AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Masked Diffusion Transformer is a Strong Image Synthesizer

Shanghua Gao, Pan Zhou, Ming-Ming Cheng, Shuicheng Yan

ICCV 2023 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis · 114 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add mask latent modeling with asymmetric structure (MDT) to DiTDiT-S/2 · ImageNet 256x256FID-50K: 68.40 -> 53.46 (400k steps)Improved
use MDT-XL/2 instead of DiT-XL/2DiT-XL/2 · ImageNet 256x256FID-50K: 9.62 (7000k) vs 9.60 (1300k)Improved
use asymmetric architecture that processes only unmasked tokens in encoder (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 51.56 -> 50.26Improved
add side-interpolater (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 51.60 -> 50.26Improved
train on both full and unmasked latents instead of unmasked only (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 76.63 -> 50.26Improved
compute loss on all tokens instead of masked tokens only (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 58.35 -> 50.26Improved
add relative positional bias (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 53.56 -> 50.26Improved
increase side-interpolater blocks from 1 to 3 (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 50.26 -> 51.96Hurt
raise mask ratio from 0.3 to 0.8 (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 50.26 -> 53.73Hurt
add Adan optimizer (ablation)MDT-S/2 · ImageNet 256x256FID-50K: 50.26 -> 49.71Improved
add Min-SNR loss weighting (ablation)MDT-S/2 with Adan · ImageNet 256x256FID-50K: 49.71 -> 45.06Improved
add encoder long-shortcuts and decoder dense input-shortcuts (ablation)MDTv2-S/2 · ImageNet 256x256FID-50K: 43.82 -> 41.12 -> 38.41Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy