AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

UltraGen: Efficient Ultra-High-Resolution Image Generation with Hierarchical Local Attention

Yuyao Zhang, Yu-Wing Tai

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis

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 hierarchical local-window attention with low-res guidance instead of dense FlashAttention at 8KFLUX.1-dev · 8K generationtotal inference speedup: 10.77x total, 6.93x per iterationImproved
use window-first local attention instead of naive FlashAttentionFLUX.1-dev · 4K generationadditional VRAM: 15.6 GB -> 7.3 GBImproved
generate at 2K-4K with model trained only on 1K dataUltraImageGen · GenEvaloverall: 0.66 -> 0.67No clear effect
let high-res tokens attend to low-res guidance tokens (ablation)UltraImageGen · qualitative 4KartifactsImproved
allow each window to attend to neighboring windows (ablation)UltraImageGen · qualitative 4Kboundary artifacts: under 2% extra computeImproved
vary window size (256/512/1024) (ablation)UltraImageGen · 4KFID: 67.23 / 66.31 / 67.83No clear effect
change guidance ratio from 4 to 2 (ablation)UltraImageGen · 4KFID: 67.23 -> 67.19No clear effect
increase LoRA rank from 4 to 16 (ablation)UltraImageGen · FID at 10K stepsFID: 82.41 -> 76.08Improved
increase LoRA rank from 16 to 64 (ablation)UltraImageGen · FID at 10K stepsFID: 76.08 -> 76.15No clear effect
skip more denoising steps from low-res warm start (ablation)UltraImageGen · 4KPSNR vs full-step: 28.78 (4 steps) -> 26.43 (18 steps), speedup 1.17x -> 2.8xMixed
switch backbone to FLUX.2UltraImageGen · 4KFID: 67.03 -> 61.38Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 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