In-context image editing on MICE-Bench Turn 2
60.77IF ScoreEdit-R2
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Edit-R22026.06 | 60.77 | 84.33 | 58.17 | 67.76 | |
| GPT-Image-1 [high]2026.06 | 60.42 | 68.1 | 59.93 | 62.82 | |
| Nano-Banana-22026.06 | 49.53 | 82.55 | 46.31 | 59.46 | |
| BAGELvariant=w/ IC-CoT2026.06 | 46.84 | 83.82 | 46.07 | 58.91 | |
| Qwen-Image-Edit [2511]Originally support in-context editing=false2026.06 | 46.47 | 77.58 | 46.5 | 56.85 | |
| BAGELvariant=w/ think2026.06 | 43.66 | 83.58 | 40.03 | 55.76 | |
| BAGEL2026.06 | 42.12 | 85.34 | 39.56 | 55.67 | |
| OmniGen2026.06 | 37.51 | 50.44 | 32.69 | 40.21 | |
| VINCIE2026.06 | 34.68 | 63.37 | 31.92 | 43.32 | |
| Flux.1-Kontext [dev]Originally support in-context editing=false2026.06 | 32.56 | 71.51 | 32.63 | 45.57 |