AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

FastVAR: Linear Visual Autoregressive Modeling via Cached Token Pruning

Hang Guo, Yawei Li, Taolin Zhang, Jiangshan Wang, Tao Dai, Shu-Tao Xia, Luca Benini

ICCV 2025 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 32 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 FastVAR cached token pruningHART · GenEval 1024x1024speedup / Overall: 1.0x -> 1.5x, 0.51 -> 0.51Improved
add FastVAR cached token pruningInfinity · GenEval 1024x1024speedup / Overall: 2.7x speedup, 0.73 -> 0.72Improved
add FastVAR cached token pruningHART · MJHQ30K peopleFID: 30.61 -> 28.19Improved
add FastVAR cached token pruningInfinity · MJHQ30K foodFID: 31.55 -> 32.54Hurt
replace ToMe token merging with FastVAR pruning (ablation)HART · MJHQ30K FIDFID: 35.22 -> 28.19Improved
prune tokens at small scales [16,21] (75%) instead of large scales [48,64] (ablation)HART · MJHQ30KFID: 28.56 -> 34.27Hurt
prune last scales [48,64] at 75% instead of [27,36] (ablation)HART · Speedupspeedup: 1.14x -> 1.64xImproved
use cache from step K-N instead of K-N-3 (ablation)HART · MJHQ30KFID: 40.61 -> 28.19Improved
raise last-scale pruning ratio from 75% to 100% on HART (ablation)HART · MJHQ30KFID: 28.19 -> 48.54Hurt
combine FastVAR with FlashAttention instead of FastVAR alone (ablation)Infinity · 1024x1024 efficiencyspeedup: 2.1x -> 2.7xImproved
use FastVAR on a smaller class-conditional VAR (d=24) on ImageNetVAR d=24 · ImageNet 256x256FID: 2.29 -> 2.64Hurt
increase pruning ratio (ablation)HART · MJHQ30KFIDMixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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