Research map / Label-efficient and robust learning / Recognition and long-tail
The Rise of AI Language Pathologists: Exploring Two-level Prompt Learning for Few-shot Weakly-supervised Whole Slide Image Classification
Linhao Qu, xiaoyuan Luo, Kexue Fu, Manning Wang, Zhijian Song
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 attention pooling with instance prompt guided pooling (with bag prompt) (ablation) | TOP · Camelyon16 bag, 1-shot | AUC: 0.6483 -> 0.6783 | Improved |
| replace attention pooling with prompt guided pooling (with CoOp prompt) (ablation) | CoOp · Camelyon16 bag, 16-shot | AUC: 0.8012 -> 0.8216 | Improved |
| replace fully learnable CoOp prompt with GPT-4 based bag prompt group (with prompt guided pooling) (ablation) | TOP · Camelyon16 bag, 8-shot | AUC: 0.7079 -> 0.7287 | Improved |
| replace CoOp prompt with bag prompt group (with attention pooling) (ablation) | Attention-pooling baseline · Camelyon16 instance, 16-shot | AUC: 0.8567 -> 0.8699 | Improved |
| add auxiliary prototype-decorrelation loss (weight 25 vs 0) (ablation) | TOP · Camelyon16 bag, 8-shot | AUC: 0.7137 -> 0.7287 | Improved |
| add auxiliary loss (weight 25 vs 0) (ablation) | TOP · Camelyon16 instance, 16-shot | AUC: 0.8834 -> 0.8896 | Improved |
| auxiliary loss weight 50 instead of 25 (ablation) | TOP · Camelyon16 bag, 1-shot | AUC: 0.6783 -> 0.6771 | No clear effect |
| use TOP two-level prompt learning instead of CoOp (Attention-pooling) | CLIP few-shot WSI · Camelyon16 bag, 1-shot | AUC: 0.6462 -> 0.6783 | Improved |
| use TOP instead of CoOp (Attention-pooling) | CLIP few-shot WSI · TCGA-Lung bag, 8-shot | AUC: 0.6824 -> 0.8059 | Improved |
| use TOP instead of CoOp (Attention-pooling) | CLIP few-shot WSI · Cervical bag, 16-shot | AUC: 0.7565 -> 0.8189 | Improved |
| use attention pooling instead of mean pooling for linear probe | Linear-Probe · Camelyon16 bag, 16-shot | AUC: 0.5816 -> 0.7866 | Improved |
Closest papers by meaning
- ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.92)
- MAPLE: Multi-scale Attribute-enhanced Prompt Learning for Few-shot Whole Slide Image Classification (NeurIPS 2025 · similarity 0.91)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.91)
- FAST: A Dual-tier Few-Shot Learning Paradigm for Whole Slide Image Classification (NeurIPS 2024 · similarity 0.89)
- Few-Shot Learning from Gigapixel Images via Hierarchical Vision-Language Alignment and Modeling (NeurIPS 2025 · similarity 0.89)
- Queryable Prototype Multiple Instance Learning with Vision-Language Models for Incremental Whole Slide Image Classification (AAAI 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.
