AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Diffusion Transformers with Representation Autoencoders

Boyang Zheng, Nanye Ma, Shengbang Tong, Saining Xie

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Latent diffusion with learned tokenizers

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
use standard DiT on RAE latents instead of SD-VAE latents (ablation)DiT-S · ImageNet 256, 80 epochsgFID: 51.74 -> 215.76Hurt
widen DiT to at least the RAE token dimension (ablation)DiT on DINOv2 RAE latents · single-image overfittingoverfitting loss: DiT-S on DINOv2-B 5.2e-1 vs DiT-B on DINOv2-B 2.4e-2Improved
increase depth from 12 to 24 at width 384 (ablation)SiT-S on RAE latent · single-image overfittingsample qualityNo clear effect
add dimension-dependent noise schedule shift (ablation)DiT-XL on RAE (DINOv2-B) · ImageNet 256, 80 epochsgFID: 23.0750453 -> 4.81Improved
add noise-augmented decoder training (ablation)RAE decoder · ImageNet 256, 80 epochsgFID: 4.81 -> 4.28Improved
add noise-augmented decoder training (ablation)RAE decoder with MAE-B encoder · ImageNet 256, 80 epochsgFID: 16.14 -> 8.38Improved
add shallow wide DDT head to DiT (ablation)DiT-XL on RAE (DINOv2-B) · ImageNet 256, 80 epochsgFID: 4.28 -> 2.16Improved
add shallow wide DDT head to DiT (ablation)DiT-XL on RAE (DINOv2-L) · ImageNet 256, 80 epochsgFID: 6.09 -> 2.73Improved
add DDT head on SD-VAE latents (ablation)DiT-XL on SD-VAE · ImageNet 256, 80 epochsgFID: 7.13 -> 11.70Hurt
diffuse in raw pixels instead of RAE latents (ablation)DiT-DH-XL · ImageNet 256, 80 epochsgFID: 2.16 -> 30.56Hurt
use decoder upsampling from 256 tokens instead of direct 512 resolution training (ablation)RAE DiT-DH-XL · ImageNet 512, 400 epochsgFID: 1.13 -> 1.61Mixed
use larger decoder (ViT-B to ViT-XL) (ablation)RAE DINOv2-B · ImageNet 256 reconstructionrFID: 0.58 -> 0.49Improved

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