Research map / Label-efficient and robust learning / Domain adaptation and generalization
Step-Wise Distribution-Aligned Style Prompt Tuning for Source-Free Cross-Domain Few-Shot Learning
Huali Xu, Li Liu, Tianpeng Liu, Shuaifeng Zhi, Shuzhou Sun, Ming‐Ming Cheng
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 style prompt to step-wise alignment (ablation) | StepSPT (ConvNeXt) · 5-way 1-shot avg | accuracy: 63.98 -> 65.31 | Improved |
| add style prompt to step-wise alignment (ablation) | StepSPT (ConvNeXt) · 5-way 5-shot avg | accuracy: 73.10 -> 73.70 | Improved |
| add step-wise distribution alignment to style prompt (ablation) | StepSPT (ConvNeXt) · 5-way 1-shot avg | accuracy: 64.11 -> 65.31 | Improved |
| add step-wise distribution alignment to style prompt (ablation) | StepSPT (ConvNeXt) · 5-way 5-shot avg | accuracy: 73.50 -> 73.70 | Improved |
| add style prompt only (no step-wise alignment) to baseline (ablation) | ConvNeXt linear baseline · 5-way 1-shot avg | accuracy: 60.95 -> 64.11 | Improved |
| add step-wise distribution alignment only (no style prompt) to baseline (ablation) | ConvNeXt linear baseline · 5-way 1-shot avg | accuracy: 60.95 -> 63.98 | Improved |
| add step-wise distribution alignment only to baseline (ablation) | ConvNeXt linear baseline · 5-way 5-shot avg | accuracy: 72.17 -> 73.10 | Improved |
| add step-wise distribution alignment only to baseline, ISIC 5-shot (ablation) | ConvNeXt linear baseline · ISIC 5-way 5-shot | accuracy: 52.51 -> 52.24 | No clear effect |
| StepSPT style prompt vs CoOp/CoCoOp text prompts in same pipeline (ablation) | StepSPT · 5-way 1-shot and 5-shot | average accuracy | Improved |
| step size E=100 vs E=2 in step-wise alignment (ablation) | StepSPT (ConvNeXt) · 5-way 1-shot avg | accuracy: 64.36 -> 65.31 | Improved |
| use StepSPT instead of LoRA on frozen CLIP | CLIP · 5-way 1-shot avg | accuracy: 61.80 -> 61.16 | Hurt |
| use StepSPT instead of LoRA on frozen ConvNeXt | ConvNeXt · 5-way 1-shot avg | accuracy: 64.31 -> 65.31 | Improved |
Closest papers by meaning
- Random Registers for Cross-Domain Few-Shot Learning (ICML 2025 · similarity 0.86)
- SVasP: Self-Versatility Adversarial Style Perturbation for Cross-Domain Few-Shot Learning (AAAI 2025 · similarity 0.85)
- Flatten Long-Range Loss Landscapes for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.85)
- A Closer Look at the CLS Token for Cross-Domain Few-Shot Learning (NeurIPS 2024 · similarity 0.84)
- TAMT: Temporal-Aware Model Tuning for Cross-Domain Few-Shot Action Recognition (CVPR 2025 · similarity 0.84)
- Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learning (CVPR 2024 · 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.
