccess=Open-source2026.01 | 12.49 | 48.32 | 3.94 | 19.31 | 11.58 | — | |
| UltraEditAccess=Open-source2026.01 | 12.41 | 85.31 | 3.54 | 11.39 | 10.35 | — | |
| Step1X-EditMechanism=w/ Thinking2026.01 | 10.48 | 82.16 | 4.11 | 28.67 | 14.17 | — | |
| Qwen-Image-EditAccess=Open-source2026.01 | 10.09 | 74.2 | 4.12 | 24.32 | 12.42 | — | |
| MagicBrushAccess=Open-source2026.01 | 7.37 | 72.08 | 3.68 | 16.6 | 8.47 | — | |
| VAREditAccess=Open-source2026.01 | 6.6 | 68.1 | 3.18 | 9.49 | 5.86 | — | |
| Step1X-EditAccess=Open-source2026.01 | 6.56 | 84.09 | 3.98 | 18.7 | 8.84 | — | |
| AnyEditAccess=Open-source2026.01 | 6.51 | 56.73 | 3.96 | 21.83 | 9.4 | — | |
| UniWorld-v1Access=Open-source2026.01 | 5.75 | 61.59 | 3.91 | 22.04 | 9.15 | — | |
| OmniGenAccess=Open-source2026.01 | 3.83 | 85.96 | 3.9 | 19.76 | 7.73 | — | |
| Instruct-Pix2PixAccess=Open-source2026.01 | 2.3 | 93.46 | 4.23 | 17.16 | 6.23 | — | |
| ICEditAccess=Open-source2026.01 | 2.28 | 64.6 | 3.42 | 18.25 | 6.43 | — |