Class-conditional Image Generation on ImageNet 256x256
365.4Inception Score (IS)VAR
Evaluation Results
| Method | Links | ||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VARTraining Epochs=350, #Params=2.0B, CFG=true2025.12 | 365.4 | — | — | — | 0.83 | 0.57 | — | — | 1.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAREpochs=350, #Params=2.0B2025.12 | 365.4 | — | — | 1.8 | 0.83 | 0.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAREpochs=350, #Params=2.0B, Guidance=w/ guidance2025.12 | 365.4 | — | — | 1.8 | 0.83 | 0.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAREpochs=350, #Params=2.0B, CFG=true2026.03 | 365.4 | — | — | — | 0.83 | 0.57 | — | — | 1.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VARTraining Epoches=350, #params=2.0B, Guidance type=classifier-free guidance (CFG)2026.03 | 365.4 | — | — | — | 0.83 | 0.57 | — | — | 1.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskGIT-reType=Bidirectional AR, #Params=227M, Steps=82025.12 | 355.6 | — | — | 4.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskGITType=Mask., Order=random, #Params=227M, #Epochs=3002025.09 | 355.6 | — | — | 4.02 | 0.78 | 0.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskGITLatent Dim=16, #Params=227M, Guidance=w/ cfg or re2026.03 | 355.6 | — | — | — | — | — | — | — | 4.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskGITTokenizer=MaskGiT [8], #Param.=675M2025.11 | 355.6 | — | — | — | — | — | — | — | 4.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskGIT-retype=Mask., epochs=300, #params=227M, rejection sampling=true, Classifier-Free Guidance (CFG)=true2026.04 | 355.6 | — | — | — | — | — | — | — | — | 4.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VARModel Category=Autoregressive (discrete)2025.11 | 350.2 | — | — | 1.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VARType=Causal AR (non-raster order), #Params=2.0B, Steps=102025.12 | 350.2 | — | — | 1.92 | 0.82 | 0.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAREpochs=350, Parameters=2.0B, Guidance=CFG, External pretrained models=false2025.12 | 350.2 | — | — | — | 0.82 | 0.6 | — | — | 1.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FractalMAR-HGeneration Space=Pixel Space2025.04 | 348.9 | — | — | 6.15 | 0.81 | 0.46 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Flow-DPMNumber of Function Evaluations (NFE)=152026.01 | 341.85 | — | — | 2.32 | 0.83 | 0.58 | 5.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskBitType=NAR, Number of Parameters=305M, Sequence Length=256, Sampling Steps=642025.03 | 341.8 | — | — | 1.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskBitAR=false, # params=305M, # steps=64, # Gflops=10.3k, Classifier-free guidance=true, Extra training data=false2026.04 | 341.8 | — | — | 1.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ARPG-XXL#Para.=1.3B, #Steps=642025.07 | 339.7 | — | — | 1.94 | 0.81 | 0.59 | — | 26.23 | — | — | — | — | 2.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskBit#Para.=305M, #Steps=642025.07 | 338.7 | — | — | 1.62 | — | — | — | 5.39 | — | — | — | — | 1.03 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LPD-XXL#Para.=1.4B, #Steps=202025.07 | 337.6 | — | — | 2 | 0.8 | 0.6 | — | 45.07 | — | — | — | — | 0.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VARLatent Dim=32, #Params=2.0B, Guidance=w/ cfg or re2026.03 | 334.7 | — | — | — | 0.81 | 0.61 | — | — | 1.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30type=VAR, epochs=350, #params=2.0B, Classifier-Free Guidance (CFG)=true2026.04 | 334.7 | — | — | — | 0.81 | 0.61 | — | — | — | 1.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MVAR-d30type=VAR, epochs=350, #params=3.0B, Classifier-Free Guidance (CFG)=true2026.04 | 331.2 | — | — | — | 0.83 | 0.61 | — | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ARPG-XL#Para.