Image Editing on MiLDEBench (test)
1.34Instruction Following ScoreNano Banana
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Nano Banana2026.01 | 1.34 | 1.76 | 1.95 | 1.84 | 1.45 | |
| MiLDEAgentMLLM=Qwen2.5VL-7B, Generative Model=Flux2026.01 | 1.28 | 2.83 | 1.27 | 1.75 | 1.37 | |
| Bagel w/ Thinking2026.01 | 0.87 | 2.05 | 1.16 | 0.58 | 0.39 | |
| MagicBrush2026.01 | 0.35 | 2.27 | 1.07 | 0.59 | 0.28 | |
| Instruct-Pix2Pix2026.01 | 0.05 | 2.85 | 1.23 | 0.67 | 0.27 |