Class-conditional Image Generation on ImageNet class-conditional 256x256
337.85Inception Score (IS)SFM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SFMSource=Spherically Projected Gaussian Distribution (N~), Target=Spherically Projected Dataset (D~)2026.05 | 337.85 | 4.62 | — | 80.83 | 48.9 | — | 7.51 | |
| AliTok-LGen Type=AR, Number of Parameters=318M, Clustering (C)=true2026.05 | 326.2 | 1.38 | 1.98 | 78 | 65 | — | — | |
| RAR-XXLGen Type=RandAR, Number of Parameters=1.5B2026.05 | 326 | 1.48 | 3.26 | 80 | 63 | — | — | |
| VAR-d30Gen Type=VAR, Number of Parameters=2.0B2026.05 | 323.1 | 1.92 | — | 82 | 59 | — | — | |
| AliTok-BGen Type=AR, Number of Parameters=177M, Clustering (C)=true2026.05 | 319.5 | 1.44 | 2.4 | 77 | 65 | — | — | |
| MAGVIT-v2Gen Type=Mask., Number of Parameters=307M2026.05 | 319.4 | 1.78 | 3.65 | — | — | — | — | |
| RAR-XLGen Type=RandAR, Number of Parameters=955M2026.05 | 306.9 | 1.5 | 3.72 | 80 | 62 | — | — | |
| VCQ B-XLGen Type=AR, Number of Parameters=684M2026.05 | 306.9 | 2.38 | 6.82 | 80.2 | 58.7 | — | — | |
| AliTok-XLGen Type=AR, Number of Parameters=662M, Clustering (C)=true2026.05 | 306.3 | 1.28 | 1.88 | 79 | 65 | — | — | |
| REPA (SiT-XL)Gen Type=Diff., Number of Parameters=675M, Semantic/VFM (S)=true, Diff./FM (D)=true2026.05 | 305.7 | 1.42 | 5.9 | 80 | 65 | — | — | |
| MAR-HGen Type=M+D, Number of Parameters=943M, Diff./FM (D)=true2026.05 | 303.7 | 1.55 | 2.35 | 81 | 62 | — | — | |
| LightningDiTGen Type=Diff., Number of Parameters=675M, Semantic/VFM (S)=true, Diff./FM (D)=true2026.05 | 295.3 | 1.35 | 2.17 | 79 | 65 | — | — | |
| ImageFolderGen Type=VAR, Number of Parameters=362M, Semantic/VFM (S)=true, Dropout (T)=true2026.05 | 295 | 2.6 | — | 75 | 63 | — | — | |
| SpectralARGen Type=AR, Number of Parameters=310M2026.05 | 282.2 | 3.02 | — | 81 | 55 | — | — | |
| TiTok-S-128Gen Type=Mask., Number of Parameters=287M2026.05 | 281.8 | 1.97 | 4.44 | — | — | — | — | |
| MAR-BGen Type=M+D, Number of Parameters=208M, Diff./FM (D)=true2026.05 | 281.7 | 2.31 | 3.48 | 82 | 57 | — | — | |
| VFMTok-LGen Type=AR, Number of Parameters=343M, Semantic/VFM (S)=true2026.05 | 278.8 | 2.75 | 2.11 | 84 | 57 | — | — | |
| DiT-XL/2Gen Type=Diff., Number of Parameters=675M, Diff./FM (D)=true2026.05 | 278.2 | 2.27 | 9.62 | 83 | 57 | — | — | |
| FlowAR-BGen Type=Flow+VAR, Number of Parameters=300M, Diff./FM (D)=true2026.05 | 272.5 | 2.9 | — | 84 | 54 | — | — | |
| SiT-XLGen Type=Diff., Number of Parameters=675M, Diff./FM (D)=true2026.05 | 270.3 | 2.06 | 8.61 | 82 | 59 | — | — | |
| VCQ B-LGen Type=AR, Number of Parameters=343M2026.05 | 258.3 | 2.75 | 7.47 | 80.1 | 57.2 | — | — | |
| ResTokGen Type=HAR, Number of Parameters=326M, Semantic/VFM (S)=true, Dropout (T)=true2026.05 | 257.8 | 2.34 | — | 79 | 60 | — | — | |
| LlamaGen-XXLGen Type=AR, Number of Parameters=1.4B2026.05 | 253.9 | 2.34 | — | 80 | 59 | — | — | |
| VCQ L-XLGen Type=AR, Number of Parameters=684M2026.05 | 250.8 | 1.71 | 2.29 | 76.2 | 67 | — | — | |
| OT-CFMSource=Gaussian Distribution (N), Target=Original Dataset (D)2026.05 | 247.71 | 5.22 | — | 82.58 | 45.28 | — | 7.85 | |
| LDM-4-GGen Type=Diff., Number of Parameters=400M, Diff./FM (D)=true2026.05 | 247.7 | 3.6 | 10.56 | 87 | 48 | — | — | |
| L2CBackbone=DiT-XL/2, Sampling steps=50-step DDIM, Latency (s)=1.6494, FLOPs=18.1264, Speed=1.24×2026.05 | 246.47 | — | — | 81 | 58 | 2.25 | 4.28 | |
| L2C + TCCBackbone=DiT-XL/2, Sampling steps=50-step DDIM, Latency (s)=1.7638, FLOPs=18.8494, Speed=1.16×2026.05 | 245.16 | — | — | 81 | 59 | 2.22 | 4.32 | |
