List your company

Image Generation on ImageNet-1K 256x256 (train val)

2.06FIDSiT-XL
-2.677229.298961.27593.2511Sep 28, 2025Nov 7, 2025Dec 17, 2025Jan 26, 2026Mar 7, 2026Apr 16, 2026May 26, 2026

Evaluation results

MethodLinks
SiT-XLSampling Steps=250, Param Count=675M, Epochs Trained=6402026.052.06———
DiT-XLSampling Steps=500, Param Count=675M, Epochs Trained=6402026.052.27———
RecFM-XLSampling Steps=16, Param Count=675M, Epochs Trained=1602026.052.49———
RecFM-XLSampling Steps=128, Param Count=675M, Epochs Trained=1602026.052.53———
RecFM-XLSampling Steps=8, Param Count=675M, Epochs Trained=1602026.053.22———
LDM-4-GSampling Steps=500, Param Count=400M, Epochs Trained=1062026.053.6———
Shortcut Model (XL)Sampling Steps=128, Param Count=676M, Epochs Trained=2502026.053.8———
ADM-GSampling Steps=250, Epochs Trained=4262026.054.59———
FPsteps=50, cfg=1.5, Bit-width (W/A)=32/322025.096.62239.521.178.49
Baselinesteps=50, cfg=4.0, Bit-width (W/A)=1.58/32025.0910.83254.9221.6885.85
RobuQsteps=50, cfg=4.0, Bit-width (W/A)=1.58/22025.0911.06273.5821.5787.51
RobuQ (w/o AMP)steps=150, cfg=4.0, Bit-width (W/A)=1.58/22025.0911.17220.4720.185.73
Baselinesteps=150, cfg=4.0, Bit-width (W/A)=1.58/32025.0911.55271.2720.7688.76
Bitnetv2steps=50, cfg=4.0, Bit-width (W/A)=1.58/42025.0911.69196.7821.4483.7
Bitnetv2steps=150, cfg=4.0, Bit-width (W/A)=1.58/42025.0911.72206.2820.6285.91
RobuQsteps=150, cfg=4.0, Bit-width (W/A)=1.58/22025.0911.86281.7321.9489.22
RobuQ (w/o AMP)steps=50, cfg=4.0, Bit-width (W/A)=1.58/22025.0911.9720021.7381.88
RobuQ (w/o AMP)steps=50, cfg=4.0, Bit-width (W/A)=1.58/32025.0912.31325.5619.9490.53
RobuQ (w/o AMP)steps=50, cfg=4.0, Bit-width (W/A)=1.58/42025.0912.64349.2219.6991.86
RobuQsteps=50, cfg=4.0, Bit-width (W/A)=1.58/32025.0912.71342.9419.8791.29
Baselinesteps=50, cfg=4.0, Bit-width (W/A)=1.58/42025.0912.82342.0720.0590.92
RobuQ (w/o AMP)steps=150, cfg=4.0, Bit-width (W/A)=1.58/32025.0913.65333.6220.7591.8
RobuQ (w/o AMP)steps=150, cfg=4.0, Bit-width (W/A)=1.58/42025.0913.82348.420.3292.25
Baselinesteps=150, cfg=4.0, Bit-width (W/A)=1.58/42025.0913.92344.4320.7591.67
RobuQsteps=150, cfg=4.0, Bit-width (W/A)=1.58/32025.0914.27342.7320.6392.47
BinaryDMsteps=150, cfg=4.0, Bit-width (W/A)=1.58/42025.0915.5124.7821.6675.75
BinaryDMsteps=50, cfg=4.0, Bit-width (W/A)=1.58/42025.0917.08115.5223.1572.3
RobuQ (w/o AMP)steps=50, cfg=1.5, Bit-width (W/A)=1.58/42025.0917.97103.2426.9565.77
FPsteps=50, cfg=4.0, Bit-width (W/A)=32/322025.0919.11478.3521.6192.98
FPsteps=150, cfg=4.0, Bit-width (W/A)=32/322025.0919.67479.7222.9493.01
Baselinesteps=50, cfg=1.5, Bit-width (W/A)=1.58/42025.0920.8295.0727.5361.52
RobuQsteps=50, cfg=1.5, Bit-width (W/A)=1.58/32025.0921.493.7526.9961.9
RobuQ (w/o AMP)steps=50, cfg=1.5, Bit-width (W/A)=1.58/32025.0924.4483.8429.1860.01
RobuQsteps=50, cfg=1.5, Bit-width (W/A)=1.58/22025.0930.366.7430.6656.8
Baselinesteps=50, cfg=1.5, Bit-width (W/A)=1.58/32025.0940.2351.3135.6449.46
Bitnetv2steps=50, cfg=1.5, Bit-width (W/A)=1.58/42025.0941.5944.3234.0950.02
RobuQ (w/o AMP)steps=50, cfg=1.5, Bit-width (W/A)=1.58/22025.0943.3145.6538.8949.17
BinaryDMsteps=50, cfg=1.5, Bit-width (W/A)=1.58/42025.0962.9125.6338.2837.65
Baselinesteps=150, cfg=4.0, Bit-width (W/A)=1.58/22025.0968.0423.2230.8429.99
Baselinesteps=50, cfg=4.0, Bit-width (W/A)=1.58/22025.0975.0220.6535.8128.12
Baselinesteps=50, cfg=1.5, Bit-width (W/A)=1.58/22025.09120.4910.6362.2920.91