Image Generation on ImageNet 512x512 (test)
1.366FIDSiD2A-EDM2-XXL
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SiD2A-EDM2-XXLCFG=false, NFE=1, #Params=1.5B2024.10 | 1.366 | — | — | — | — | — | — | |
| SiD2A-EDM2-XLCFG=false, NFE=1, #Params=1.1B2024.10 | 1.379 | — | — | — | — | — | — | |
| SiD2A-EDM2-LCFG=false, NFE=1, #Params=777M2024.10 | 1.413 | — | — | — | — | — | — | |
| SiD2A-EDM2-MCFG=false, NFE=1, #Params=498M2024.10 | 1.488 | — | — | — | — | — | — | |
| SiD2A-EDM2-SCFG=false, NFE=1, #Params=280M2024.10 | 1.669 | — | — | — | — | — | — | |
| MARCFG=true, NFE=100x2, #Params=481M2024.10 | 1.73 | — | — | — | — | — | — | |
| SiT-XL/2 + Self-TranscendenceDiffusion Space=Latent diffusion, Architecture=Transformer, Epochs=2002026.01 | 1.76 | 4.16 | — | 286.6 | 0.82 | 0.62 | — | |
| MAGVIT-v2CFG=true, NFE=64x2, #Params=307M2024.10 | 1.91 | — | — | — | — | — | — | |
| SiT-XL/2 + Self-TranscendenceDiffusion Space=Latent diffusion, Architecture=Transformer, Epochs=1002026.01 | 2 | 4.11 | — | 265 | 0.83 | 0.58 | — | |
| SiT-XL/2 + REPADiffusion Space=Latent diffusion, Architecture=Transformer, Epochs=2002026.01 | 2.08 | 4.19 | — | 274.6 | 0.83 | 0.58 | — | |
| SiD2A+ELCFG=N, NFE=1, Batch Size=2048, Model Scale=XS, Parameters=125M2026.04 | 2.132 | — | — | — | — | — | — | |
| SiD2A-EDM2-XSCFG=false, NFE=1, #Params=125M2024.10 | 2.156 | — | — | — | — | — | — | |
| SiD2ACFG=N, NFE=1, Batch Size=2048, Model Scale=XS, Parameters=125M2026.04 | 2.156 | — | — | — | — | — | — | |
| SiD2ACFG=N, NFE=1, Batch Size=2048, Reproduced=true, Model Scale=XS, Parameters=125M2026.04 | 2.191 | — | — | — | — | — | — | |
| SiDACFG=N, NFE=1, Batch Size=2048, Model Scale=XS, Parameters=125M2026.04 | 2.228 | — | — | — | — | — | — | |
| SiD2A+ELCFG=N, NFE=1, Batch Size=16, Model Scale=XS, Parameters=125M2026.04 | 2.371 | — | — | — | — | — | — | |
| StyleGAN-XLResolution=512x512, Number of parameters=168.4M2023.03 | 2.41 | 4.06 | 51.54 | 267.75 | 77 | 52 | — | |
| StyleGAN-XLCFG=false, NFE=1x2, #Params=168M2024.10 | 2.41 | — | — | — | — | — | — | |
| MaskDiT-GCFG=true, NFE=79x2, #Params=736M2024.10 | 2.5 | — | — | — | — | — | — | |
| MaskDiTDiffusion Space=Latent diffusion, Architecture=Transformer, Epochs=8002026.01 | 2.5 | 5.1 | — | 256.3 | 0.83 | 0.56 | — | |
| Large-DiT-3B-GCFG=true, NFE=250x2, #Params=3B2024.10 | 2.52 | — | — | — | — | — | — | |
| Ours-L/32Params (M)=520, GFLOPs=84.8, Epoch=6002026.05 | 2.57 | — | — | — | — | — | — | |
| SiT-XLCFG=true, NFE=250x2, #Params=675M2024.10 | 2.62 | — | — | — | — | — | — | |
