Instruction-based Image Editing on KRIS-Bench 1.0 (test)
81.02Attribute PerceptionThinkRL-Edit
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ThinkRL-Edit2026.01 | 81.02 | 81.45 | 75.67 | 71.25 | 49.07 | 79.71 | 81.13 | 72.31 | 57.44 | 71.65 | 71.16 | 77.52 | 97.12 | |
| Qwen-Edit2026.01 | 72.57 | 79.92 | 61.45 | 56.38 | 48.57 | 78.44 | 74.53 | 57.6 | 56.68 | 62.77 | 56.54 | 76.37 | 95.86 | |
| Flux-Kontext2026.01 | 70.78 | 69.2 | 51.27 | 52.05 | 45.82 | 73.67 | 70.38 | 51.86 | 53.55 | 57.35 | 46.61 | 77.09 | 94.08 | |
| UniCoT2026.01 | 67.94 | 73.72 | 59.45 | 53.19 | 40.97 | 54.67 | 69.38 | 54.7 | 44.78 | 56.76 | 57.24 | 59.52 | 92.6 | |
| OmniGen22026.01 | 65.41 | 53.36 | 50.46 | 45.3 | 32.19 | 56.36 | 63.57 | 46.55 | 38.83 | 49.24 | 39.4 | 66.72 | 93.16 | |
| Bagel2026.01 | 61.39 | 62.08 | 50.21 | 46.26 | 30.21 | 48.44 | 61.55 | 47.21 | 35.23 | 48.69 | 51.99 | 52.49 | 86.98 | |
| Bagel-Think2026.01 | 60.39 | 61.19 | 49.06 | 47.44 | 29.44 | 48.36 | 60.61 | 47.83 | 34.58 | 48.71 | 55.68 | 70 | 96.35 |