Research map / Vision-language and multimodal / CLIP and zero-shot
Libra-MIL: Multimodal Prototypes Stereoscopic Infused with Task-specific Language Priors for Few-shot Whole Slide Image Classification
Zhenfeng Zhuang, Fangyu Zhou, Liansheng 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 |
|---|---|---|---|
| add instance-level language priors (ablation) | Libra-MIL · TCGA-RCC 1-shot | accuracy (%): 67.0 -> 71.2 | Improved |
| add vision prototypes (ablation) | Libra-MIL · TCGA-RCC 1-shot | accuracy (%): 66.0 -> 71.2 | Improved |
| use SOT fusion instead of concatenation (ablation) | Libra-MIL · TCGA-RCC 1-shot | accuracy (%): 66.4 -> 71.2 | Improved |
| use SOT fusion instead of cross-attention (ablation) | Libra-MIL · TCGA-RCC 4-shot | accuracy (%): 88.8 -> 91.4 | Improved |
| add bag priors in semantic aggregation (ablation) | Libra-MIL · TCGA-RCC 1-shot | accuracy (%): 55.4 -> 71.2 | Improved |
| add bag priors in semantic aggregation (ablation) | Libra-MIL · TCGA-RCC 4-shot | accuracy (%): 88.0 -> 91.4 ACC, 88.5 -> 88.5 F1 | No clear effect |
| add instance priors (ablation) | Libra-MIL · TCGA-RCC 16-shot | accuracy (%): 92.0 -> 93.8 | No clear effect |
| use CONCH instead of Phikon/GigaPath/UNI vision encoder (ablation) | Libra-MIL · TCGA-NSCLC 4-shot | accuracy (%): 72.9 -> 83.2 | Improved |
| vary number of vision prototypes Kv (ablation) | Libra-MIL · TCGA-NSCLC 4-shot | accuracy (%): 83.2 at Kv=6, 79.5 at Kv=12 | Mixed |
| Libra-MIL vs best baseline (FOCUS) | Libra-MIL · CAMELYON16 4-shot | accuracy (%): 61.2 -> 65.0 | Improved |
| Libra-MIL vs best baseline (CLAM MB) | Libra-MIL · TCGA-NSCLC 16-shot | accuracy (%): 90.7 -> 90.8 | No clear effect |
Closest papers by meaning
- ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification (CVPR 2024 · similarity 0.90)
- Few-Shot Learning from Gigapixel Images via Hierarchical Vision-Language Alignment and Modeling (NeurIPS 2025 · similarity 0.89)
- SlideChat: A Large Vision-Language Assistant for Whole-Slide Pathology Image Understanding (CVPR 2025 · similarity 0.88)
- MLLM-HWSI: A Multimodal Large Language Model for Hierarchical Whole Slide Image Understanding (CVPR 2026 · similarity 0.88)
- PathFLIP: Fine-grained Language-Image Pretraining for Versatile Computational Pathology (AAAI 2026 · similarity 0.88)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.88)
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.
