Research map / Architectures and efficiency / Attention and transformers
Semantic Equitable Clustering: A Simple and Effective Strategy for Clustering Vision Tokens
Qihang Fan, Huaibo Huang, Mingrui Chen, Ran He
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 window attention with SEC in Swin-T | Swin-T · ImageNet-1K | top-1 accuracy (%): 81.3 -> 83.8 | Improved |
| replace window attention with SEC in Swin-S | Swin-S · ImageNet-1K | top-1 accuracy (%): 83.0 -> 85.0 | Improved |
| add SEC to DeiT-S | DeiT-S · COCO Mask R-CNN | box AP: 44.5 -> 47.7 | Improved |
| add SEC to DeiT-S | DeiT-S · ADE20K | mIoU: 43.0 -> 47.5 | Improved |
| add SEC to DeiT-B (fewer FLOPs) | DeiT-B · ImageNet-1K | top-1 accuracy (%): 81.8 -> 82.4 | Improved |
| use SEC instead of EViT token pruning in DeiT-S | DeiT-S · ImageNet-1K | top-1 accuracy (%): 79.8 -> 80.5 | Improved |
| add SEC as pretrained backbone under SimMIM | Swin-B · ImageNet-1K | top-1 accuracy (%): 84.0 -> 85.9 | Improved |
| use SEC connector at 288+1 tokens instead of full 576+1 tokens | LLaVA-1.5 · TextVQA | accuracy: 58.2 -> 60.1 | Improved |
| use SEC instead of EViT connector at 144+1 tokens | LLaVA-1.5 · TextVQA | accuracy: 51.2 -> 56.8 | Improved |
| reduce SEC output tokens from 288+1 to 144+1 (ablation) | LLaVA-1.5 · TextVQA | accuracy: 60.1 -> 56.8 | Hurt |
| use interleaved instead of sequential token merging in SEC connector (144+1) (ablation) | LLaVA-1.5 · TextVQA | accuracy: 47.2 -> 56.8 | Improved |
| increase tokens per cluster from 98 to 196 (ablation) | SECViT-T · ImageNet-1K | top-1 accuracy (%): 82.7 -> 83.0 | No clear effect |
Closest papers by meaning
- Self-Enhanced Image Clustering with Cross-Modal Semantic Consistency (AAAI 2026 · similarity 0.91)
- Hierarchical Semantic Alignment for Image Clustering (AAAI 2026 · similarity 0.90)
- Language-Assisted Image Clustering Guided by Discriminative Relational Signals and Adaptive Semantic Centers (NeurIPS 2026 · similarity 0.90)
- Semantic-Enhanced Image Clustering (AAAI 2023 · similarity 0.90)
- Image Clustering Conditioned on Text Criteria (ICLR 2024 · similarity 0.87)
- Image Clustering with External Guidance (ICML 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.
