AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Stratified Knowledge-Density Super-Network for Scalable Vision Transformers

Longhua Li, Lei Qi, Xin Geng

AAAI 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 2 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
Use WPAC instead of Taylor FO criterion for attention pruningDeiT-S · ImageNet-1k, 2:4 size, no fine-tuningtop-1 accuracy: 39.4 -> 72.9Improved
Use WPAC instead of Taylor FO criterion for attention pruningDeiT-B · ImageNet-1k, 1:4 size, no fine-tuningtop-1 accuracy: 6.0 -> 41.8Improved
Add importance weighting to PCA over all tokens (versus uniform weight 0.01) (ablation)WPAC DeiT-B · ImageNet-1k, 50% dimstop-1 accuracy: 76.1 -> 76.9Improved
Use all tokens with importance weighting instead of class token only (ablation)WPAC DeiT-B · ImageNet-1k, 50% dimstop-1 accuracy: 65.1 -> 76.9Improved
Increase randomly sampled tokens from 1 to 2 (ablation)WPAC DeiT-S · ImageNet-1k, 50% dimstop-1 accuracy: 64.5 -> 71.0Improved
Use 5 random tokens instead of 2 on DeiT-B (ablation)WPAC DeiT-B · ImageNet-1k, 50% dimstop-1 accuracy: 75.7 -> Ill-Cond.Hurt
Use unweighted PCA over all tokens (weight 1.0) (ablation)WPAC · ImageNet-1k, 50% dimstop-1 accuracy: 76.9 -> Ill-Cond.Hurt
Train with PIAD instead of layerdrop (ablation)SKD DeiT-S · ImageNet-1k, 4:12 sub-networktop-1 accuracy: 39.7 -> 70.6Improved
Train with PIAD instead of weighted channel dropout (ablation)SKD DeiT-S · ImageNet-1k, 4:12 sub-networktop-1 accuracy: 34.2 -> 70.6Improved
Train with PIAD instead of channel dropout (ablation)SKD DeiT-S · ImageNet-1k, 10:12 sub-networktop-1 accuracy: 46.5 -> 79.0Improved
Train with PIAD instead of plain WPAC baseline (ablation)SKD DeiT-Ti · ImageNet-1k, 4:12 sub-networktop-1 accuracy: 1.4 -> 61.4Improved
Use SKD instead of WAVE for scaled models without fine-tuningDeiT-S · ImageNet-1k, 8:12 sizetop-1 accuracy: 74.1 -> 78.2Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 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