| SiT-XL/2Diffusion Space=Latent diffusion, Architecture=Transformer, Epochs=6002026.01 | 2.62 | 4.18 | — | 252.2 | 0.84 | 0.57 | — | |
| VAR-d36-sCFG=true, NFE=10x2, #Params=2.3B2024.10 | 2.63 | — | — | — | — | — | — | |
| VDM++CFG=true, NFE=512x2, #Params=2B2024.10 | 2.65 | — | — | — | — | — | — | |
| VDM++Diffusion Space=Pixel diffusion2026.01 | 2.65 | — | — | 278.1 | — | — | — | |
| DiffiTClass=Diffusion2023.12 | 2.67 | — | — | 252.12 | 83 | 55 | — | |
| DiffiTCFG=true, NFE=250x2, #Params=561M2024.10 | 2.67 | — | — | — | — | — | — | |
| SiD2ACFG=N, NFE=1, Batch Size=16, Reproduced=true, Model Scale=XS, Parameters=125M2026.04 | 2.684 | — | — | — | — | — | — | |
| JiT-L/32Params (M)=461, GFLOPs=84.8, Epoch=6002026.05 | 2.69 | — | — | — | — | — | — | |
| MARCFG=false, NFE=100, #Params=481M2024.10 | 2.74 | — | — | — | — | — | — | |
| DiS-H/2CFG=true, NFE=250x2, #Params=900M2024.10 | 2.88 | — | — | — | — | — | — | |
| DiMR-XL/3RCFG=true, NFE=250x2, #Params=525M2024.10 | 2.89 | — | — | — | — | — | — | |
| EDM2CFG=Y, NFE=63×2, Batch Size=–, Model Scale=XS, Parameters=125M2026.04 | 2.91 | — | — | — | — | — | — | |
| DRWKV-H/2CFG=true, NFE=250x2, #Params=779M2024.10 | 2.95 | — | — | — | — | — | — | |
| VDM++CFG=false, NFE=512, #Params=2B2024.10 | 2.99 | — | — | — | — | — | — | |
| DiT-XL/2-GResolution=512x512, Number of parameters=675.0M2023.03 | 3.04 | 5.04 | — | 240.82 | 84 | 54 | — | |
| DiT-XL/2-GClass=Diffusion2023.12 | 3.04 | — | — | 240.82 | 84 | 54 | — | |
| DiT-XL/2CFG=true, NFE=250x2, #Params=675M2024.10 | 3.04 | — | — | — | — | — | — | |
| DiT-XL/2Diffusion Space=Latent diffusion, Architecture=Transformer, Epochs=6002026.01 | 3.04 | 5.02 | — | 240.8 | 0.84 | 0.54 | — | |
| MAGVIT-v2CFG=false, NFE=64, #Params=307M2024.10 | 3.07 | — | — | — | — | — | — | |
| Ours-L/32Params (M)=520, GFLOPs=84.8, Epoch=2002026.05 | 3.08 | — | — | — | — | — | — | |
| JiT-L/32Params (M)=461, GFLOPs=84.8, Epoch=2002026.05 | 3.28 | — | — | — | — | — | — | |
| SiDCFG=N, NFE=1, Batch Size=2048, Model Scale=XS, Parameters=125M2026.04 | 3.353 | — | — | — | — | — | — | |
| DiffuSSM-XL-GCFG=true, NFE=250x2, #Params=660M2024.10 | 3.41 | — | — | — | — | — | — | |
| EDM2CFG=N, NFE=63, Batch Size=–, Model Scale=XS, Parameters=125M2026.04 | 3.53 | — | — | — | — | — | — | |
| DyDiT-XLCFG=true, NFE=100x2, #Params=675M2024.10 | 3.61 | — | — | — | — | — | — | |
| LEGO-XL-PGCFG=true, NFE=250x2, #Params=681M2024.10 | 3.74 | — | — | — | — | — | — | |
| DiM-HCFG=true, NFE=250x2, #Params=860M2024.10 | 3.78 | — | — | — | — | — | — | |
