AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

QUEST: A robust attention formulation using query-modulated spherical attention

Hariprasath Govindarajan, Per Sidén, Jacob Roll, Fredrik Lindsten

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 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 standard attention with QUESTViT-Tiny DeiT · ImageNet-1K valtop-1 accuracy (%): 72.6 -> 73.4Improved
use QUEST instead of QNorm (ablation)ViT-Tiny · ImageNet-1K valtop-1 accuracy (%): 72.7 -> 73.4Improved
use QUEST instead of QKNorm-DSViT-L/16 100 epochs · ImageNet-1K valtop-1 accuracy (%): 72.5 -> 74.9Improved
replace standard attention with QUESTViT-L/16 DeiT-3 · ImageNet-1K valtop-1 accuracy (%): 83.9 -> 84.1Improved
replace standard attention with QUESTViT-B/16 DeiT-3 · ImageNet-Atop-1 accuracy (%): 34.8 -> 37.6Improved
replace standard attention with QUESTViT-Ti · PGD adversarial attacktop-1 accuracy (%): 43.65 -> 45.26Improved
add QUEST to Elliptical attentionElliptical attention ViT-Ti · clean ImageNettop-1 accuracy (%): 71.53 -> 72.48Improved
replace standard attention with QUESTViT-B DeiT-3 · ImageNet AG-CAM localizationmIoU: 35.53 -> 53.54Improved
replace standard attention with QUESTGraphGPS · GNN benchmarks (ZINC, CIFAR10, PATTERN, CLUSTER)accuracy / MAE: ZINC 0.070 -> 0.069No clear effect
replace standard attention with QUESTGraphGPS · LRGB Peptides-struct / PascalVOC-SPMAE / F1: 0.250 -> 0.251No clear effect
replace standard attention with QUESTGraphGPS · LRGB COCO-SPF1: 0.341 -> 0.349Improved
replace standard attention with QUESTTransformer Small 44M · WikiText-103 testperplexity: 34.076 -> 33.966No clear effect

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