Generative Modeling on ImageNet 256x256 (train)
3.56FID (50K)Ours
Evaluation results
| Method | Links | |
|---|---|---|
| OursData Budget=0.8% (10K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 3.56 | |
| D2CData Budget=0.8% (10K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 3.98 | |
| RandomData Budget=0.8% (10K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 4.35 | |
| OursData Budget=4.0% (50K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 7.26 | |
| OursData Budget=8.0% (100K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 8.83 | |
| D2CData Budget=4.0% (50K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 11.21 | |
| K-CenterData Budget=0.8% (10K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 14.77 | |
| D2CData Budget=8.0% (100K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 15.01 | |
| HerdingData Budget=0.8% (10K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 22.96 | |
| HerdingData Budget=4.0% (50K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 29.11 | |
| RandomData Budget=4.0% (50K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 31.13 | |
| HerdingData Budget=8.0% (100K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 32.3 | |
| RandomData Budget=8.0% (100K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 36.64 | |
| K-CenterData Budget=4.0% (50K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 61.66 | |
| K-CenterData Budget=8.0% (100K), Backbone=SiT-L/2, Training Iterations=100k2026.06 | 66.96 |