| Poly-INRResolution=512x512, Number of parameters=46.0M2023.03 | 3.81 | 5.06 | 54.31 | 267.44 | 70 | 34 | — | |
| ADM-GResolution=512x512, Number of parameters=559.0M2023.03 | 3.85 | 5.86 | 210.83 | 221.72 | 84 | 53 | — | |
| ADM-UCFG=true, NFE=250x2, #Params=>559M2024.10 | 3.85 | — | — | — | — | — | — | |
| Ours-B/32Params (M)=151, GFLOPs=24.3, Epoch=6002026.05 | 3.92 | — | — | — | — | — | — | |
| RINCFG=false, NFE=1000, #Params=320M2024.10 | 3.95 | — | — | — | — | — | — | |
| LEGO-XL-PRCFG=true, NFE=250x2, #Params=681M2024.10 | 3.99 | — | — | — | — | — | — | |
| U-ViT-H/4CFG=true, NFE=250x2, #Params=501M2024.10 | 4.05 | — | — | — | — | — | — | |
| JiT-B/32Params (M)=133, GFLOPs=24.3, Epoch=6002026.05 | 4.12 | — | — | — | — | — | — | |
| MAGVIT-v2CFG=false, NFE=12, #Params=307M2024.10 | 4.61 | — | — | — | — | — | — | |
| U-ViT-L/4CFG=true, NFE=250x2, #Params=287M2024.10 | 4.67 | — | — | — | — | — | — | |
| Ours-B/32Params (M)=151, GFLOPs=24.3, Epoch=2002026.05 | 4.98 | — | — | — | — | — | — | |
| JiT-B/32Params (M)=133, GFLOPs=24.3, Epoch=2002026.05 | 5.84 | — | — | — | — | — | — | |
| MaskGITCFG=false, NFE=12, #Params=227M2024.10 | 7.32 | — | — | — | — | — | — | |
| ADM-GCFG=true, NFE=250x2, #Params=>559M2024.10 | 7.72 | — | — | — | — | — | — | |
| DiMR-XL/3RCFG=false, NFE=250, #Params=525M2024.10 | 7.93 | — | — | — | — | — | — | |
| BigGANResolution=512x512, Number of parameters=164.7M2023.03 | 8.43 | 8.13 | 312 | 177.9 | 88 | 29 | — | |
| BigGAN-deepCFG=false, NFE=1, #Params=160M2024.10 | 8.43 | — | — | — | — | — | — | |
| LEGO-XL-PRCFG=false, NFE=250, #Params=681M2024.10 | 9.01 | — | — | — | — | — | — | |
| ADM-UCFG=false, NFE=250, #Params=>559M2024.10 | 9.96 | — | — | — | — | — | — | |
| MaskDiTCFG=false, NFE=79, #Params=736M2024.10 | 10.79 | — | — | — | — | — | — | |
| DiT-XL/2CFG=false, NFE=250, #Params=675M2024.10 | 12.03 | — | — | — | — | — | — | |
| U-DiT-BCFG=false, NFE=250, #Params=204M2024.10 | 15.39 | — | — | — | — | — | — | |
| ADMResolution=512x512, Number of parameters=559.0M2023.03 | 23.24 | 10.19 | 561.32 | 58.06 | 73 | 60 | — | |
| ADM-GCFG=false, NFE=250, #Params=559M2024.10 | 23.24 | — | — | — | — | — | — | |
| DiT-XL/2Guidance=w/o guidance2026.05 | — | — | — | — | — | — | 12.03 | |
| EOSTok-LGuidance=w/o guidance2026.05 | — | — | — | — | — | — | 1.98 | |
| MaskDiTGuidance=w/o guidance2026.05 | — | — | — | — | — | — | 10.79 | |
| TiTok-B-128Guidance=w/o guidance2026.05 | — | — | — | — | — | — | 4.17 | |
| TiTok-L-64Guidance=w/o guidance2026.05 | — | — | — | — | — | — | 3.99 |