=719M, #Steps=642025.07 | 331 | — | — | 2.1 | 0.79 | 0.61 | — | 36.53 | — | — | — | — | 1.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskBittype=Mask., epochs=1080, #params=305M, Classifier-Free Guidance (CFG)=true2026.04 | 328.6 | — | — | — | — | — | — | — | — | 1.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Flow-UniPCNumber of Function Evaluations (NFE)=152026.01 | 327.72 | — | — | 2.59 | 0.82 | 0.58 | 8.37 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Euler-SolverNumber of Function Evaluations (NFE)=152026.01 | 326.91 | — | — | 2.56 | 0.82 | 0.57 | 4.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LPD-XL#Para.=752M, #Steps=202025.07 | 326.7 | — | — | 2.1 | 0.8 | 0.59 | — | 75.2 | — | — | — | — | 0.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BA-solverNumber of Function Evaluations (NFE)=152026.01 | 326.67 | — | — | 1.65 | 0.81 | 0.62 | 4.43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAR-XLType=Causal AR (non-raster order), #Params=1.5B, Steps=2562025.12 | 326 | — | — | 1.48 | 0.8 | 0.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAR-XXLType=AR, Order=hybrid, Tokenizer=VQ, #Params=1.5B2026.02 | 326 | — | — | 1.48 | 0.8 | 0.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAR-XXL#Para.=1.5B, #Steps=2562025.07 | 326 | — | — | 1.48 | 0.8 | 0.63 | — | 6.72 | — | — | — | — | 6.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdaGen-VAR-d30Type=AR, Params=2.0B, Steps=10, TFLOPs=2.02026.03 | 324.3 | — | — | — | — | — | — | — | — | 1.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RQ-Trans.Model Category=Autoregressive (discrete)2025.11 | 323.7 | — | — | 3.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RQTran.-reType=Causal AR (raster order), #Params=3.8B, Steps=642025.12 | 323.7 | — | — | 3.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RQTran.-re#Para.=3.8B, #Steps=2562025.07 | 323.7 | — | — | 3.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RQTran.Latent Dim=16, #Params=3.8B, Guidance=w/ cfg or re2026.03 | 323.7 | — | — | — | — | — | — | — | 3.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RQTran-retype=AR, epochs=50, #params=3.8B, rejection sampling=true, Classifier-Free Guidance (CFG)=true2026.04 | 323.7 | — | — | — | — | — | — | — | — | 3.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAREpochs=350, Parameters=2.0B, Guidance=None, External pretrained models=false2025.12 | 323.1 | — | — | — | 0.82 | 0.59 | — | — | 1.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30#params=2.0B, Classifier-Free Guidance (CFG)=true, Semantics distillation=false2025.07 | 323.1 | — | — | 1.92 | 0.82 | 0.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30Type=VAR, Order=-, Tokenizer=VQ, #Params=2B2026.02 | 323.1 | — | — | 1.92 | 0.82 | 0.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30Tokenizer=-, # token × # dim=256 × 32, Training Epochs=350, CFG=true2025.09 | 323.1 | — | — | — | 0.82 | 0.59 | — | — | 1.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30Type=N.S., #Params=2B, #Epochs=3502025.09 | 323.1 | — | — | 1.92 | 0.82 | 0.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30Type=AR, Params=2.0B, Steps=10, TFLOPs=2.02026.03 | 323.1 | — | — | — | — | — | — | — | — | 1.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30#Para.=2.0B, #Steps=102025.07 | 323.1 | — | — | 1.92 | 0.82 | 0.59 | — | 25.89 | — | — | — | — | 0.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30#Params=2B+61M, NFE=10×22026.05 | 323.1 | — | — | 1.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XXL#params=1.4B, Classifier-Free Guidance (CFG)=true, Semantics distillation=false2025.07 | 322 | — | — | 2.15 | 0.79 | 0.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XXLType=AR, Order=random, Tokenizer=VQ, #Params=1.4B2026.02 | 322 | — | — | 2.15 | 0.79 | 0.