AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

From smooth to sharp: Frequency-Decoupled Latent Optimization for Realistic Image Generation

Tejaswini Medi, Hsien-Yi Wang, Arianna Rampini, Margret Keuper

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Hierarchical VQ-VAE generation

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
decouple low and high frequency bands into separate encoders and decoders (ablation)VA-VAE with wavelet input · ImageNet-1K val reconstructionreconstruction loss: 0.0125 -> 0.0044Improved
decouple low and high frequency bands (ablation)VA-VAE with wavelet input · ImageNet-1K val reconstructionhigh-frequency loss: 0.0098 -> 0.0020Improved
decouple low and high frequency bands (ablation)VA-VAE with wavelet input · ImageNet-1K val reconstructionPSNR: 26.01 -> 30.03Improved
decouple low and high frequency bands (ablation)VA-VAE with wavelet input · ImageNet-1K val reconstructionrFID: 0.6509 -> 0.4156Improved
decouple bands in KL-VAE and MS-VQ-VAE style tokenizers (coupled baselines) (ablation)KL-VAE with wavelet input · ImageNet-1K val reconstructionreconstruction loss: 0.1146 -> 0.0044Improved
use DeBaT tokenizer instead of VA-VAE in LightningDiTLightningDiT · ImageNet-1K 256 w/o CFGgFID: 5.14 -> 3.24Improved
use DeBaT tokenizer instead of VA-VAE in LightningDiTLightningDiT · ImageNet-1K 256 w/ CFGgFID: 2.11 -> 1.52Improved
use DeBaT tokenizer instead of VA-VAE in LightningDiTLightningDiT · ImageNet-1K 256 w/ CFGIS: 252.3 -> 317.4Improved
use DeBaT instead of VA-VAELightningDiT · FFHQ 1024gFID: 6.92 -> 6.86No clear effect
use DeBaT instead of VA-VAELightningDiT · FFHQ 1024PSNR: 33.84 -> 34.27Improved
use DeBaT instead of LiteVAE wavelet tokenizerLiteVAE · ImageNet-1K val reconstructionhigh-frequency loss: 0.0050 -> 0.0020Improved
use dual encoder-decoder (DeBaT) instead of VA-VAEVA-VAE · tokenizer efficiencyparameters (M): 72.60 -> 137.91Hurt

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