List your company

Class-conditional Image Generation on ImageNet 512x512

1.73FIDMAR-L, Diff Loss
1.50643.01574.5256.0343Sep 25, 2022May 12, 2023Dec 27, 2023Aug 12, 2024Mar 29, 2025Nov 13, 2025Jul 1, 2026

Evaluation results

MethodLinks
MAR-L, Diff LossToken Type=continuous-valued tokens, #params=481M, CFG=true2024.061.73279.9——————
MAR-LType=Mask, #Para.=481M2025.071.73279.9——————
SiT-XL/2 + IGEpochs=602025.121.78286.30.810.62—3.92——
EDM2-XXLToken Type=continuous-valued tokens, #params=1.5B, CFG=true2024.061.81———————
MAGVIT-v2Token Type=vector-quantized tokens, #params=307M, CFG=true2024.061.91324.3——————
EDM2-XXLToken Type=continuous-valued tokens, #params=1.5B, CFG=false2024.061.91———————
MAGVIT-v2Type=Mask, #Para.=307M, #Steps=642025.071.91324.3——————
FreqFlow-L#params.=507M2026.042.02285.3——————
SiT-XL/2 + REPAEpochs=2002025.122.08274.60.830.58—4.19——
Raster Counterpart-XLType=AR, #Para.=752M, #Steps=1024, Latency(s)=20.93, Throughput(img/s)=2.362025.072.093158157————
LPD-XLType=Parallelized AR, #Para.=752M, #Steps=48, Latency(s)=1.01, Throughput(img/s)=18.182025.072.13268063————
TiTok-BTok. Model=VQ, # Tokens=128, Classifier-Free Guidance (CFG)=true2024.122.13261.2——1.52———
SiT-XLTok. Model=SoftVQ-BL, # Tokens=64, Gflops=29.23, Throughput (imgs/sec)=5.38, Classifier-Free Guidance (CFG)=true2024.122.21290.5——0.71———
UVIT-2B+Tok. Model=AE, # Tokens=64, Gflops=9.44, Classifier-Free Guidance (CFG)=true2024.122.25———104.18———
SiT-XL/2 + REPAEpochs=1002025.122.32255.70.840.56—4.16——
StyleGAN-XLArchitecture Type=GAN2022.092.41———————
StyleGAN-XLClassifier-Free Guidance (CFG)=true2024.122.41267.7——————
StyleGAN-XLParams(M)=1662026.032.41—7752————
StyleGAN-XL2026.042.41267.8——————
SiT-XL/2 + REPAEpochs=802025.122.44247.30.840.56—4.21——
MPDiT-XLParams(M)=482, GFLOPs=228.42026.032.47—8356————
MaskDiTEpochs=8002025.122.5256.30.830.56—5.1——
ComposerBackbone=VAR d-362026.032.51305.7————1.0721.96
UVIT-H+Tok. Model=AE, # Tokens=256, Gflops=6.14, Classifier-Free Guidance (CFG)=true2024.122.53———128.9———
DiCo-XLParams(M)=701, GFLOPs=349.82026.032.53—8356————
Raster Counterpart-LType=AR, #Para.=337M, #Steps=1024, Latency(s)=14.25, Throughput(img/s)=3.792025.072.54278.58058————
LPD-LType=Parallelized AR, #Para.=337M, #Steps=48, Latency(s)=0.69, Throughput(img/s)=35.162025.072.54292.28155————
Test-time TrainingBackbone=VAR d-362026.032.54305.4————112.3732.45
SiT-XL/2Tok. Model=KL, # Tokens=4096, Gflops=0.10, Classifier-Free Guidance (CFG)=true2024.122.62252.2——373.32———
SiT-XL/2Epochs=6002025.122.62252.20.840.57—4.18——
SiT-XL/2Type=Diffusion, #Para.=675M, #Steps=2502025.072.62252.218457————
SiT-XLParams(M)=675, GFLOPs=524.72026.032.62—8457————
VAR-d36-sType=VAR, Epoch=350, #Params=2.35B, Gflops=-2024.062.63303.2——————
VAR-d36-sType=VAR, #Para.=2.3B, #Steps=10, Latency(s)=0.45, Throughput(img/s)=OOM2025.072.63303.2——————
VAR-d36-sParams(M)=23002026.032.63———————
StandardBackbone=VAR d-362026.032.63303.2————121.56
VAR-d36-s#params.=2.35B2026.042.63303.2——————
VAR-d36-sEpoch=3502026.072.63303.2——————
VDM++Token Type=pixel-based, #params=2B, CFG=true2024.062.65278.1——————
VDM++2025.122.65278.1——————
VDM++Params(M)=20002026.032.65———————
UVIT-H+Tok. Model=AE, # Tokens=64, Gflops=15.23, Classifier-Free Guidance (CFG)=true2024.122.66———32.99———
DiffiTToken Type=continuous-valued tokens, CFG=true2024.062.67252.1——————
DiffiTType=Diff., Epoch=800, #Params=561M, Gflops=-2024.062.67252.18355————
DiffiT#params.=561M2026.042.67252.1——————
MAR-L, Diff LossToken Type=continuous-valued tokens, #params=481M, CFG=false2024.062.74205.2——————
ComposerBackbone=DiT-XL/22026.032.81248.1————81.038.5
DiT-XLTok. Model=AE, # Tokens=256, Gflops=1.02, Classifier-Free Guidance (CFG)=true2024.122.84———80.75———
ADM-G, ADM-UEpochs=4002025.122.85221.70.840.53—5.86——
Test-time TrainingBackbone=DiT-XL/22026.032.87245.2————152.2414.3