| LlamaGen-XLGen Type=AR, Number of Parameters=775M2026.05 | 244.1 | 2.62 | — | 80 | 57 | — | — | |
| VCQ L-LGen Type=AR, Number of Parameters=343M2026.05 | 243.6 | 1.9 | 2.72 | 75.2 | 65.8 | — | — | |
| ReDi2Steps=4, Aux. Score Net=✗2026.05 | 240 | 7.86 | — | 87 | 44 | — | — | |
| DiTBackbone=DiT-XL/2, Sampling steps=50-step DDIM, Latency (s)=2.0435, FLOPs=23.7637, Speed=1.00×2026.05 | 238.21 | — | — | 81 | 60 | 2.26 | 4.32 | |
| FORA(N = 2)Backbone=DiT-XL/2, Sampling steps=50-step DDIM, N=2, Latency (s)=1.1929, FLOPs=11.9085, Speed=1.71×2026.05 | 238.04 | — | — | 80 | 59 | 2.65 | 4.71 | |
| FORA(N = 2) + TCCBackbone=DiT-XL/2, Sampling steps=50-step DDIM, N=2, Latency (s)=1.3163, FLOPs=12.6695, Speed=1.55×2026.05 | 237 | — | — | 80 | 59 | 2.49 | 4.69 | |
| ToCa(N = 2)Backbone=DiT-XL/2, Sampling steps=50-step DDIM, N=2, R=93%, Latency (s)=1.5386, FLOPs=13.5404, Speed=1.33×2026.05 | 236.57 | — | — | 80 | 58 | 2.6 | 4.43 | |
| I-CFMSource=Gaussian Distribution (N), Target=Original Dataset (D)2026.05 | 236.37 | 5.29 | — | 82.5 | 46.04 | — | 8.24 | |
| I-CFMSource=Gaussian Distribution (N), Target=Spherically Projected Dataset (D~)2026.05 | 236.13 | 5.02 | — | 80.73 | 48.27 | — | 7.98 | |
| SOT-CFMSource=Gaussian Distribution (N), Target=Original Dataset (D)2026.05 | 235.99 | 5.15 | — | 81.81 | 46.54 | — | 7.8 | |
| ToCa(N = 2) + TCCBackbone=DiT-XL/2, Sampling steps=50-step DDIM, N=2, R=93%, Latency (s)=1.6186, FLOPs=14.3775, Speed=1.26×2026.05 | 235.84 | — | — | 80 | 59 | 2.55 | 4.49 | |
| SOT-CFMSource=Gaussian Distribution (N), Target=Spherically Projected Dataset (D~)2026.05 | 234.76 | 5.1 | — | 81.28 | 46.11 | — | 7.69 | |
| ReDi1Steps=4, Aux. Score Net=✗2026.05 | 228 | 7.58 | — | 87 | 46 | — | — | |
| MaskGitSteps=16, Aux. Score Net=-2026.05 | 224 | 6.6 | — | 83 | 40 | — | — | |
| MaskGitSteps=8, Aux. Score Net=-2026.05 | 222 | 6.66 | — | 83 | 40 | — | — | |
| VCQ B-BGen Type=AR, Number of Parameters=115M2026.05 | 221.9 | 4.79 | 14.8 | 82.8 | 45.2 | — | — | |
| DetailFlow-16Gen Type=PAR, Number of Parameters=326M2026.05 | 221.4 | 2.96 | — | 82 | 57 | — | — | |
| GigaTok-B-LGen Type=AR, Number of Parameters=111M, Semantic/VFM (S)=true2026.05 | 221 | 3.26 | — | 81 | 56 | — | — | |
| PAR-L-4xGen Type=PAR, Number of Parameters=343M2026.05 | 218.9 | 3.76 | — | 84 | 50 | — | — | |
| VCQ L-BGen Type=AR, Number of Parameters=115M2026.05 | 218.7 | 2.73 | 5.14 | 77.7 | 59 | — | — | |
| FPD (Ours)Steps=1, Aux. Score Net=✗2026.05 | 215 | 6.9 | — | 81 | 35 | — | — | |
| di4c-dSteps=4, Aux. Score Net=✗2026.05 | 214 | 6.57 | — | — | — | — | — | |
| DiMOSteps=1, Aux. Score Net=✓2026.05 | 214 | 6.91 | — | 83 | 38 | — | — | |
| di4cSteps=4, Aux. Score Net=✗2026.05 | 209 | 6.79 | — | — | — | — | — | |
| SDTTSteps=4, Aux. Score Net=✗2026.05 | 205 | 8.97 | — | 88 | 41 | — | — | |
| TiTok-L-32Gen Type=Mask., Number of Parameters=177M2026.05 | 199.8 | 2.77 | 3.15 | — | — | — | — | |
| MaskGitSteps=4, Aux. Score Net=-2026.05 | 192 | 10.73 | — | 75 | 31 | — | — | |
| MaskGITGen Type=Mask., Number of Parameters=227M2026.05 | 182.1 | — | 6.18 | 80 | 51 | — | — | |
| ReDi3-distillSteps=1, Aux. Score Net=✗2026.05 | 182 | 11.68 | — | 83 | 44 | — | — | |
| RQ-Trans.Gen Type=AR, Number of Parameters=3.8B2026.05 | 134 | — | 7.55 | — | — | — | — | |
| VQGANGen Type=AR, Number of Parameters=1.4B2026.05 | 74.3 | — | 15.78 | — | — | — | — | |
| SDTTSteps=1, Aux. Score Net=✗2026.05 | 14 | 90.4 | — | 31 | 13 | — | — | |
| MaskGitSteps=2, Aux. Score Net=-2026.05 | 13 | 91.35 | — | 18 | 16 | — | — |