Research map / Architectures and efficiency / Pruning and efficient tuning
CVPT: Cross Visual Prompt Tuning
Lingyun Huang, Jianxu Mao, Junfei Yi, Ziming Tao, Yaonan Wang
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 |
|---|---|---|---|
| replace concatenated prompts with cross-attention prompts | VPT-Deep · VTAB-1k avg | accuracy: 72.0 -> 77.2 | Improved |
| replace concatenated prompts with cross-attention prompts | VPT · FGVC avg | accuracy: 89.1 -> 90.5 | Improved |
| replace concatenated prompts with cross-attention prompts | VPT · ADE20K ViT-L | mIoU-SS: 42.11 -> 45.66 | Improved |
| use cross-attention with 200 prompts instead of VPT (ablation) | VPT · VTAB-1k avg (200 prompts) | accuracy: 64.0 -> 74.8 | Improved |
| use cross-attention with 1 prompt instead of VPT (ablation) | VPT · VTAB-1k avg (1 prompt) | accuracy: 71.0 -> 69.5 | Hurt |
| use cross-attention with 200 prompts instead of VPT (ablation) | VPT · CIFAR-100 VTAB-1k | accuracy: 35.7 -> 72.1 | Improved |
| increase prompt count from 1 to 200 (ablation) | CVPT · VTAB-1k avg | accuracy: 69.5 -> 74.8 | Improved |
| increase prompt count from 5 to 200 (ablation) | VPT · VTAB-1k avg | accuracy: 73.0 -> 64.0 | Hurt |
| add weight sharing from self-attention to cross-attention (frozen CA) (ablation) | CVPT · VTAB-1k avg | accuracy: 63.7 -> 74.0 | Improved |
| add weight sharing to learnable cross-attention (ablation) | CVPT · VTAB-1k avg | accuracy: 74.0 -> 74.6 | Improved |
| freeze shared cross-attention instead of making it learnable (ablation) | CVPT · VTAB-1k avg | accuracy: 74.6 -> 74.0 | No clear effect |
| use CVPT instead of DM-LoRA | CVPT · VTAB-1k avg | accuracy: 77.0 -> 77.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- E2VPT — E^2VPT: An Effective and Efficient Approach for Visual Prompt Tuning (ICCV 2023) · better on 1 of 1 shared result
- AdaptFormer — AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognition (NeurIPS 2022) · 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
- Facing the Elephant in the Room: Visual Prompt Tuning or Full finetuning? (ICLR 2024 · similarity 0.91)
- DA-VPT: Semantic-Guided Visual Prompt Tuning for Vision Transformers (CVPR 2025 · similarity 0.89)
- Visual Fourier Prompt Tuning (NeurIPS 2024 · similarity 0.89)
- Fair-VPT: Fair Visual Prompt Tuning for Image Classification (CVPR 2024 · similarity 0.89)
- Revisiting the Power of Prompt for Visual Tuning (ICML 2024 · similarity 0.88)
- LoR-VP: Low-Rank Visual Prompting for Efficient Vision Model Adaptation (ICLR 2025 · similarity 0.87)
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.
