AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

WeTok: Powerful Discrete Tokenization for High-Fidelity Visual Reconstruction

Shaobin Zhuang, Yiwei Guo, Fangyikang Wang, Canmiao Fu, Zhipeng Huang, Zeyue Tian, Xiaohui Li, Ying Zhang, Chen Li, Yali Wang

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Generative latent space tooling

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 generative decoder (stage 2) after reconstruction training (ablation)WeTok · ImageNet valrFID: 5.37 -> 3.90Improved
add generative decoder (stage 2) (ablation)WeTok · ImageNet valPSNR: 20.53 -> 20.72Improved
use group-wise lookup-free quantization (GQ) instead of LFQ at same compression (ablation)WeTok · ImageNetreconstructionImproved
use GQ instead of LFQ for large embedding dimension (ablation)WeTok · memory at d=24GPU memory: OOM -> 10.6 GBImproved
increase number of groups g in GQ (ablation)WeTok · ImageNetreconstructionImproved
train on 400M general-domain data instead of ImageNet (ablation)WeTok · ImageNet valrFID, LPIPS vs validation loss, SSIM, PSNRMixed
use constant learning rate instead of warmup + cosine decay (ablation)WeTok · ImageNetreconstructionImproved
scale base channels/blocks: 256 channels, 4 residual blocks (ablation)WeTok · ImageNetreconstructionImproved
use WeTok tokenizer vs Open-MAGVIT2 in AR model with growing parameters (ablation)WeTok-AR · ImageNet generationFIDMixed
use WeTok tokenizer in AR generation vs Open-MAGVIT2-AR-XLWeTok-AR-XL · ImageNet 256FID: 2.33 -> 2.31No clear effect

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