Research map / Architectures and efficiency / Attention and transformers
Strip-MLP: Efficient Token Interaction for Vision MLP
Guiping Cao, Shengda Luo, Wenjian Huang, Xiangyuan Lan, Dongmei Jiang, Yaowei Wang, Jianguo Zhang
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 |
|---|---|---|---|
| split channels into groups with unshared strip MLP (CGSMM patch splitting, P>1) (ablation) | Strip-MLP-S · Caltech-101 | Top-1 accuracy: 90.93 -> 92.09 | Improved |
| split channels into groups with unshared strip MLP (CGSMM patch splitting, P>1) (ablation) | Strip-MLP-S · CIFAR-100 | Top-1 accuracy: 85.03 -> 86.50 | Improved |
| use cascade rather than parallel row/column group strip MLP (ablation) | Strip-MLP-S · CIFAR-100 | Top-1 accuracy: 85.89 -> 86.50 | Improved |
| add LSMM local branch to CGSMM (ablation) | Strip-MLP-S · Caltech-101 | Top-1 accuracy: 91.10 -> 91.38 | Improved |
| add CGSMM branch to LSMM (ablation) | Strip-MLP-S · CIFAR-100 | Top-1 accuracy: 86.13 -> 86.50 | Improved |
| increase strip width from 1 to 3 (ablation) | Strip-MLP-S · Caltech-101 | Top-1 accuracy: 91.38 -> 92.09 | Improved |
| increase strip width beyond 3 (5, 7) (ablation) | Strip-MLP-S · CIFAR-100 | Top-1 accuracy: 86.50 -> 86.39 (5), 86.35 (7) | No clear effect |
| apply Strip MLP layer to Wave-MLP-T (ablation) | Wave-MLP-T · ImageNet-1K | Top-1 accuracy: 80.6 -> 81.2 | Improved |
| Strip-MLP-T vs Wave-MLP-S on ImageNet-1K | Strip-MLP-T · ImageNet-1K | Top-1 accuracy: 82.6 (Wave-MLP-S) vs 82.2 | Hurt |
| Strip-MLP vs transformers on small datasets | Strip-MLP-B · Caltech-101 | Top-1 accuracy: 78.42 (Swin-B) -> 92.26 | Improved |
Closest papers by meaning
- AS-MLP: An Axial Shifted MLP Architecture for Vision (ICLR 2022 · similarity 0.85)
- VisionZip: Longer is Better but Not Necessary in Vision Language Models (CVPR 2025 · similarity 0.84)
- SA²VP: Spatially Aligned-and-Adapted Visual Prompt (AAAI 2024 · similarity 0.83)
- Patch-as-Decodable-Token: Towards Unified Multi-Modal Vision Tasks in MLLMs (ICLR 2026 · similarity 0.83)
- CrossGET: Cross-Guided Ensemble of Tokens for Accelerating Vision-Language Transformers (ICML 2024 · similarity 0.83)
- DeepStack: Deeply Stacking Visual Tokens is Surprisingly Simple and Effective for LMMs (NeurIPS 2024 · similarity 0.83)
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.
