AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

SpeeDiff: Scalable Pixel-Anchored End-to-End Latent Diffusion Model

Bingliang Zhang, Wenda Chu, Yizhuo Li, Linjie Yang, Yisong Yue, Katherine Bouman, Yang Song, Qiushan Guo

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis · 0 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
backpropagate diffusion loss into VAE encoder without any pixel feedback (vanilla end-to-end) (ablation)joint VAE + diffusion training · ImageNet 256x256, 80 epochsgFID: 13.02 -> 33.95 (vs two-stage baseline)Hurt
add Tweedie Pixel Reconstruction (TPR) loss to end-to-end training (ablation)Vanilla E2E · ImageNet 256x256, 80 epochsgFID: 33.95 -> 5.79Improved
add refined architecture (ViT-VAE and Refined-DiT) to E2E with TPR (ablation)E2E w/ TPR · ImageNet 256x256, 80 epochsgFID: 5.79 -> 3.66Improved
add REPA++ alignment to E2E (ablation)SpeeDiff (w/o REPA++) · ImageNet 256x256, 80 epochsgFID: 3.66 -> 1.69Improved
allow diffusion gradient to flow into VAE (end-to-end) versus stop-gradient, with all components (ablation)SpeeDiff · ImageNet 256x256, 80 epochsgFID: 3.47 -> 1.69Improved
add TPR loss under stop-gradient single-stage training (ablation)Single-stage Baseline · ImageNet 256x256, 80 epochsgFID: 14.21 -> 11.65Improved
use two-stage VAE-then-DiT vs single-stage with stop-gradient (ablation)latent diffusion baseline · ImageNet 256x256, 80 epochsgFID: 13.02 vs 14.21No clear effect
add Latent-REPA and Diff-REPA together (REPA++) instead of either alone (ablation)SpeeDiff-XL · ImageNet 256x256, 80 epochsgFID: 2.28 / 2.15 -> 1.69Improved
use DINOv3-L instead of DINOv2-L as the alignment VFM (ablation)REPA++ · ImageNet 256x256, 80 epochsgFID: 1.76 -> 1.69Improved
use 32 latent channels instead of 16, 64 or 128 (ablation)ViT-VAE-XL + Refined-DiT-XL · ImageNet 256x256, 80 epochsgFID: 2.82 (16), 2.27 (64), 3.26 (128) -> 1.69 (32)Improved
increase latent channels (rFID improves but generation worsens at high counts) (ablation)ViT-VAE-B + Refined-DiT-B · ImageNet 256x256, 80 epochsgFID / rFID: rFID 0.84 -> 0.26 and gFID 6.20 -> 12.06 (32 -> 128 channels)Mixed
train a new diffusion model on the frozen VAE learned by SpeeDiff (ablation)SpeeDiff-XL · ImageNet 256x256, 80 epochsgFID: 1.73 vs 1.69No clear effect

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