Research map / Label-efficient and robust learning / Domain adaptation and generalization
Towards Effective Foundation Model Adaptation for Extreme Cross-Domain Few-Shot Learning
Fei Zhou, Peng Wang, Lei Zhang, Wei Wei, Chen Ding, Guosheng Lin, Yanning Zhang
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 masked cross-model unidirectional reconstruction (MCR) (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 39.70 -> 41.96 | Improved |
| add masked cross-model unidirectional reconstruction (MCR) (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 5-shot | accuracy (%): 64.20 -> 65.06 | Improved |
| add decision graph association (DGA) (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 39.70 -> 42.03 | Improved |
| add decision graph association (DGA) (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 5-shot | accuracy (%): 64.20 -> 64.99 | Improved |
| combine MCR and DGA rather than either alone (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 39.70 -> 43.35 | Improved |
| add MCR on ISIC (ablation) | F-ViT fine-tuning · ISIC 1-shot | accuracy (%): 26.35 -> 32.27 | Improved |
| add DGA on ISIC (ablation) | F-ViT fine-tuning · ISIC 1-shot | accuracy (%): 26.35 -> 33.11 | Improved |
| replace plain MAE reconstruction with cross-model unidirectional reconstruction plus DGA (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 41.49 -> 43.35 | Improved |
| replace bidirectional MAE with unidirectional reconstruction (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 40.56 -> 43.35 | Improved |
| replace feature or logits distillation with MCR plus DGA (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 39.90 -> 43.35 | Improved |
| replace decision fusion with MCR plus DGA (ablation) | F-ViT fine-tuning · Extreme CD-FSL average 1-shot | accuracy (%): 35.13 -> 43.35 | Improved |
| jointly train expert T-ViT to guide F-ViT instead of T-ViT alone (ablation) | T-ViT · Extreme CD-FSL average 1-shot | accuracy (%): 27.71 -> 43.35 | Improved |
Closest papers by meaning
- Flatten Long-Range Loss Landscapes for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.87)
- Leveraging Normalization Layer in Adapters with Progressive Learning and Adaptive Distillation for Cross-Domain Few-Shot Learning (AAAI 2024 · similarity 0.87)
- Unleashing the Power of Meta-tuning for Few-shot Generalization Through Sparse Interpolated Experts (ICML 2024 · similarity 0.86)
- Overcoming Generic Knowledge Loss with Selective Parameter Update (CVPR 2024 · similarity 0.86)
- Adapter Naturally Serves as Decoupler for Cross-Domain Few-Shot Semantic Segmentation (ICML 2025 · similarity 0.86)
- Towards Few-Shot Adaptation of Foundation Models via Multitask Finetuning (ICLR 2024 · 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.
