AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Latent Denoising Makes Good Tokenizers

Jiawei Yang, Tianhong Li, Lijie Fan, Yonglong Tian, Yue Wang

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Hierarchical VQ-VAE generation · 34 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
train tokenizer with latent denoising (l-DeTok) instead of MAR-VAEMAR-B · ImageNet 256FID: 3.71 -> 2.43Improved
train tokenizer with latent denoising (l-DeTok) instead of MAR-VAERandomAR-B · ImageNet 256FID: 11.78 -> 5.22Improved
train tokenizer with latent denoising (l-DeTok) instead of MAR-VAERasterAR-B · ImageNet 256FID: 7.99 -> 4.46Improved
add semantics distillation (VA-VAE, MAETok) to tokenizersMAR-B / RandomAR-B · ImageNet 256FID: VA-VAE 16.66 vs MAR-VAE 3.71 (MAR)Hurt
use interpolative instead of additive latent noise (ablation)tokenizer for SiT/MAR · ImageNet 256FIDImproved
add latent noising with tau~U(0,1) (ablation)tokenizer MAR-B · ImageNet 256FID: 3.31 -> 2.77Improved
add latent noising with tau~U(0,1) (ablation)tokenizer SiT-B · ImageNet 256FID: 6.97 -> 5.56Improved
shift noise level to higher values, logit(tau)~N(0.8,1) vs uniform (ablation)tokenizer MAR-B · ImageNet 256FID: 2.77 -> 2.58Improved
use random instead of constant masking ratio (ablation)tokenizer · ImageNet 256FIDImproved
add joint denoising (noise+masking) over baseline tokenizer (ablation)tokenizer MAR-B · ImageNet 256FID: 3.31 -> 2.65Improved
add masking on top of latent noising (ablation)tokenizer SiT-B · ImageNet 256FIDNo clear effect
scale generative model with l-DeTok vs baseline tokenizerSiT-XL · ImageNet 256FID: 4.47 -> 3.14Improved

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