AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Hint-Aug: Drawing Hints From Foundation Vision Transformers Towards Boosted Few-Shot Parameter-Efficient Tuning

Zhongzhi Yu, Shang Wu, Yonggan Fu, Shunyao Zhang, Yingyan (Celine) Lin

CVPR 2023 · Architectures and efficiency › Pruning and efficient tuning · approach: Efficient ViT adaptation · 5 citations (OpenAlex)

Read the paper

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 Hint-Aug rather than the strongest listed augmentation baselineAdapter tuning · MAE-pretrained ViT-Base, 8-shot and 16-shot benchmarksaccuracy: 0.31% \sim 3.60% (8-shot); 0.45% \sim 3.27% (16-shot)Improved
Apply Hint-Aug rather than the strongest listed augmentation baselineVPT tuning · MAE-pretrained ViT-Base, 8-shot and 16-shot benchmarksaccuracy: 0.31% \sim 3.60% (8-shot); 0.45% \sim 3.27% (16-shot)Improved
Apply Hint-Aug rather than the strongest listed augmentation baselineLoRA tuning · MAE-pretrained ViT-Base, 8-shot and 16-shot benchmarksaccuracy: 0.31% \sim 3.60% (8-shot); 0.45% \sim 3.27% (16-shot)Improved
Apply Hint-AugAdapter tuning · Caltech-101, CIFAR-100, DTD, Flowers, Pets transfer learningaccuracy: as high as 2.47%Improved
Apply Hint-AugVPT tuning · Caltech-101, CIFAR-100, DTD, Flowers, Pets transfer learningaccuracy: 0.90%Improved
Apply Hint-AugLoRA tuning · Caltech-101, CIFAR-100, DTD, Flowers, Pets transfer learningaccuracy: 3.77%Improved
Vary λ among 0.2, 0.1, and 0.05 (ablation)Hint-Aug + Adapter · Food, 8-shotaccuracy: 70.88; 71.04; 70.85No clear effect

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