AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

CaTok: Taming Mean Flows for One-Dimensional Causal Image Tokenization

Yitong Chen, Zuxuan Wu, Xipeng Qiu, Yu-Gang Jiang

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 0 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 MeanFlow loss to rectified-flow decoder (one-step rFID) (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 ablation (80 epochs)rFID@1: 183.69 -> 4.71Improved
add MeanFlow loss (generation) (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 ablation (80 epochs)gFID: 19.67 -> 24.39Hurt
add REPA objective (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 ablation (80 epochs)gFID: 24.39 -> 17.92Improved
add REPA-A objective on encoder features (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 ablation (80 epochs)gFID: 17.92 -> 13.54Improved
add token selection in interval [r,t] (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 ablation (80 epochs)gFID: 13.54 -> 4.91Improved
add token selection in [r,t] (one-step reconstruction) (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 ablation (80 epochs)rFID@1: 3.92 -> 4.89Hurt
select tokens in [r,t] instead of all tokens (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 token selectiongFID: 13.54 -> 4.91Improved
select tokens in [r,t] instead of first k tokens (ablation)CaTok-B-256 · ImageNet-1K 256, CaTok-B-256 token selectiongFID: 9.21 -> 4.91Improved
add MeanFlow-style causality (with token selection) in non-autoregressive generator (ablation)CaTok-L-32 · ImageNet-1K 256, eps-MaskGiT 8 stepsgFID: 3.26 -> 2.69Improved
use CaTok tokenizer instead of LlamaGen tokenizer (VQ)LlamaGen generator · ImageNet-1K 256 (CaTok-VQ with FSQ)gFID: 3.80 -> 3.35Improved
use CaTok-L-256 instead of Semanticist-L-256 without CFGAR generator · ImageNet-1K 256 generation, no CFGgFID: 7.60 -> 5.52Improved
increase number of tokens from 32 to 128CaTok-L · ImageNet-1K 256 generationgFID: 3.40 -> 2.95Improved

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