Research map / Label-efficient and robust learning / Domain adaptation and generalization
Uncertainty-Guided Model Generalization to Unseen Domains
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 uncertainty-guided feature perturbations and learnable label mixup | Single-source domain generalization · Digits: average across unseen domains | accuracy (%): 59.5 -> 61.3 | Helped |
| Use uncertainty-guided feature perturbations and learnable label mixup | Single-source domain generalization · CIFAR-10-C with WRN | accuracy (%): 80.2 -> 83.4 | Helped |
| Use uncertainty-guided feature perturbations and learnable label mixup | Single-source domain generalization · Amazon Reviews: dvd -> electronics | accuracy (%): 75.2 -> 75.0 | Hurt |
| Make feature perturbations learnable rather than fixed random Gaussian noise (ablation) | Uncertainty-guided domain generalization · Digits: average across unseen domains | accuracy (%): 51.0±0.36 -> 61.3±0.73 | Helped |
| Sample perturbations rather than apply only the learned mean (ablation) | Uncertainty-guided domain generalization · Digits: average across unseen domains | accuracy (%): 59.7±0.70 -> 61.3±0.73 | Helped |
| Learn both μ and σ instead of fixing either distribution parameter (ablation) | Uncertainty-guided domain generalization · Digits: average across unseen domains | accuracy (%): 60.5±0.75 -> 61.3±0.73; 60.7±0.65 -> 61.3±0.73 | Helped |
| Add label mixup (ablation) | Uncertainty-guided domain generalization · CIFAR-10-C | accuracy (%): 67.4±0.64 -> 70.2±0.62 | Helped |
| Learn the mixup distribution from uncertainty instead of sampling λ from a fixed Beta(1,1) (ablation) | Uncertainty-guided domain generalization · CIFAR-10-C | accuracy (%): 69.4±0.58 -> 70.2±0.62 | Helped |
| Add adversarial training for domain augmentation (ablation) | Uncertainty-guided domain generalization · Digits: average across unseen domains | accuracy (%): 51.8±0.71 -> 61.3±0.73 | Helped |
| Use meta-learning rather than train source and augmentations together (ablation) | Uncertainty-guided domain generalization · CIFAR-10-C | accuracy (%): 68.7±0.81 -> 70.2±0.62 | Helped |
| Optimize φ_p with the adversarial constraint (ablation) | Uncertainty-guided domain generalization · CIFAR-10-C | accuracy (%): 69.6±0.75 -> 70.2±0.62 | Helped |
Closest papers by meaning
- Robust Domain Generalization under Divergent Marginal and Conditional Distributions (NeurIPS 2026 · similarity 0.89)
- Understanding Domain Generalization: A Noise Robustness Perspective (ICLR 2024 · similarity 0.88)
- Set Valued Predictions For Robust Domain Generalization (ICML 2025 · similarity 0.87)
- Rethinking Multi-domain Generalization with A General Learning Objective (CVPR 2024 · similarity 0.87)
- PEER Pressure: Model-to-Model Regularization for Single Source Domain Generalization (CVPR 2025 · similarity 0.87)
- AdvST: Revisiting Data Augmentations for Single Domain Generalization (AAAI 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.