AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Adaptive Visual Autoregressive Acceleration via Dual-Linkage Entropy Analysis

Yu Zhang, Jingyi Liu, Feng Liu, Duoqian Miao, Qi Zhang, Kexue Fu, Changwei Wang, Longbing Cao

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer neural operators for PDEs · 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
use entropy instead of attention to select tokens (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.69 -> 0.72Improved
use entropy instead of MSE to select tokens (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.71 -> 0.72Improved
adaptive activation scale instead of fixed 6th start scale (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.62 -> 0.72Improved
adaptive activation scale instead of fixed 10th start scale (ablation)NOVA on Infinity-2B · GenEval / speedupspeedup: 1.63x -> 2.89xImproved
add scale-linkage to layer-linkage-only variant (ablation)NOVA on Infinity-2B · GenEval / speedupspeedup: 2.34x -> 2.85xImproved
add layer-linkage on top of scale-linkage (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.71 -> 0.72Improved
add layer-linkage to no-linkage variant (dagger) (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.69 -> 0.70Improved
dynamic residual cache reuse instead of static caching from 6-th scale (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.67 -> 0.72Improved
dynamic residual cache reuse instead of static caching from 8-th scale (ablation)NOVA on Infinity-2B · GenEvaloverall: 0.71 -> 0.72Improved
apply NOVA to Infinity-2B instead of FastVARInfinity-2B · GenEval / speedupspeedup: 2.45x -> 2.89xImproved
apply NOVA to Infinity-8BInfinity-8B · ImageRewardscore: 1.030 -> 1.035No clear effect
apply NOVA to VAR-d36 on ImageNet 512VAR-d36 · ImageNet class-conditionallatency: 2.89s -> 1.97s; FID 2.63 -> 2.79Mixed

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