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XXLType=X-AR, #Para=1.4B, #Train=300e, #Step=882025.08 | 322 | — | — | 2.15 | 0.79 | 0.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XXL#Para.=1.4B, #Steps=882025.07 | 322 | — | — | 2.15 | 0.79 | 0.62 | — | 11.49 | — | — | — | — | 3.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XXLType=Causal AR (non-raster order), #Params=1.4B, Steps=882025.12 | 321.97 | — | — | 2.15 | 0.79 | 0.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BA-solverNumber of Function Evaluations (NFE)=72026.01 | 320.49 | — | — | 1.96 | 0.78 | 0.64 | 6.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2Type=NAR, Number of Parameters=307M, Sequence Length=256, Sampling Steps=642025.03 | 319.4 | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2Type=Bidirectional AR, #Params=307M, Steps=642025.12 | 319.4 | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MagViT-v2Training Epochs=1080, #Params=307M, CFG=true2025.12 | 319.4 | — | — | — | — | — | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2Type=Mask., Order=random, Tokenizer=LFQ, #Params=307M2026.02 | 319.4 | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MagViT-v2Tokenizer=-, # token × # dim=256 × 5, Training Epochs=1080, CFG=true2025.09 | 319.4 | — | — | — | — | — | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2Type=Mask, #Para=307M, #Train=1080e, #Step=642025.08 | 319.4 | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2Type=Mask., Order=random, #Params=307M, #Epochs=2702025.09 | 319.4 | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2Type=Mask., Params=307M, Steps=642026.03 | 319.4 | — | — | — | — | — | — | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVIT-v2#Para.=307M, #Steps=642025.07 | 319.4 | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LPD-XL#Para.=752M, #Steps=322025.07 | 319.4 | — | — | 1.92 | 0.79 | 0.61 | — | 61.24 | — | — | — | — | 0.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MagViT-v2Epochs=1080, #Params=307M, CFG=true2026.03 | 319.4 | — | — | — | — | — | — | — | 1.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XL#Para.=775M, #Steps=882025.07 | 317.8 | — | — | 2.25 | 0.8 | 0.6 | — | 17.06 | — | — | — | — | 2.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XLType=Causal AR (non-raster order), #Params=775M, Steps=882025.12 | 317.77 | — | — | 2.25 | 0.8 | 0.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BigGAN-deepResolution=256x256, Selection strategy=max IS2021.05 | 317 | 27 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NVG-d24Type=Ours*, Architecture=856M+215M, #Para=1.1B, #Train=350e, #Step=92025.08 | 317 | — | — | 2.06 | 0.79 | 0.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Raster Counterpart-XXL#Para.=1.4B, #Steps=2562025.07 | 316 | — | — | 2.01 | 0.8 | 0.59 | — | 8.99 | — | — | — | — | 7.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Training Epochs=1080, #Params=675M, CFG=true2025.12 | 314.7 | — | — | — | 0.79 | 0.65 | 4.52 | — | 1.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Epochs=1080, #Params=675M2025.12 | 314.7 | — | — | 1.58 | 0.79 | 0.65 | 4.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Epochs=1080, #Params=675M, Guidance=w/ guidance2025.12 | 314.7 | — | — | 1.58 | 0.79 | 0.65 | 4.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Epochs=1080, Parameters=675M, Guidance=CFG, External pretrained models=false2025.12 | 314.7 | — | — | — | 0.79 | 0.65 | — | — | 1.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Type=Mask.+Diff., Order=-, Tokenizer=VAE, #Params=675M2026.02 | 314.7 | — | — | 1.58 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2-XL/2Epochs=1080, CFG scheduling=true2025.10 | 314.7 | — | — | 1.58 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Tokenizer=SD-VAE, # token × # dim=1024 × 4, Training Epochs=1080, CFG=true2025.09 | 314.7 | — | — | — | 0.79 | 0.65 | — | — | 1.