Research map / Architectures and efficiency / Pruning and efficient tuning
E^2VPT: An Effective and Efficient Approach for Visual Prompt Tuning
Cheng Han, Qifan Wang, Yiming Cui, Zhiwen Cao, Wenguan Wang, Siyuan Qi, Dongfang Liu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add key-value prompts in self-attention (ablation) | E2VPT (visual prompts) · VTAB-1k SVHN | accuracy: 78.1% -> 83.8% | Improved |
| add pruning and rewinding of input prompts (ablation) | E2VPT (visual prompts) · VTAB-1k SVHN | accuracy: 78.1% -> 79.0% | Improved |
| add pruning and rewinding on top of visual and key-value prompts (ablation) | E2VPT · VTAB-1k SVHN | accuracy: 83.8% -> 85.3% | Improved |
| add key-value prompts and pruning to visual prompts (ablation) | E2VPT · FGVC NABirds | accuracy: 84.2% -> 84.6% | Improved |
| increase key-value prompts to 20 (versus 5) (ablation) | E2VPT · VTAB-1k SVHN | accuracy: 85.3% -> 80.2% | Hurt |
| place key-value prompts before rather than after K and V (ablation) | E2VPT · VTAB-1k | average accuracy: 73.91% -> 73.94% | No clear effect |
| use He initialization instead of truncated normal (ablation) | E2VPT · VTAB-1k | accuracy: Natural 79.77% -> 80.01% | Mixed |
| replace VPT input-only prompts with E2VPT on MAE-pretrained ViT | ViT-Base MAE · VTAB-1k Natural | accuracy: 36.02% -> 59.52% | Improved |
Later papers that compared with this method
2 later accepted papers report E2VPT as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- CVPT: Cross Visual Prompt Tuning (ICCV 2025 · better on 1 of 1 shared result)
- Dropout Mixture Low-Rank Adaptation for Visual Parameters-Efficient Fine-Tuning (ECCV 2024 · better on 1 of 1 shared result)
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
- DA-VPT: Semantic-Guided Visual Prompt Tuning for Vision Transformers (CVPR 2025 · similarity 0.91)
- Facing the Elephant in the Room: Visual Prompt Tuning or Full finetuning? (ICLR 2024 · similarity 0.90)
- LoR-VP: Low-Rank Visual Prompting for Efficient Vision Model Adaptation (ICLR 2025 · similarity 0.90)
- Learning to Learn Better Visual Prompts (AAAI 2024 · similarity 0.89)
- Revisiting the Power of Prompt for Visual Tuning (ICML 2024 · similarity 0.89)
- Visual Fourier Prompt Tuning (NeurIPS 2024 · similarity 0.89)
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.
