Research map / Label-efficient and robust learning / Recognition and long-tail
Post-Hoc Uncertainty Calibration for Domain Drift Scenarios
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 |
|---|---|---|---|
| Tune post-hoc calibrators on a Gaussian-perturbed validation set instead of the standard validation set | Post-hoc calibrators on CIFAR-10 and ImageNet · Domain drift scenarios | ECE | Helped |
| Use perturbed-validation tuning with IR | IR on CIFAR-10 and ImageNet · In-domain through large domain shifts | ECE | Helped |
| Use perturbed-validation tuning with temperature-scaling methods (ablation) | TS, ETS, and TS-IR · ImageNet in-domain and small domain shifts | ECE | Hurt |
| Use perturbed-validation tuning with temperature-scaling methods (ablation) | TS and ETS · Overall domain drift scenarios | ECE | Helped |
| Use perturbed-validation tuning with IRM rather than IR (ablation) | Isotonic-regression calibrators · Domain drift, especially small shifts | ECE | Hurt |
| Use image-specific validation perturbations (-H) instead of generic Gaussian perturbations (ablation) | Calibrators on CIFAR-10 VGG19 · Affine domain drift scenarios | ECE: 0.127 vs 0.049 | Hurt |
| Use IR-P with a small validation set rather than a sufficiently large one (ablation) | IR-P · CIFAR-10 | accuracy | Hurt |
| Add a perturbed validation set when tuning additional calibrators (ablation) | Platt scaling, histogram binning, and PBMC · CIFAR-10 affine perturbations | ECE | Helped |
Closest papers by meaning
- A Benchmark Study on Calibration (ICLR 2024 · similarity 0.89)
- Pseudo-Calibration: Improving Predictive Uncertainty Estimation in Unsupervised Domain Adaptation (ICML 2024 · similarity 0.88)
- Calibration Bottleneck: Over-compressed Representations are Less Calibratable (ICML 2024 · similarity 0.87)
- Consistency-Guided Temperature Scaling Using Style and Content Information for Out-of-Domain Calibration (AAAI 2024 · similarity 0.87)
- Uncertainty Weighted Gradients for Model Calibration (CVPR 2025 · similarity 0.87)
- Towards Certification of Uncertainty Calibration under Adversarial Attacks (ICLR 2025 · 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.