AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

ViT-Linearizer: Distilling Quadratic Knowledge into Linear-Time Vision Models

Guoyizhe Wei, Rama Chellappa

ICCV 2025 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 6 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
add activation matching to masked prediction (ablation)ViT-Linearizer · ImageNet-1kaccuracy: 83.6 -> 84.3Improved
add activation matching to masked prediction (ablation)ViT-Linearizer · ADE20kmIoU: 49.7 -> 51.3Improved
add masked prediction to activation matching (ablation)ViT-Linearizer · ImageNet-1kaccuracy: 83.8 -> 84.3Improved
add masked prediction to activation matching (ablation)ViT-Linearizer · ADE20kmIoU: 50.1 -> 51.3Improved
use token-wise masking instead of block-wise masking (ablation)ViT-Linearizer · ImageNet-1kaccuracy: 84.0 -> 84.3Improved
match visible tokens only instead of class token only (ablation)ViT-Linearizer · ImageNet-1kaccuracy: 83.7 -> 84.3Improved
match visible tokens only instead of all tokens (ablation)ViT-Linearizer · ADE20kmIoU: 49.0 -> 51.3Improved
distill CLIP into Adventurer-Base vs supervised trainingAdventurer-Base · ImageNet-1k 224accuracy: 82.6 -> 84.3Improved
distill CLIP into Adventurer-Base vs supervised trainingAdventurer-Base · ADE20kmIoU: 47.8 -> 51.3Improved
distill CLIP into Adventurer-Base vs supervised trainingAdventurer-Base · CityscapesmIoU: 80.2 -> 82.0Improved
scale student from Base to Large with CLIP ViT-B/16 teacherAdventurer student · ImageNet-1kaccuracy: 84.3 -> 85.0Improved
use larger CLIP ViT-L/14 teacher instead of ViT-B/16 for Large studentAdventurer-L · ImageNet-1kaccuracy: 85.0 -> 85.2Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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

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