Class-conditional Image Generation on ImageNet-1k (train)

1.13FID

GAE

Updated 16d ago

Evaluation Results

Method	Links
GAE 2026.05		1.13
RAE + DiT-XL 2026.05		1.13
REPA-E 2026.05		1.15
LightningDiT-XL/2 2026.05		1.35
SiT-XL/2 + REG 2026.05		1.36
SiT-XL/2 + REPA 2026.05		1.42
mmDiT-XL/2 (ours) 2024.11		1.59
DiT-XL/2 w/Att 2024.11		1.71
mmDiT-XL/2 (SD3) 2024.11		1.71
iMF-XL/2 2026.05		1.72
mDT-v2-XL/2 w/ LN 2024.11		1.79
DiT-XL/2 w/LN 2024.11		1.95
UNet (SDXL) 2024.11		2.05
PixArt-a-XL/2 2024.11		2.06
SiT-XL/2 2026.05		2.06
Sphere Latent Encoder (Ours-XL/2) 2026.05		2.11
Sphere Latent Encoder (Ours-XL/2) 2026.05		2.25
DiT-XL2 w/ LN 2024.11		2.27
DiT-XL/2 2026.05		2.27
mDT-v2-XL/2 w/ LN 2024.11		2.51
α-Flow-XL/2+ 2026.05		2.58
mmDiT-XL/2 (ours) 2024.11		2.76
DiT-XL/2 w/LN 2024.11		2.85
mmDiT-XL/2 (SD3) 2024.11		3.02
DiT-XL2 w/ LN 2024.11		3.04
PixArt-a-XL/2 2024.11		3.05
MeanFlow-XL/2 2026.05		3.43
UNet (SD/LDM-G4) 2024.11		3.6
mDT-v2-XL/2 w/ LN 2024.11		3.75
Sphere Encoder 2026.05		4.02
UNet (SDXL) 2024.11		4.81