AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Self-Slimmed Vision Transformer

Zhuofan Zong, Kunchang Li, Guanglu Song, Yali Wang, Yu Qiao, Biao Leng, Yu Liu

ECCV 2022 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 41 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 token hard dropping (DynamicViT) with soft slimming TSM at FLOPs ratio 0.25SiT on DeiT-P · ImageNetTop-1 accuracy: 65.6 -> 75.9Improved
replace DynamicViT with SiT at FLOPs ratio 0.5SiT on DeiT-P · ImageNetTop-1 accuracy: 78.4 -> 79.4Improved
replace static aggregation (3x3 conv) with dynamic TSM (ablation)SiT-Ti · ImageNetTop-1 accuracy: 79.3 -> 80.1Improved
use learnable aggregation instead of 3x3 average pooling (ablation)SiT-Ti · ImageNetTop-1 accuracy: 77.4 -> 80.1Improved
use Token-MLP plus MLP recalibration (RTSM) instead of interpolation (ablation)SiT-Ti · ImageNetTop-1 accuracy: 78.3 -> 80.1Improved
add soft logits supervision (ablation)SiT-Ti · ImageNetTop-1 accuracy: 77.7 -> 79.0 (+1.3)Improved
add block-to-block token supervision (ablation)SiT-Ti · ImageNetTop-1 accuracy: 79.0 -> 80.1Improved
add hard label supervision (ablation)SiT-Ti · ImageNetTop-1 accuracy: 80.1 -> 80.2Improved
add FRD (feature recalibration distillation) to TSM (ablation)SiT-Ti · ImageNetTop-1 accuracy: 77.7 -> 80.1Improved
use convolutional patch embedding instead of original DeiT patch embedding at FLOPs ratio 0.25 (ablation)DeiT student · ImageNetTop-1 accuracy: 75.3 -> 78.8Improved
use FRD combined with CNN distillation at FLOPs ratio 0.5 on DeiT_C (ablation)DeiT_C student · ImageNetTop-1 accuracy: 80.0 -> 80.6Improved
use token slimming (TSM) instead of scaling model depth (ablation)SiT-Ti · ImageNetTop-1 accuracy: 69.4 -> 77.7Improved

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