Text-to-Image Generation on Pick-a-Pic 10,000 prompts
2.34Inference TimeAccelAes
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| AccelAesBackbone=SD3-Medium2026.03 | 2.34 | 1.5 | 0.2744 | 0.9042 | 0.3014 | 5.9898 | 0.9437 | |
| BaselineBackbone=SD3-Medium2026.03 | 3.52 | 1 | 0.2662 | 0.8788 | 0.2895 | 5.733 | 0.916 | |
| AccelAesBackbone=Lumina-Next2026.03 | 5.86 | 2.11 | 0.264 | 0.841 | 0.274 | 6.1414 | 0.6285 | |
| AccelAesBackbone=FLUX.1-dev2026.03 | 7.31 | 1.73 | 0.2772 | 1.317 | 0.3214 | 6.372 | 0.59 | |
| BaselineBackbone=Lumina-Next2026.03 | 12.37 | 1 | 0.2531 | 0.7518 | 0.271 | 5.9407 | 0.5829 | |
| BaselineBackbone=FLUX.1-dev2026.03 | 12.66 | 1 | 0.2753 | 1.233 | 0.32 | 6.243 | 0.56 |