Class-unconditional Image Generation on ImageNet 256x256
1.67FIDVQ only
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VQ onlyType=validation data2024.10 | 1.67 | 175.56 | — | — | |
| DiGIT + VQType=validation data2024.10 | 1.92 | 184.4 | — | — | |
| RCG-GBackbone=MAGE-L, Epochs=800, Params=502M2025.12 | 2.15 | 253.4 | — | — | |
| SFDEpochs=200, Params=675M, AutoGuidance (AG)=true2025.12 | 2.9 | 148.5 | — | — | |
| RCGBackbone=MAGE-L, Epochs=800, Params=502M2025.12 | 3.44 | 186.9 | — | — | |
| SFDEpochs=80, Params=675M, AutoGuidance (AG)=true2025.12 | 3.77 | 127.9 | — | — | |
| DIGIT (+VQGAN)Type=AR, #Param=732M, #Epoch=2002024.10 | 4.59 | 141.29 | — | — | |
| RCGBackbone=DiT-XL/2, Epochs=400, Params=675M2025.12 | 4.89 | 143.2 | — | — | |
| RAEEpochs=200, Params=839M, AutoGuidance (AG)=true2025.12 | 4.96 | 123.1 | — | — | |
| Latent-CFM (pretrained)# Params.=675.2M, Solver=SDE (Euler-Maruyama)2025.05 | 5.955 | — | — | — | |
| Latent-CFM (pretrained)# Params.=675.2M, Solver=ODE (adaptive dopri5)2025.05 | 6.076 | — | — | — | |
| I-CFM# Params.=675.1M, Solver=SDE (Euler-Maruyama)2025.05 | 6.745 | — | — | — | |
| I-CFM# Params.=675.1M, Solver=ODE (adaptive dopri5)2025.05 | 6.893 | — | — | — | |
| SFDEpochs=200, Params=675M, AutoGuidance (AG)=false2025.12 | 8.46 | 89.9 | — | — | |
| DIGIT (+MaskGIT)Type=MIM, #Param=219M, #Epoch=2002024.10 | 9.04 | 75.04 | — | — | |
| MAGEType=MIM, #Param=463M, #Epoch=16002024.10 | 9.1 | 105.1 | — | — | |
| DIGIT (+VQGAN)Type=AR, #Param=219M, #Epoch=4002024.10 | 9.13 | 73.85 | — | — | |
| DiT-XL/2 + DGType=Uncond, DG=True, Backbone=DiT-XL/22026.07 | 9.68 | 122.22 | 66 | 67 | |
| SFDEpochs=80, Params=675M, AutoGuidance (AG)=false2025.12 | 10.24 | 78.5 | — | — | |
| GIVTType=AR, #Param=1.67B, #Epoch=5002024.10 | 11.02 | — | — | — | |
| MAGEType=MIM, #Param=200M, #Epoch=16002024.10 | 11.1 | 81.17 | — | — | |
| DiT-XL/2Type=Uncond, DG=False, Backbone=DiT-XL/22026.07 | 16.95 | 105.64 | 61 | 76 | |
| GIVTType=AR, #Param=304M, #Epoch=5002024.10 | 17.7 | — | — | — | |
| MaskGITType=MIM, #Param=227M, #Epoch=3002024.10 | 20.7 | 42.08 | — | — | |
| ReDiEpochs=80, Params=675M, Representation Guidance (RG)=true2025.12 | 22.6 | — | — | — | |
| VQGANType=AR, #Param=214M, #Epoch=2002024.10 | 24.38 | 30.93 | — | — | |
| ReDiEpochs=80, Params=675M2025.12 | 25.1 | — | — | — | |
| ADMType=Diff., #Param=554M2024.10 | 26.2 | 39.7 | — | — | |
| DiT-XLEpochs=400, Params=675M2025.12 | 30.68 | 32.7 | — | — | |
| BigGANType=GAN, #Param=70M2024.10 | 38.6 | 24.7 | — | — | |
| LDMType=Diff., #Param=395M2024.10 | 39.1 | 22.83 | — | — | |
| NoiseRaterStep=+80k, Model Backbone=DiT-S/22026.05 | 59.95 | — | — | — | |
| NoiseRaterStep=+60k, Model Backbone=DiT-S/22026.05 | 60.37 | — | — | — | |
| Naive (max)Step=+80k, Model Backbone=DiT-S/22026.05 | 60.82 | — | — | — | |
| NoiseRaterStep=+40k, Model Backbone=DiT-S/22026.05 | 61.03 | — | — | — | |
| Naive (max)Step=+60k, Model Backbone=DiT-S/22026.05 | 61.41 | — | — | — | |
| VanillaStep=+80k, Model Backbone=DiT-S/22026.05 | 61.56 | — | — | — | |
| NoiseRaterStep=+20k, Model Backbone=DiT-S/22026.05 | 61.87 | — | — | — | |
| Naive (max)Step=+40k, Model Backbone=DiT-S/22026.05 | 62 | — | — | — | |
| VanillaStep=+60k, Model Backbone=DiT-S/22026.05 | 62.03 | — | — | — | |
| Naive (max)Step=+20k, Model Backbone=DiT-S/22026.05 | 62.65 | — | — | — | |
| VanillaStep=+40k, Model Backbone=DiT-S/22026.05 | 62.78 | — | — | — | |
| VanillaStep=+20k, Model Backbone=DiT-S/22026.05 | 63.21 | — | — | — | |
| Naive (min)Step=+80k, Model Backbone=DiT-S/22026.05 | 63.31 | — | — | — | |
| Naive (min)Step=+60k, Model Backbone=DiT-S/22026.05 | 63.97 | — | — | — | |
| Naive (min)Step=+40k, Model Backbone=DiT-S/22026.05 | 64.5 | — | — | — | |
| Naive (min)Step=+20k, Model Backbone=DiT-S/22026.05 | 64.69 | — | — | — |