AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Latent Fourier Transform

Mason Long Wang, Cheng-Zhi Anna Huang

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Latent diffusion with learned tokenizers · 2 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
use latent-spectrum conditioning (LatentFT-MLP) vs best baseline ILVRdiffusion autoencoder · MTG-Jamendo conditional generationFAD: 1.061 -> 0.337Improved
use LatentFT-MLP vs ILVRdiffusion autoencoder · MTG-Jamendo conditional generationloudness correlation: 0.575 -> 0.815Improved
use LatentFT vs ILVR for blendingdiffusion autoencoder · MTG-Jamendo blendingtimbre MCD/100: 0.825 -> 1.021Hurt
use LatentFT-UNet vs ILVR for blendingdiffusion autoencoder · MTG-Jamendo blendingFAD: 2.696 -> 1.357Improved
add random frequency masking during training (ablation)LatentFT-MLP · MTG-Jamendo conditional generationFAD: 5.341 -> 0.349Improved
add correlated bin scores in mask sampling (ablation)LatentFT-MLP · MTG-Jamendo conditional generationFAD: 2.744 -> 0.349Improved
use logarithmic frequency axis for correlation (ablation)LatentFT-MLP · MTG-Jamendo conditional generationloudness correlation: 0.512 -> 0.800Improved
add encoder (ablation)LatentFT-MLP · MTG-Jamendo conditional generationloudness correlation: 0.028 -> 0.800Improved
add bandpass augmentation (ablation)LatentFT-MLP · MTG-Jamendo conditional generationFAD: 0.349 -> 1.511Hurt
train without frequency masking (ablation)LatentFT-MLP · MTG-Jamendo blendingFAD: 4.789 -> 1.371Improved
post-hoc latent frequency masking of pretrained DAC/RAVEDAC, RAVE · MTG-Jamendo conditional generationFAD: 7.016 (DAC), 4.695 (RAVE) vs 0.337Hurt

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