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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace standard attention with QUEST | ViT-Tiny DeiT · ImageNet-1K val | top-1 accuracy (%): 72.6 -> 73.4 | Improved |
| use QUEST instead of QNorm (ablation) | ViT-Tiny · ImageNet-1K val | top-1 accuracy (%): 72.7 -> 73.4 | Improved |
| use QUEST instead of QKNorm-DS | ViT-L/16 100 epochs · ImageNet-1K val | top-1 accuracy (%): 72.5 -> 74.9 | Improved |
| replace standard attention with QUEST | ViT-L/16 DeiT-3 · ImageNet-1K val | top-1 accuracy (%): 83.9 -> 84.1 | Improved |
| replace standard attention with QUEST | ViT-B/16 DeiT-3 · ImageNet-A | top-1 accuracy (%): 34.8 -> 37.6 | Improved |
| replace standard attention with QUEST | ViT-Ti · PGD adversarial attack | top-1 accuracy (%): 43.65 -> 45.26 | Improved |
| add QUEST to Elliptical attention | Elliptical attention ViT-Ti · clean ImageNet | top-1 accuracy (%): 71.53 -> 72.48 | Improved |
| replace standard attention with QUEST | ViT-B DeiT-3 · ImageNet AG-CAM localization | mIoU: 35.53 -> 53.54 | Improved |
| replace standard attention with QUEST | GraphGPS · GNN benchmarks (ZINC, CIFAR10, PATTERN, CLUSTER) | accuracy / MAE: ZINC 0.070 -> 0.069 | No clear effect |
| replace standard attention with QUEST | GraphGPS · LRGB Peptides-struct / PascalVOC-SP | MAE / F1: 0.250 -> 0.251 | No clear effect |
| replace standard attention with QUEST | GraphGPS · LRGB COCO-SP | F1: 0.341 -> 0.349 | Improved |
| replace standard attention with QUEST | Transformer Small 44M · WikiText-103 test | perplexity: 34.076 -> 33.966 | No clear effect |
Closest papers by meaning
- Theory, Analysis, and Best Practices for Sigmoid Self-Attention (ICLR 2025 · similarity 0.87)
- Bayes optimal learning of attention-indexed models (NeurIPS 2025 · similarity 0.87)
- Transformers Learn Faster with Semantic Focus (NeurIPS 2025 · similarity 0.86)
- Bridging the Divide: Reconsidering Softmax and Linear Attention (NeurIPS 2024 · similarity 0.86)
- SAS: Simulated Attention Score (NeurIPS 2025 · similarity 0.86)
- Mixture-of-Top-$k$ Attention: Efficient Attention as Scalable Fast Weights (NeurIPS 2026 · similarity 0.86)
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.
