Multi-image editing on MMIE-Bench 1.0 (test)
4.46Addition ScoreMIE-UMM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| MIE-UMMEvaluator=Doubao-1.6, Backbone=Qwen-Edit-2509, Sampling steps=40, Output image resolution=1328x1024, Classifier-guidance scale=42026.01 | 4.46 | 4.1 | 3.28 | 3.59 | 3.15 | 3.65 | 3.7 | |
| DreamOmni2Evaluator=Doubao-1.62026.01 | 3.92 | 3.94 | 3.21 | 3.23 | 3.14 | 3.22 | 3.44 | |
| MIE-UMMEvaluator=Qwen2.5-VL (72B), Backbone=Qwen-Edit-2509, Sampling steps=40, Output image resolution=1328x1024, Classifier-guidance scale=42026.01 | 3.77 | 3.51 | 3.09 | 3.22 | 3.12 | 3.3 | 3.34 | |
| OmniGen2Evaluator=Doubao-1.62026.01 | 3.68 | 3.08 | 3.17 | 3.29 | 2.65 | 3.32 | 3.2 | |
| Qwen-EditEvaluator=Doubao-1.62026.01 | 3.66 | 3.35 | 2.45 | 2.79 | 2.83 | 2.63 | 2.95 | |
| OmniGen2Evaluator=Qwen2.5-VL (72B)2026.01 | 3.26 | 2.82 | 2.93 | 3.07 | 2.79 | 3.09 | 3.03 | |
| DreamOmni2Evaluator=Qwen2.5-VL (72B)2026.01 | 3.23 | 3.35 | 2.93 | 2.97 | 2.83 | 2.93 | 3.04 | |
| Qwen-EditEvaluator=Qwen2.5-VL (72B)2026.01 | 2.99 | 3 | 2.56 | 2.72 | 2.75 | 2.67 | 2.77 |