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Epochs=1080, #Params=675M, CFG=true2026.03 | 314.7 | — | — | — | 0.79 | 0.65 | — | — | 1.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Training Epoches=1080, #params=675M, Guidance type=classifier-free guidance (CFG)2026.03 | 314.7 | — | — | — | 0.79 | 0.65 | 4.52 | — | 1.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2-XL/2type=Diff., epochs=800, #params=676M, Classifier-Free Guidance (CFG)=true2026.04 | 314.7 | — | — | — | 0.79 | 0.65 | — | — | — | 1.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2Epochs=1080, Parameters=675M, AutoGuidance=true2026.06 | 314.7 | — | — | 1.58 | 0.79 | 0.65 | 4.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Heun-SolverNumber of Function Evaluations (NFE)=152026.01 | 314.36 | — | — | 1.99 | 0.81 | 0.61 | 5.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XLType=Causal AR (non-raster order), #Params=775M, Steps=2562025.12 | 314.21 | — | — | 2.22 | 0.8 | 0.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XLType=AR, Order=random, Tokenizer=VQ, #Params=775M2026.02 | 314.2 | — | — | 2.22 | 0.8 | 0.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandAR-XLType=AR, Order=random, #Params=775M, #Epochs=3002025.09 | 314.2 | — | — | 2.22 | 0.8 | 0.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d24Type=VAR, #Para=1.0B, #Train=350e, #Step=102025.08 | 312.9 | — | — | 2.09 | 0.82 | 0.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d24#Para.=1.0B, #Steps=102025.07 | 312.9 | — | — | 2.09 | 0.82 | 0.59 | — | 39.3 | — | — | — | — | 0.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MergeTok-BLGenerator Method=SiT-XL/1, Training Epochs=800, #Param.=675M, CFG=true2026.05 | 311.7 | — | — | — | — | — | — | — | 1.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SRAEpochs=800, CFG scheduling=true2025.10 | 311.4 | — | — | 1.58 | 0.8 | 0.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEPA-d16Type=VAR, Param=585M, Step=10, Epoch=2002026.07 | 311.26 | — | — | 2.32 | 0.82 | 0.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL + MAETok#params=675M, Classifier-Free Guidance (CFG)=true, Semantics distillation=true2025.07 | 311.2 | — | — | 1.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDT-XLParams=675M, Epochs=400, CFG=true2025.04 | 310.6 | — | — | 1.26 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDT#Params=675M, Epochs=400, Guidance=true2026.02 | 310.6 | — | — | 1.26 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDTEpochs=400, #Params=675M2025.12 | 310.6 | — | — | 1.26 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDTEpochs=400, #Params=675M, Guidance=w/ guidance2025.12 | 310.6 | — | — | 1.26 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDTEpochs=400, Parameters=675M, Guidance=CFG, External pretrained models=true2025.12 | 310.6 | — | — | — | 0.79 | 0.65 | — | — | 1.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDT-XL#params=675M, Classifier-Free Guidance (CFG)=true, Semantics distillation=true2025.07 | 310.6 | — | — | 1.26 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDTEpochs=400, Parameters=675M, AutoGuidance=true2026.06 | 310.6 | — | — | 1.26 | 0.79 | 0.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDT-LParams=458M, Epochs=80, CFG=true2025.04 | 310.5 | — | — | 1.64 | 0.81 | 0.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NVG-d20Type=Ours*, Architecture=497M+125M, #Para=622M, #Train=250e, #Step=92025.08 | 310.4 | — | — | 2.44 | 0.8 | 0.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAE (DiT-XL)Training Epochs=800, #Params=676M, CFG=true2025.12 | 309.4 | — | — | — | 0.8 | 0.63 | — | — | 1.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAE (DiT-XL)Epochs=800, #Params=676M2025.12 | 309.4 | — | — | 1.41 | 0.8 | 0.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |