AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

SparVAR: Exploring Sparsity in Visual AutoRegressive Modeling for Training-Free Acceleration

Zekun Li, Ning Wang, Tongxin Bai, Changwang Mei, Peisong Wang, Shuang Qiu, Jian Cheng

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: Next-scale and token autoregressive images · 6 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 early-scale attention sink to CSLA sparse mask (ablation)CSLA on Infinity-8B · GenEval/PSNRPSNR: 23.541 -> 25.897Improved
add cached dense output O_cache to CS4A (ablation)CS4A on Infinity-8B · Table 5PSNR: 25.665 -> 26.359Improved
add CSLA block-wise local sparse attention on top of CS4A (ablation)SparVAR on Infinity-8B · Table 5PSNR: 26.359 -> 29.481; speedup 1.43x -> 1.57xImproved
use block size 128 instead of 64 in CSLA kernel (ablation)CSLA FlexAttention kernel · last-scale attention forwardspeedup vs FlashAttention2: 3.60x -> 5.61xImproved
use smaller local windows ([3,3,3] vs [3,5,7]) in CSLA (ablation)CSLA · Infinity-8Blatency/PSNR: latency 0.5346 -> 0.4480 ms; PSNR 25.897 -> 25.837Mixed
use a larger sink range (scale <=8 vs <=5) (ablation)CSLA · Infinity-8BPSNR / latency: PSNR 25.897 -> 27.582; latency 0.5346 -> 0.7729 msMixed
choose a mid sparse decision scale (10) rather than early scales (6-7) (ablation)CS4A · Infinity-8Bgeneration qualityImproved
replace scale skipping (FastVAR) with SparVAR sparse attention without skippingInfinity-8B · Infinity-8B GenEval 1024PSNR vs baseline: 17.403 -> 29.481Improved
combine SparVAR with skip-last-2-scalesInfinity-8B · GenEvalspeedup: 1.57x -> 2.28xImproved
SparVAR vs unaccelerated baseline on semantic qualityInfinity-8B · GenEval overallGenEval: 0.798 -> 0.796No clear effect

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