Image Generation on ImageNet condensed 2015 (train)
3.43FIDGADC
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GADCData Budget=0.8% (10K), Eval. Samples=50K, Model Architecture=DiT-L/2, Resolution=256 × 2562026.06 | 3.43 | 414.3 | 78 | 28 | |
| K-Medoids†Data Budget=0.8% (10K), Eval. Samples=50K, Model Architecture=DiT-L/2, Resolution=256 × 256, Attachment Phase=D2C2026.06 | 4.13 | 315.7 | 77 | 24 | |
| K-Centers†Data Budget=0.8% (10K), Eval. Samples=50K, Model Architecture=DiT-L/2, Resolution=256 × 256, Attachment Phase=D2C2026.06 | 39.67 | 142.1 | 27 | 34 |