Research map / Vision-language and multimodal / CLIP and zero-shot
BioDPP: Dynamic Prompt Policy Learning for Biomedical Vision-Language Models
Pingyi Miao, Xianlai Chen, Kai Sun, Yunbo Wang, Shuang Zhao, Ying An
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 semantic representation reward Rrep to classification reward (ablation) | BioDPP · RETINA base-to-novel | harmonic mean (%): 64.41 -> 67.00 | Improved |
| add adaptive baseline stabilization on top of Rcls + Rrep (ablation) | BioDPP · RETINA base-to-novel | harmonic mean (%): 67.00 -> 67.41 | Improved |
| add ABS on top of Rcls + Rrep (ablation) | BioDPP · RETINA 1-shot | accuracy (%): 34.99 -> 37.15 | Improved |
| add ABS on top of Rcls + Rrep (ablation) | BioDPP · RETINA 16-shot | accuracy (%): 65.56 -> 66.32 | Improved |
| add Rrep to Rcls (ablation) | BioDPP · RETINA few-shot 1-shot | accuracy (%): 34.20 -> 34.99 | Improved |
| BioDPP dynamic prompt policy vs BiomedCoOp | BiomedCLIP · 10 datasets avg, 16-shot | accuracy (%): 72.70 -> 73.61 | Improved |
| BioDPP vs BiomedCoOp on base-to-novel generalization | BiomedCLIP · Base-to-novel average | harmonic mean (%): 73.35 -> 76.75 | Improved |
| use BiomedCLIP ViT-B/16 backbone instead of other CLIP-based backbones (ablation) | BioDPP · BUSI | accuracy (%): zero-shot 59.8, 16-shot 69.8 (best) | Improved |
| use shorter context length (4) (ablation) | BioDPP · BTMRI base-to-novel | harmonic mean (%): HM 89.21 at length 4 | Improved |
| increase number of LLM prompts beyond 50 (ablation) | BioDPP · 8-shot | accuracy (%): 79.95 at 50 prompts, saturating | No clear effect |
Closest papers by meaning
- BiomedCoOp: Learning to Prompt for Biomedical Vision-Language Models (CVPR 2025 · similarity 0.93)
- LLaDA-MedV: Exploring Large Language Diffusion Models for Biomedical Image Understanding (CVPR 2026 · similarity 0.87)
- IPO: Interpretable Prompt Optimization for Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- ProAPO: Progressively Automatic Prompt Optimization for Visual Classification (CVPR 2025 · similarity 0.86)
- ZIP: An Efficient Zeroth-order Prompt Tuning for Black-box Vision-Language Models (ICLR 2025 · similarity 0.86)
- pMoE: Prompting Diverse Experts Together Wins More in Visual Adaptation (ICLR 2025 · similarity 0.86)
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.
