Research map / Label-efficient and robust learning / Domain adaptation and generalization
Guiding Pseudo-Labels With Uncertainty Estimation for Source-Free Unsupervised Domain Adaptation
Read the paperCheck 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 uncertainty-weighted pseudo-label adaptation | Source-free UDA · PACS single-source average | accuracy: 72.4 -> 84.7 | Helped |
| Use uncertainty-weighted pseudo-label adaptation | Source-free UDA · PACS multi-target average | accuracy: 68.4 -> 75.0 | Helped |
| Use uncertainty-weighted pseudo-label adaptation | Source-free UDA · VisDA-C | accuracy: 88.2 -> 90.0 | Helped |
| Use uncertainty-weighted pseudo-label adaptation | Source-free UDA · DomainNet-126 average | accuracy: 67.8 -> 69.6 | Helped |
| Add contrastive regularisation to pseudo-label refinement (ablation) | VisDA-C adaptation · VisDA-C | accuracy: 52.3 -> 78.9; +26.6% | Helped |
| Add negative learning loss (ablation) | VisDA-C adaptation with pseudo-label refinement and contrastive regularisation · VisDA-C | accuracy: 78.9 -> 82.1; +3.2% | Helped |
| Add temporal-queue negative-pair exclusion (ablation) | VisDA-C adaptation with negative learning · VisDA-C | accuracy: 82.1 -> 85.8; +3.7% | Helped |
| Add uncertainty reweighting (ablation) | VisDA-C adaptation with temporal-queue exclusion · VisDA-C | accuracy: 85.8 -> 90.0; 4.2% | Helped |
| Use exponential rather than linear uncertainty weighting (ablation) | VisDA-C adaptation · VisDA-C | accuracy: 85.1 -> 90.0; +4.9% | Helped |
| Use smooth uncertainty weighting rather than hard entropy-margin selection (ablation) | VisDA-C adaptation · VisDA-C | accuracy: 85.9 -> 90.0 | Helped |
| Use negative learning instead of positive classification loss (ablation) | VisDA-C adaptation · VisDA-C | accuracy: 83.0 -> 90.0; +4.8% | Helped |
| Use only negative learning rather than a positive-plus-negative loss (ablation) | VisDA-C adaptation · VisDA-C | accuracy: 85.2 -> 90.0 | Helped |
Closest papers by meaning
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.91)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.90)
- Toward Improving Robustness and Accuracy in Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.90)
- Pseudo-Calibration: Improving Predictive Uncertainty Estimation in Unsupervised Domain Adaptation (ICML 2024 · similarity 0.90)
- Pseudo-Labeling for Unsupervised Domain Adaptation with Kernel GLMs (NeurIPS 2026 · similarity 0.90)
- Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias (CVPR 2024 · similarity 0.90)
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.