AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Toward Diffusible High-Dimensional Latent Spaces: A Frequency Perspective

Bolin Lai, XuDong Wang, Saketh Rambhatla, James M. Rehg, Zsolt Kira, Rohit Girdhar, Ishan Misra

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis · 7 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 FreqWarm frequency warm-up to diffusion trainingUSiT-H on Wan2.2-AE · ImageNet 512x512 w/o CFGgFID: 43.67 -> 29.56Improved
add FreqWarm frequency warm-upUSiT-H on LTX-AE · ImageNet 512x512 w/o CFGgFID: 24.18 -> 18.05Improved
add FreqWarm frequency warm-upUSiT-H on DC-AE-f32c128 · ImageNet 512x512 w/o CFGgFID: 13.84 -> 9.42Improved
add FreqWarm frequency warm-upUSiT-2B on DC-AE-f64c128 · ImageNet 512x512 w/o CFGgFID: 5.67 -> 4.77Improved
add FreqWarm frequency warm-upDiT-XL on DC-AE-f32c128 · ImageNet 512x512 with CFGgFID: 3.11 -> 2.87Improved
add FreqWarm frequency warm-upDiT, UViT, USiT on DC-AE-f32c128 · ImageNet 256x256gFID: 26.30 -> 17.89 (DiT)Improved
increase latent channels from 16 to 512 with FreqWarm (ablation)DC-AE f32 · ImageNet 512x512gFID gain: 0.02 at c16 -> 12.18 at c512Improved
apply FreqWarm to low-channel autoencoder (ablation)DC-AE-f32c16 · ImageNet 512x512gFID: 12.59 -> 12.57No clear effect
use FreqWarm threshold 0.2 instead of 0.05 (ablation)DC-AE-f32c128 + USiT-H · ImageNet 512x512gFID: 23.11 -> 9.42Improved
use FreqWarm threshold 0.2 instead of 0.4 or 0.6 (ablation)DC-AE-f32c128 + USiT-H · ImageNet 512x512gFID: 12.88 / 13.24 -> 9.42Improved
add FreqWarm to 512-channel autoencoder (ablation)DC-AE-f32c512 · ImageNet 512x512gFID: 54.84 -> 42.66Improved

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