Research map / Architectures and efficiency / Attention and transformers
Transformer in Transformer
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 inner transformer on visual words (TNT) | DeiT-S · ImageNet | top-1 accuracy: 79.8 -> 81.5 | Helped |
| add word-level position encoding (ablation) | TNT-S · ImageNet | top-1 accuracy: 80.8 -> 81.5 | Helped |
| add sentence-level position encoding (ablation) | TNT-S · ImageNet | top-1 accuracy: 80.7 -> 81.5 | Helped |
| vary inner heads (1 to 8) (ablation) | TNT-S · ImageNet | top-1 accuracy: 81.0/81.4/81.5/81.3/81.1 | Mixed |
| change visual words per patch m (ablation) | TNT-S · ImageNet | top-1 accuracy: 81.0 / 81.5 / 81.1 | No clear effect |
| replace TNT blocks with vanilla blocks in later layers (ablation) | TNT-S · ImageNet | top-1 accuracy: 81.5 -> 80.8 | Hurt |
| add SE module (ablation) | TNT-S · ImageNet | top-1 accuracy: 81.5 -> 81.7 | Helped |
| TNT-S backbone instead of PVT-Small | DETR · COCO val2017 | AP: 34.7 -> 38.2 | Helped |
Closest papers by meaning
- Viewing Transformers Through the Lens of Long Convolutions Layers (ICML 2024 · similarity 0.86)
- CAT: Circular-Convolutional Attention for Sub-Quadratic Transformers (NeurIPS 2025 · similarity 0.85)
- SLAB: Efficient Transformers with Simplified Linear Attention and Progressive Re-parameterized Batch Normalization (ICML 2024 · similarity 0.85)
- Rethinking Attention: Exploring Shallow Feed-Forward Neural Networks as an Alternative to Attention Layers in Transformers (Student Abstract) (AAAI 2024 · similarity 0.85)
- AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers (ICML 2024 · similarity 0.85)
- Wavy Transformer (NeurIPS 2025 · similarity 0.85)
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.