Text-to-Image Generation on 80 distinct prompts (Portrait, Landscape, Painting, Objects) (test)
1,835Auto-MOG Wins (Count)Auto-MOG
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Auto-MOGMetric Criterion=Overall Quality, Baseline=Standard CFG, Total Trials=2,960, Architectures=Lumina, SD-2.1, SD-XL, SD-3.5, Protocol=2AFC2026.03 | 1,835 | — | — | — | — | 62 | 60.2 | 0.42 | 0.001 | |
| Auto-MOGMetric Criterion=Color Fidelity, Baseline=Standard CFG, Total Trials=2,960, Architectures=Lumina, SD-2.1, SD-XL, SD-3.5, Protocol=2AFC2026.03 | 1,823 | — | — | — | — | 61.6 | 59.8 | 0.45 | 0.001 | |
| Auto-MOGMetric Criterion=Realism, Baseline=Standard CFG, Total Trials=2,960, Architectures=Lumina, SD-2.1, SD-XL, SD-3.5, Protocol=2AFC2026.03 | 1,794 | — | — | — | — | 60.6 | 58.8 | 0.48 | 0.001 | |
| Auto-MOGMetric Criterion=Texture Detail, Baseline=Standard CFG, Total Trials=2,960, Architectures=Lumina, SD-2.1, SD-XL, SD-3.5, Protocol=2AFC2026.03 | 1,732 | — | — | — | — | 58.5 | 56.7 | 0.38 | 0.001 |