DiMR-XL/3RType=Diff., Epoch=800, #Params=525M, Gflops=2062024.062.89289.88355————
DiMR-XL/3R#params.=525M2026.042.89289.8——————
VDM++Token Type=pixel-based, #params=2B, CFG=false2024.062.99232.2——————
PPF-XL-2Size=512, Training steps=400k, Training FLOPs (%)=7.6, Testing FLOPs (%)=58.72025.063.01249.988454—5.24——
DiT-XL/2Token Type=continuous-valued tokens, #params=675M, CFG=true2024.063.04240.8——————
DiT-XL/2Type=Diff., Epoch=599, #Params=675M, Gflops=5252024.063.04240.88454————
DiT-XL/2Tok. Model=KL, # Tokens=4096, Gflops=0.10, Classifier-Free Guidance (CFG)=true2024.123.04240.8——373.34———
DiT-XL/2Epochs=6002025.123.04240.80.840.54—5.02——
DiT-XL/2Size=512, Training steps=-, Training FLOPs (%)=100, Testing FLOPs (%)=1002025.063.04240.828454—5.02——
DiT-XL/2Type=Diffusion, #Para.=675M, #Steps=250, Latency(s)=11.32, Throughput(img/s)=0.102025.073.04240.828454————
DiT-XL/2Params(M)=675, GFLOPs=524.72026.033.04—8454————
StandardBackbone=DiT-XL/22026.033.04240.8————818.3
DiT-XL/2#params.=675M2026.043.04240.8——————
DiT-XL/22026.073.04240.8——————
PPF-XL-3Size=512, Training steps=400k, Training FLOPs (%)=5.8, Testing FLOPs (%)=45.42025.063.06249.918353—5.31——
MAGVIT-v2Token Type=vector-quantized tokens, #params=307M, CFG=false2024.063.07213.1——————
DiMR-XL/3RType=Diff., Epoch=400, #Params=525M, Gflops=2062024.063.23285.18254————
DiT-XL/2 [33]*NFE=50, MACs=26.3T, Speed=1×2026.033.33203.583.4254.5—4.53——
ARPG-XLType=Parallelized AR, #Para.=719M2025.073.38257.8——————
DIFFUSSM-XLType=Diff., Epoch=236, #Params=673M, Gflops=10662024.063.412558549————
DiffuSSM-XLParams(M)=673, GFLOPs=1066.22026.033.41—8549————
DIFFUSSM-XL#params.=673M2026.043.41255——————
MAR-HTok. Model=SoftVQ-BL, # Tokens=64, Gflops=86.55, Throughput (imgs/sec)=1.50, Classifier-Free Guidance (CFG)=true2024.123.42261.8——0.71———
Timestep-Aware Block Masking (Ours)NFE=50, Extra Data=✗, Training Time=2.8h, MACs=17.9T, Speed=1.48×2026.033.64202.883.1854.5—5.06——
L2C [29]*NFE=50, Extra Data=✓, Training Time=18.4h, MACs=16.9T, Speed=1.54×2026.033.76200.783.1654.3—5.1——
DiM-HParams(M)=860, GFLOPs=7082026.033.78———————
ADM-G, ADM-UArchitecture Type=Diffusion based on U-Net, #Params=422M + 43M (Cls) + 309M (SR)2022.093.85———————
ADM-G, ADM-UType=Diff., Epoch=1081, #Params=731M, Gflops=28132024.063.85221.78453————
ADMClassifier-Free Guidance (CFG)=true2024.123.85221.7——————
ADM-UParams(M)=731, GFLOPs=28132026.033.85—8453————
ADM-G, ADM-U#params.=731M2026.043.85221.7——————
U-ViT-H/4Params(M)=501, GFLOPs=133.32026.034.03—8448————
GAT-XL/2Resolution=512×512, Epochs=152025.094.04———————
U-ViT-H/4Architecture Type=Diffusion based on ViT, #Params=501M + 84M (AE), Latent representations & Classifier-free guidance=true2022.094.05———————
U-ViT-H/2-GToken Type=continuous-valued tokens, #params=501M, CFG=true2024.064.05263.8——————
U-ViT-H/4Type=Diff., Epoch=400, #Params=501M, Gflops=1332024.064.05263.88448————
U-ViT-H/4Classifier-Free Guidance (CFG)=true, Sampler=DPM-Solver, Steps=502024.124.05263.8——128.92———
U-ViT-H/4#params.=501M2026.044.05263.8——————
Simple diffusion (U-Net)Epochs=8002025.124.28171——————
MaskGIT-reType=Mask., Epoch=300, #Params=227M, Gflops=-2024.064.46342——————
MaskGIT-re#params.=227M2026.044.46342——————
Simple diffusion (U-ViT, L)Epochs=8002025.124.53205.3——————
Simple DiffParams(M)=20002026.034.53———————
U-ViT-L/4Type=Diff., Epoch=400, #Params=287M, Gflops=772024.064.67213.38745————
U-ViT-L/4Params(M)=287, GFLOPs=76.52026.034.67—8745————
U-ViT-L/4#params.=287M2026.044.67213.3——————
UVIT-2B+Tok. Model=AE, # Tokens=64, Gflops=9.44, Classifier-Free Guidance (CFG)=false2024.126.5———104.18———
MEPA-d20Epoch=662026.077.29199.2——————
MaskGITToken Type=vector-quantized tokens, #params=227M, CFG=false2024.067.32156——————
MaskGITType=Mask., Epoch=300, #Params=227M, Gflops=-2024.067.321567850————