Research map / Architectures and efficiency / Attention and transformers
Vicinity Vision Transformer
Weixuan Sun, Zhen Qin, Hui Deng, Jianyuan Wang, Yi Zhang, Kaihao Zhang, Nick M. Barnes, Stan T. Birchfield, Lingpeng Kong, Yiran Zhong
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 |
|---|---|---|---|
| add 2D vicinity locality (ablation) | VVT-Tiny · CIFAR-100 | top-1 accuracy (%): 82.13 -> 82.92 | Improved |
| add 2D vicinity locality (ablation) | VVT-Tiny · ImageNet-1k | top-1 accuracy (%): 78.0 -> 79.4 | Improved |
| use 1D locality (cosFormer) instead of no locality (ablation) | VVT-Tiny · CIFAR-100 | top-1 accuracy (%): 82.13 -> 76.78 | Hurt |
| use 1D locality (cosFormer) in 2D vision data (ablation) | VVT-Tiny · ImageNet-1k | top-1 accuracy (%): failed to converge | Hurt |
| raise feature reduction ratio from 2 to 8 (ablation) | VVT · CIFAR-100 | top-1 accuracy (%): 82.92 -> 80.60 | Hurt |
| use feature reduction ratio 1 instead of 2 (ablation) | VVT · CIFAR-100 | top-1 accuracy (%): 82.89 vs 82.92 | No clear effect |
| raise feature reduction ratio from 2 to 4 (ablation) | VVT · CIFAR-100 | top-1 accuracy (%): 82.92 -> 82.00 | Hurt |
| add feature preserving connection (FPC) (ablation) | VVT-Tiny · ImageNet-1k | top-1 accuracy (%): 77.9 -> 79.4 | Improved |
| add FRA and FPC to Performer attention (ablation) | VVT-T with Performer · ImageNet-1k | top-1 accuracy (%): 77.2 -> 77.3 | No clear effect |
| use vanilla softmax attention instead of vicinity attention (ablation) | VVT-T · ImageNet-1k | top-1 accuracy (%): 79.4 -> 80.0 | Improved |
| use vicinity attention instead of vanilla attention (ablation) | VVT-T · CIFAR-100 | top-1 accuracy (%): 82.68 -> 82.92 | Improved |
| use vicinity attention instead of Performer (ablation) | VVT-T · ImageNet-1k | top-1 accuracy (%): 77.2 -> 79.4 | Improved |
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.
- SOFT (13M) — SOFT: Softmax-free Transformer with Linear Complexity (NeurIPS 2021) · better on 1 of 1 shared result
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
- RegionViT: Regional-to-Local Attention for Vision Transformers (ICLR 2022 · similarity 0.91)
- Locality-Attending Vision Transformer (ICLR 2026 · similarity 0.91)
- Slide-Transformer: Hierarchical Vision Transformer With Local Self-Attention (CVPR 2023 · similarity 0.90)
- You Only Need Less Attention at Each Stage in Vision Transformers (CVPR 2024 · similarity 0.89)
- Bridging the Divide: Reconsidering Softmax and Linear Attention (NeurIPS 2024 · similarity 0.88)
- Skip-Attention: Improving Vision Transformers by Paying Less Attention (ICLR 2024 · similarity 0.87)
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.
