AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Efficient Visual Transformer by Learnable Token Merging

Yancheng Wang, Yingzhen Yang

TPAMI 2025 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 15 citations (OpenAlex)

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 changedWhereMeasuredEffect
replace transformer blocks with LTM blocks (learnable token merging)MobileViT-S · ImageNet-1k val (from scratch)top-1 accuracy: 78.4 -> 79.7Improved
replace transformer blocks with LTM blocksSwin-T · ImageNet-1k val (from scratch)top-1 accuracy: 81.3 -> 81.8Improved
replace transformer blocks with LTM blocksViT-B · ImageNet-1k val (from scratch)top-1 accuracy: 83.7 -> 83.9Improved
LTM instead of ToMe/LTMP token mergingMobileViT-S · ImageNet-1k val (from scratch)top-1 accuracy: 76.7 -> 79.7Improved
LTM instead of LTMP token merging, fine-tuned 50 epochsSwin-B · ImageNet-1k val (fine-tuned 50 epochs)top-1 accuracy: 82.55 -> 83.64Improved
LTM instead of LTMP token merging, fine-tuned 50 epochsSwin-T · ImageNet-1k val (fine-tuned 50 epochs)top-1 accuracy: 80.30 -> 81.38Improved
LTM backbone for SSDLite detectionMobileViT-S · COCO detection (SSDLite)mAP: 27.7 -> 28.4Improved
LTM backbone for Mask R-CNNEfficientViT-B1 · COCO instance segmentation (Mask R-CNN)mAP box: 33.5 -> 34.3Improved
add token merging with learnable mask (LTM) to reduce IB loss (ablation)MobileViT-S · ImageNet-1k val (from scratch, IB loss)IB loss: -0.00432 -> -0.01725Improved
add token merging with learnable mask (LTM) vs ToMe (ablation)ViT-B · ImageNet-1k val (from scratch, IB loss)IB bound: 0.04583 -> 0.03094Improved
lower compression ratio from 0.95 to 0.65 (ablation)LTM-ViT-B · ImageNet-1k val (from scratch, compression ratio ablation)top-1 accuracy: 84.43 -> 83.75Hurt
lower compression ratio from 0.65 to 0.40 (ablation)LTM-ViT-B · ImageNet-1k val (from scratch, compression ratio ablation)top-1 accuracy: 83.75 -> 82.23Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy