Instruction-based Image Editing on KRIS Bench 38 (test)
79.8Factual ScoreGPT-4o
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4oFine-tuning Source=Base Model2026.03 | 79.8 | 81.37 | 78.32 | 80.09 | |
| ScaleEditFine-tuning Source=Bagel, Category=Our Dataset2026.03 | 70.24 | 60.78 | 60.54 | 63.58 | |
| ImgEditFine-tuning Source=Bagel, Category=Open-source Datasets2026.03 | 69.61 | 57.28 | 58.82 | 62.01 | |
| OpenGPT-4o-ImageFine-tuning Source=Bagel, Category=Commercial Datasets2026.03 | 67.6 | 57.59 | 59.58 | 61.09 | |
| Pico-BananaFine-tuning Source=Bagel, Category=Commercial Datasets2026.03 | 67.38 | 57.96 | 58.91 | 61.04 | |
| AnyEditFine-tuning Source=Bagel, Category=Open-source Datasets2026.03 | 66.96 | 38.89 | 56.71 | 61.2 | |
| ShareGPT-4o-ImageFine-tuning Source=Bagel, Category=Commercial Datasets2026.03 | 66.92 | 58.31 | 59.85 | 61.28 | |
| GPT-Image-EditFine-tuning Source=Bagel, Category=Commercial Datasets2026.03 | 66.86 | 44.32 | 60.7 | 62.92 | |
| UltraEditFine-tuning Source=Bagel, Category=Open-source Datasets2026.03 | 66.26 | 56.55 | 52.91 | 58.64 | |
| Nano-consistentFine-tuning Source=Bagel, Category=Commercial Datasets2026.03 | 66.21 | 58.17 | 58.61 | 60.71 | |
| OmniEditFine-tuning Source=Bagel, Category=Open-source Datasets2026.03 | 66.04 | 57.59 | 59.46 | 60.59 | |
| NHR-EditFine-tuning Source=Bagel, Category=Open-source Datasets2026.03 | 65.74 | 57.93 | 58.04 | 60.31 | |
| Pico-BananaFine-tuning Source=UniWorld-V1, Category=Commercial Datasets2026.03 | 58.86 | 63.87 | 38.61 | 56.51 | |
| ScaleEditFine-tuning Source=UniWorld-V1, Category=Our Dataset2026.03 | 57.76 | 64.29 | 39.95 | 56.6 | |
| ShareGPT-4o-ImageFine-tuning Source=UniWorld-V1, Category=Commercial Datasets2026.03 | 56.07 | 64.29 | 36.15 | 55.24 | |
| Nano-consistentFine-tuning Source=UniWorld-V1, Category=Commercial Datasets2026.03 | 54.92 | 57.88 | 36.64 | 52.03 | |
| OpenGPT-4o-ImageFine-tuning Source=UniWorld-V1, Category=Commercial Datasets2026.03 | 53.49 | 60.03 | 39.51 | 53.22 | |
| OmniEditFine-tuning Source=UniWorld-V1, Category=Open-source Datasets2026.03 | 51.71 | 57.85 | 27.5 | 48.87 | |
| GPT-Image-EditFine-tuning Source=UniWorld-V1, Category=Commercial Datasets2026.03 | 51.62 | 60.71 | 33.73 | 51.6 | |
| ImgEditFine-tuning Source=UniWorld-V1, Category=Open-source Datasets2026.03 | 51.15 | 58.68 | 28.92 | 49.44 | |
| NHR-EditFine-tuning Source=UniWorld-V1, Category=Open-source Datasets2026.03 | 50.65 | 60.57 | 28.65 | 50.07 | |
| AnyEditFine-tuning Source=UniWorld-V1, Category=Open-source Datasets2026.03 | 47.81 | 54.35 | 23.24 | 45.09 | |
| UniWorld-V1Fine-tuning Source=UniWorld-V1, Category=Baseline2026.03 | 47.71 | 44.8 | 47.92 | 50.27 | |
| BagelFine-tuning Source=Bagel, Category=Baseline2026.03 | 47.71 | 44.8 | 47.92 | 50.27 | |
| Step1X-EditFine-tuning Source=Base Model2026.03 | 45.52 | 48.01 | 31.82 | 43.29 | |
| UltraEditFine-tuning Source=UniWorld-V1, Category=Open-source Datasets2026.03 | 36.61 | 43.59 | 20.72 | 36.17 | |
| OmniGenFine-tuning Source=Base Model2026.03 | 33.11 | 28.02 | 23.89 | 28.85 |