AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

AVQ-Attention: Adaptive Vector-Quantized Attention

Winfried van den Dool, Patrick Forré, Amirhossein Habibian, Yuki Asano, Max Welling

ECCV 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse attention for long-context LLMs · 1 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
replace flat VQ codebook with adaptive parent-child refinementVQ-attention · ADE20K (DPT-L, 30-epoch comparison)mIoU: 42.70 -> 43.33 (vs flat M=128)Improved
fuse VQ precompute into a single kernel (ablation)VQ-attention · Synthetic N=65k kernel benchmarkkernel time (ms): 20.7 -> 10.6Improved
replace Flash Attention with fused AVQ-attentionattention kernel · Synthetic N=65k kernel benchmarkkernel time (ms): 847 -> 6.67Improved
replace Flash Attention with AVQ-attention at short sequenceattention kernel · Synthetic N=1k kernel benchmarkkernel time (ms): 0.21 -> 0.13 (AVQ 64/8/8); 0.21 -> 0.21 (128/16/8)Mixed
replace attention with AVQ in SD1.5 UNet (level-0 blocks)SD1.5 · SD1.5 COCO 1024^2 (ScaleCrafter)FID: 41.03 -> 35.39Improved
replace attention with AVQ in SD1.5 UNet (level-0 blocks)SD1.5 · SD1.5 COCO 512^2FID: 12.75 -> 12.50No clear effect
also replace level-1 (N=1k) attention blocks with AVQ (ablation)AVQ M64P16C8 · SD1.5 COCO 1024^2 (ScaleCrafter)UNet time per step (ms): 135.06 -> 128.59 ms, FID 35.65 -> 36.29Mixed
reduce refined parents P at inference from 32 to 16 (ablation)AVQ M0=64 C=8 · ImageNet-1k (ViT-B elastic inference, M0=64 C=8)top-1 accuracy: 82.29 -> 82.04No clear effect
reduce refined parents P at inference from 8 to 4 (ablation)AVQ M0=64 C=8 · ImageNet-1k (ViT-B elastic inference, M0=64 C=8)top-1 accuracy: 81.16 -> 79.23Hurt
use tiled approximate denominator for importance instead of exact (ablation)AVQ importance scores · ImageNet (ViT-Base), P=32attention mass captured: 82.0% -> 77.7%Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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