Research map / Architectures and efficiency / Attention and transformers
LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference
Benjamin Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze
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 multi-resolution pyramid shape (vs straight stack) (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 56.5 -> 71.9 | Improved |
| add PatchConv (four 3x3 convolutions in patch embedding) (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 65.3 -> 71.9 | Improved |
| add BatchNorm (instead of LayerNorm) with zero-init residuals (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 66.6 -> 71.9 | Improved |
| add hard distillation from RegNetY-16GF teacher (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 69.7 -> 71.9 | Improved |
| add attention bias (replacing positional embedding) (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 70.4 -> 71.9 | Improved |
| add wider blocks (smaller Q,K, V of 2D, MLP expansion 2) (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 70.9 -> 71.9 | Improved |
| add Hardswish activation after attention product (ablation) | LeViT-128S · ImageNet-1k val (LeViT-128S, 100 epochs) | top-1 accuracy (%): 71.1 -> 71.9 | Improved |
| graft 2 ResNet-50 stages below 6 DeiT-S layers | DeiT-S · ImageNet-1k val (300 epochs) | top-1 accuracy (%): 79.9 -> 80.9 | Improved |
| graft 3 ResNet-50 stages below 3 DeiT-S layers | DeiT-S · ImageNet-1k val (300 epochs) | top-1 accuracy (%): 79.9 -> 80.1 | Mixed |
| use LeViT-256 instead of EfficientNet B3 | LeViT-256 · ImageNet-V2 matched frequency | top-1 accuracy (%): 70.6 -> 70.0 | Hurt |
| use LeViT-384 instead of DeiT-Small | LeViT-384 · ImageNet-Real | top-1 accuracy (%): 87.8 -> 87.6 | Hurt |
| use LeViT-384 instead of DeiT-Small | LeViT-384 · ImageNet-V2 matched frequency | top-1 accuracy (%): 71.7 -> 71.3 | Hurt |
Later papers that compared with this method
5 later accepted papers report LeViT-256 as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- GmNet: Revisiting Gating Mechanisms From A Frequency View (ICLR 2026 · better on 1 of 2 shared results)
- MIA-Former: Efficient and Robust Vision Transformers via Multi-Grained Input-Adaptation (AAAI 2022 · better on 3 of 4 shared results)
- EfficientFormer: Vision Transformers at MobileNet Speed (NeurIPS 2022 · better on 1 of 1 shared result)
- Self-Slimmed Vision Transformer (ECCV 2022 · better on 1 of 1 shared result)
- TinyViT: Fast Pretraining Distillation for Small Vision Transformers (ECCV 2022 · 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
- A Survey on Efficient Vision Transformers: Algorithms, Techniques, and Performance Benchmarking (TPAMI 2024 · similarity 0.88)
- ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biases (ICML 2021 · similarity 0.88)
- FasterViT: Fast Vision Transformers with Hierarchical Attention (ICLR 2024 · similarity 0.87)
- Slicing Vision Transformer for Flexible Inference (NeurIPS 2024 · similarity 0.87)
- Structured Initialization for Vision Transformers (NeurIPS 2025 · similarity 0.87)
- 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.
