Inconsistency and intrinsic difficulty for different guidance scales and number of diffusion steps for Stable Diffusion 3. Classification error rate and obtained with Dual Vision Transformer –DaViT– model.
| Diffusion steps/Guidance | 2 | 4 | 8 | 12 | 16 | 32 | 64 |
|---|---|---|---|---|---|---|---|
| 0.5 | 3.49% | 2.55% | 2.38% | 2.32% | 2.49% | 2.53% | 3.33% |
| 1 | 3.40% | 2.48% | 2.34% | 2.31% | 2.44% | 2.48% | 3.27% |
| 2 | 3.33% | 2.44% | 2.31% | 2.34% | 2.41% | 2.50% | 3.22% |
| 3 | 3.31% | 2.44% | 2.36% | 2.43% | 2.57% | 3.27% | 3.92% |
| 4 | 3.42% | 2.76% | 2.57% | 2.60% | 2.65% | 2.69% | 4.12% |
| 8 | 3.88% | 3.18% | 3.01% | 2.98% | 2.99% | 3.03% | 4.64% |
| 16 | 4.55% | 3.91% | 3.69% | 3.75% | 3.80% | 3.77% | 4.97% |
| Diffusion steps/Guidance | 2 | 4 | 8 | 12 | 16 | 32 | 64 |
|---|---|---|---|---|---|---|---|
| 0.5 | 3.49% | 2.55% | 2.38% | 2.32% | 2.49% | 2.53% | 3.33% |
| 1 | 3.40% | 2.48% | 2.34% | 2.31% | 2.44% | 2.48% | 3.27% |
| 2 | 3.33% | 2.44% | 2.31% | 2.34% | 2.41% | 2.50% | 3.22% |
| 3 | 3.31% | 2.44% | 2.36% | 2.43% | 2.57% | 3.27% | 3.92% |
| 4 | 3.42% | 2.76% | 2.57% | 2.60% | 2.65% | 2.69% | 4.12% |
| 8 | 3.88% | 3.18% | 3.01% | 2.98% | 2.99% | 3.03% | 4.64% |
| 16 | 4.55% | 3.91% | 3.69% | 3.75% | 3.80% | 3.77% | 4.97% |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.