Research map / Label-efficient and robust learning / Domain adaptation and generalization
Interpretable Cross-Domain Few-Shot Learning with Rectified Target-Domain Local Alignment
Yaze Zhao, Yixiong Zou, Yuhua Li, Ruixuan Li
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 CC-CDFSL cycle-consistency regularizer to CLIP-LoRA | CLIP-LoRA ViT · CDFSL 5-way 1-shot avg | accuracy (%): 55.89 -> 58.83 | Improved |
| add CC-CDFSL to CLIP-LoRA | CLIP-LoRA ViT · CDFSL 5-way 5-shot avg | accuracy (%): 65.99 -> 67.90 | Improved |
| add CC-CDFSL to CoOp | CoOp · CDFSL 5-way 5-shot avg | accuracy (%): 60.95 -> 61.44 | Improved |
| add CC-CDFSL to Maple | Maple · CDFSL 5-way 1-shot avg | accuracy (%): 53.04 -> 54.89 | Improved |
| add CC-CDFSL to CLIP-LoRA on ChestX | CLIP-LoRA ViT · ChestX 5-way 1-shot | accuracy (%): 21.73 -> 22.21 | Improved |
| add all components (T-I-T, I-T-I, SA) (ablation) | CLIP-LoRA · CDFSL 5-way 5-shot avg | accuracy (%): 65.98 -> 67.90 | Improved |
| use cross-view instead of intra-image retrieval in I-T-I (ablation) | CC-CDFSL · CDFSL 5-way 5-shot avg | accuracy (%): 67.25 -> 67.37 | Improved |
| use cross-view instead of all-images retrieval in I-T-I (ablation) | CC-CDFSL · CDFSL 5-way 5-shot avg | accuracy (%): 67.19 -> 67.37 | Improved |
| increase anchor patch number k beyond 13 (ablation) | CC-CDFSL · ISIC2018 5-way 5-shot | accuracy (%): 54.72 -> 54.21 | Hurt |
| use very small anchor patch number k<=3 (ablation) | CC-CDFSL · ISIC2018 5-way 5-shot | accuracy (%): 54.67 -> 54.28 | Hurt |
| add cycle consistency to CLIP-LoRA for prototype classification | CLIP-LoRA · CropDiseases | accuracy (%): 95.85 -> 96.24 | Improved |
| add cycle consistency to Maple for base-to-new generalization | Maple · EuroSAT new classes | accuracy (%): +3.6% | Improved |
Closest papers by meaning
- Random Registers for Cross-Domain Few-Shot Learning (ICML 2025 · similarity 0.89)
- Attention Temperature Matters in ViT-Based Cross-Domain Few-Shot Learning (NeurIPS 2024 · similarity 0.89)
- A Closer Look at the CLS Token for Cross-Domain Few-Shot Learning (NeurIPS 2024 · similarity 0.89)
- Adapter Naturally Serves as Decoupler for Cross-Domain Few-Shot Semantic Segmentation (ICML 2025 · similarity 0.89)
- Revisiting Continuity of Image Tokens for Cross-domain Few-shot Learning (ICML 2025 · similarity 0.88)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (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.
