Research map / Label-efficient and robust learning / Domain adaptation and generalization
Sparse Task Vector Mixup with Hypernetworks for Efficient Knowledge Transfer in Whole-Slide Image Prognosis
Pei Liu, Xiangxiang Zeng, Tengfei Ma, Yucheng Xing, Xuanbai Ren, Yiping Liu
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 |
|---|---|---|---|
| merge cross-cancer task vectors with STEPH instead of cancer-specific training | cancer-specific model · TCGA 13 cancers avg | C-Index: 0.6609 -> 0.6949 | Improved |
| STEPH instead of representation-based transfer ROUPKT | ROUPKT · TCGA 13 cancers avg | C-Index: 0.6812 -> 0.6949 | Improved |
| STEPH vs ROUPKT on COAD | ROUPKT · TCGA-COAD | C-Index: 0.6708 -> 0.6698 | Hurt |
| add task vector mixup between target and source vectors (ablation) | STEPH · TCGA 13 cancers avg | C-Index: 0.6860 -> 0.6949 | Improved |
| use hypernetwork-driven lambda instead of trainable parameter (ablation) | STEPH · TCGA 13 cancers avg | C-Index: 0.6921 -> 0.6949 | Improved |
| use hypernetwork-driven w instead of trainable parameter w (ablation) | STEPH · TCGA 13 cancers avg | C-Index: 0.6490 -> 0.6949 | Improved |
| add sparsity (top-K) to aggregation (ablation) | STEPH · TCGA 13 cancers avg | C-Index: 0.6912 -> 0.6949 | Improved |
| add hypernetwork-driven w to existing merging methods (ablation) | AdaMerging · TCGA 13 cancers avg | C-Index: 0.5689 -> 0.6877 | Improved |
| add hypernetwork-driven w to TIES (ablation) | TIES AM · TCGA 13 cancers avg | C-Index: 0.6396 -> 0.6802 | Improved |
| use existing merging methods (Model Avg, AdaMerging, Surgery, Iso-C) for cross-cancer transfer | target model · TCGA 13 cancers avg | C-Index: 0.6609 -> 0.5689..0.6396 | Hurt |
| add mixup coefficient penalty (beta=0.05 vs 0) (ablation) | STEPH · TCGA 13 cancers avg | C-Index: 0.6927 -> 0.6949 | No clear effect |
| add aggregation weight penalty (gamma>0 vs 0) (ablation) | STEPH · TCGA 13 cancers avg | C-Index: 0.6879 -> 0.6896 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AdaMerging — AdaMerging: Adaptive Model Merging for Multi-Task Learning (ICLR 2024) · 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
- Interpretable Vision-Language Survival Analysis with Ordinal Inductive Bias for Computational Pathology (ICLR 2025 · similarity 0.87)
- Learning Heterogeneous Tissues with Mixture of Experts for Gigapixel Whole Slide Images (CVPR 2025 · similarity 0.86)
- Leveraging Tumor Heterogeneity: Heterogeneous Graph Representation Learning for Cancer Survival Prediction in Whole Slide Images (NeurIPS 2024 · similarity 0.86)
- Revisiting End-to-End Learning with Slide-level Supervision in Computational Pathology (NeurIPS 2025 · similarity 0.85)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.85)
- Tumor Micro-environment Interactions Guided Graph Learning for Survival Analysis of Human Cancers from Whole-slide Pathological Images (CVPR 2024 · similarity 0.85)
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.
