Research map / Vision-language and multimodal / CLIP and zero-shot
Visual Prompt-Agnostic Evolution
Junze Wang, Lei Fan, Dezheng Zhang, Weipeng Jing, Donglin Di, Yang Song, Sidong Liu, Cong Cong
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 PAE | VPT, ViT-B/16 · VTAB-1k mean (ViT-B/16) | accuracy: 71.96 -> 74.84 | Improved |
| add PAE | VPT, ViT-B/16 · FGVC | accuracy: 89.11 -> 91.02 | Improved |
| add PAE | SA2VP, ViT-B/16 · VTAB-1k mean (ViT-B/16) | accuracy: 75.83 -> 77.49 | Improved |
| add PAE | VPT · convergence speed | speedup: 1.78x | Improved |
| add MPA init alone (ablation) | VPT · VTAB-1k mean (ViT-B/16) | accuracy: 71.96 -> 74.02 | Improved |
| add Koopman consistency loss alone (ablation) | VPT · VTAB-1k mean (ViT-B/16) | accuracy: 71.96 -> 73.13 | Improved |
| add Lyapunov stability loss on top of L_kp (ablation) | VPT+L_kp · VTAB-1k mean (ViT-B/16) | accuracy: 73.13 -> 74.42 | Improved |
| add MPA on top of L_kp and L_stab (ablation) | VPT+KLD · VTAB-1k mean (ViT-B/16) | accuracy: 74.42 -> 74.84 | Improved |
| use MPA instead of random init (ablation) | PAE · VTAB-1k mean (ViT-B/16) | accuracy: 71.96 -> 74.84 | Improved |
| use MPA instead of Xavier init (ablation) | PAE · VTAB-1k mean (ViT-B/16) | accuracy: 72.30 -> 74.84 | Improved |
| use MPA instead of unsupervised prototype init (ablation) | PAE · VTAB-1k mean (ViT-B/16) | accuracy: 73.85 -> 74.84 | Improved |
| propagate first-layer prompt through frozen backbone instead of copying it (ablation) | PAE · VTAB-1k mean (ViT-B/16) | accuracy: 73.17 -> 74.84 | Improved |
Closest papers by meaning
- Revisiting the Power of Prompt for Visual Tuning (ICML 2024 · similarity 0.92)
- Facing the Elephant in the Room: Visual Prompt Tuning or Full finetuning? (ICLR 2024 · similarity 0.91)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.90)
- DA-VPT: Semantic-Guided Visual Prompt Tuning for Vision Transformers (CVPR 2025 · similarity 0.90)
- Black-Box Test-Time Prompt Tuning for Vision-Language Models (AAAI 2025 · similarity 0.89)
- VIPAMIN: Visual Prompt Initialization via Embedding Selection and Subspace Expansion (NeurIPS 2025 · 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.
