AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

$\textit{MADFormer}$: Mixed Autoregressive and Diffusion Transformers for Continuous Image Generation

Junhao Chen, Yulia Tsvetkov, Xiaochuang Han

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images

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
allocate more layers to AR (e.g. 3:1 AR:diffusion) under low inference NFE (ablation)MADFormer · FFHQ-1024 and ImageNet, low NFEFID: 60-75% lower FIDImproved
allocate more layers to diffusion when compute is high (ablation)MADFormer · high NFE, high resolutionFIDImproved
increase diffusion depth from 7 to 28 layers (ablation)MADFormer · FFHQ-1024FID: 20.2 -> 15.9Improved
increase diffusion depth from 7 to 28 layers (ablation)MADFormer · ImageNet 256FID: 34.0 -> 27.4Improved
split FFHQ-1024 into 16 blocks instead of 4 or 64 (ablation)MADFormer · FFHQ-1024FID: 18.9 -> 17.8Improved
increase AR length (more blocks) (ablation)MADFormer · ImageNet 256FID: 28.4 -> 39.7Hurt
add AR condition on top of clean blocks (ablation)MADFormer · FFHQ-1024FID: 20.1 -> 17.8Improved
add clean blocks on top of AR condition (ablation)MADFormer · ImageNet 256FID: 31.2 -> 30.0Improved
use separate parameter sets for text, clean and noisy blocks (ablation)MADFormer · FFHQ-1024 and ImageNetFID: ImageNet 30.4 -> 30.0No clear effect
keep sequence-level causal attention during diffusion instead of within-block MLP-style denoising (ablation)MADFormer · ImageNet 256FID: 96.5 -> 30.0Improved
add hidden loss (lambda 0.1) (ablation)MADFormer · FFHQ-1024FID: 19.4 -> 17.8Improved
raise hidden loss weight from 0.1 to 1 (ablation)MADFormer · FFHQ-1024FID: 17.8 -> 18.4Hurt

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