AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

MIA-Former: Efficient and Robust Vision Transformers via Multi-Grained Input-Adaptation

Zhongzhi Yu, Yonggan Fu, Sicheng Li, Chaojian Li, Yingyan Lin

AAAI 2022 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 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
Apply multi-grained input-adaptive skipping of blocks, heads, and tokensDeiT-Small · ImageNet-1KFLOPs: 20.1% reductionImproved
Apply multi-grained input-adaptive skippingDeiT-Small · ImageNet-1K under PGD attackrobust accuracy: 2.4% higherImproved
Apply multi-grained input-adaptive skippingDeiT-Small · ImageNet-1K under FGSM attackrobust accuracy: 3.0% higherImproved
Enable only head-wise input-adaptive skipping instead of all three dimensions (ablation)MIA-Former on DeiT-Small · ImageNet-1Kaccuracy: 79.9 -> 78.7Hurt
Enable only depth-wise input-adaptive skipping instead of all three dimensions (ablation)MIA-Former on DeiT-Small · ImageNet-1Kaccuracy: 79.9 -> 76.2Hurt
Enable only token-wise input-adaptive skipping instead of all three dimensions (ablation)MIA-Former on DeiT-Small · ImageNet-1Kaccuracy: 79.9 -> 79.7Hurt
Use a 75% inherited controller weight initialization rather than 50% (ablation)MIA-Former RL skipping-policy fine-tuning · ImageNet-1K, target 3.4 GFLOPsrobust accuracy: 21.3 -> 19.8Hurt
Use a 75% inherited controller weight initialization rather than 50% (ablation)MIA-Former RL skipping-policy fine-tuning · ImageNet-1K, target 3.4 GFLOPsclean accuracy: 79.2 -> 79.9Improved
Use a 100% inherited controller weight initialization rather than 75% (ablation)MIA-Former RL skipping-policy fine-tuning · ImageNet-1K, target 3.4 GFLOPsrobust accuracy: 19.8 -> 15.7Hurt
Use a 100% inherited controller weight initialization rather than 75% (ablation)MIA-Former RL skipping-policy fine-tuning · ImageNet-1K, target 3.4 GFLOPsclean accuracy: 79.9 -> 79.9No clear effect
Use a 75% inherited controller weight initialization rather than 50% (ablation)MIA-Former RL skipping-policy fine-tuning · ImageNet-1K, target 3.9 GFLOPsrobust accuracy: 22.8 -> 20.2Hurt
Use a 75% inherited controller weight initialization rather than 50% (ablation)MIA-Former RL skipping-policy fine-tuning · ImageNet-1K, target 3.9 GFLOPsclean accuracy: 78.3 -> 79.6Improved

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