Research map / Label-efficient and robust learning / Domain adaptation and generalization
Transferability Metrics for Selecting Source Model Ensembles
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use SoftIoU-EEP rather than the target-agnostic BASE metric to rank ensembles | Transferability-based ensemble selection · 5 target datasets (average) | weighted Kendall's τ: 0.49 -> 0.69 | Helped |
| Use SoftIoU-EEP rather than MS-LEEP to rank ensembles (ablation) | Transferability-based ensemble selection · 5 target datasets (average) | weighted Kendall's τ: 0.66 -> 0.69 | Helped |
| Use SoftIoU-EEP rather than E-LEEP to rank ensembles (ablation) | Transferability-based ensemble selection · 5 target datasets (average) | weighted Kendall's τ: 0.66 -> 0.69 | Helped |
| Use SoftIoU-EEP rather than IoU-EEP to rank ensembles (ablation) | Transferability-based ensemble selection · 5 target datasets (average) | weighted Kendall's τ: 0.61 -> 0.69 | Helped |
| Use an ensemble selected by SoftIoU-EEP instead of selecting one model from the same pool | Transfer learning for semantic segmentation · 17 target datasets | relative mean IoU: +6.0% | Helped |
| Use an ensemble selected by SoftIoU-EEP instead of one large, capacity-matched model | Transfer learning for semantic segmentation · 17 target datasets | relative mean IoU: +2.5% | Helped |
| Combine three selected source models rather than using the ensemble's best individual member | SoftIoU-EEP-selected ensemble · 17 target datasets | relative mean IoU: 4.6% | Helped |
| Combine three selected source models rather than using the ensemble's worst individual member | SoftIoU-EEP-selected ensemble · 17 target datasets | relative mean IoU: 18.6% | Helped |
| Use an unweighted average instead of the learned combination head (ablation) | Selected ensemble · 17 target datasets | relative mean IoU gain over best member: +2.9% | Helped |
| Use probability-distribution-based E-LEEP instead of hard-label IoU-EEP (ablation) | Transferability-based ensemble selection · 5 target datasets (average) | weighted Kendall's τ: 0.61 -> 0.66 | Helped |
Closest papers by meaning
- Transferability Estimation Using Bhattacharyya Class Separability (CVPR 2022 · similarity 0.88)
- Scalable Diverse Model Selection for Accessible Transfer Learning (NeurIPS 2021 · similarity 0.87)
- H-ensemble: An Information Theoretic Approach to Reliable Few-Shot Multi-Source-Free Transfer (AAAI 2024 · similarity 0.85)
- Not All Areas Are Equal: Transfer Learning for Semantic Segmentation via Hierarchical Region Selection (CVPR 2019 · similarity 0.85)
- A High-Dimensional Statistical Method for Optimizing Transfer Quantities in Multi-Source Transfer Learning (NeurIPS 2025 · similarity 0.85)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.84)
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.