Research map / Label-efficient and robust learning / Test-time adaptation
A Learning Based Hypothesis Test for Harmful Covariate Shift
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 constrained-disagreement classifiers instead of a standard deep ensemble (ablation) | CIFAR-10.1 shift detection, |Q|=10 · CIFAR-10.1 | TPR@5: .33 ± .05 -> .37 ± .05 | Helped |
| Use constrained-disagreement classifiers instead of a standard deep ensemble (ablation) | CIFAR-10.1 shift detection, |Q|=20 · CIFAR-10.1 | TPR@5: .52 ± .05 -> .56 ± .05 | Helped |
| Use constrained-disagreement classifiers instead of a standard deep ensemble (ablation) | CIFAR-10.1 shift detection, |Q|=50 · CIFAR-10.1 | TPR@5: .74 ± .04 -> .92 ± .03 | Helped |
| Use constrained-disagreement classifiers instead of MMD-D (ablation) | Camelyon17 shift detection, |Q|=10 · Camelyon17 | TPR@5: .42 ± .05 -> .97 ± .02 | Helped |
| Use constrained-disagreement classifiers instead of MMD-D (ablation) | Camelyon17 shift detection, |Q|=20 · Camelyon17 | TPR@5: .62 ± .05 -> 1.0 ± .00 | Helped |
| Use constrained-disagreement classifiers instead of Relative Mahalanobis (ablation) | Camelyon17 shift detection, |Q|=50 · Camelyon17 | TPR@5: .89 ± .03 -> 1.0 ± .00 | Helped |
| Use constrained-disagreement classifiers instead of CTST (ablation) | UCI Heart Disease shift detection, |Q|=10 · UCI Heart Disease | TPR@5: .15 ± .04 -> .45 ± .05 | Helped |
| Use constrained-disagreement classifiers instead of CTST (ablation) | UCI Heart Disease shift detection, |Q|=20 · UCI Heart Disease | TPR@5: .51 ± .05 -> .88 ± 0.03 | Helped |
| Use constrained-disagreement classifiers instead of CTST (ablation) | UCI Heart Disease shift detection, |Q|=50 · UCI Heart Disease | TPR@5: .98 ± .01 -> 1.0± .00 | Mixed |
| Use entropy rather than disagreement as the Detectron statistic (ablation) | Detectron on UCI Heart Disease · UCI Heart Disease, |Q|=10, 20, and 50 | TPR@5: .24 ± 0.04 -> .45 ± .05; .57 ± 0.05 -> .88 ± 0.03; .82 ± 0.04 -> 1.0± .00 | Helped |
| Use entropy rather than disagreement as the Detectron statistic (ablation) | Detectron on CIFAR-10.1 · CIFAR-10.1, |Q|=10, 20, and 50 | TPR@5: .37 ± .05 -> .35 ± .05; .54 ± .05 -> .56 ± .05; .83 ± .04 -> .92 ± .03 | Mixed |
| Use entropy rather than disagreement as the Detectron statistic (ablation) | Detectron on Camelyon17 · Camelyon17, |Q|=10, 20, and 50 | TPR@5: .97 ± .02 -> .97 ± .02; 1.0 ± .00 -> 1.0 ± .00; .96 ± .02 -> 1.0 ± .00 | No clear effect |
Closest papers by meaning
- Overparameterization Improves Robustness to Covariate Shift in High Dimensions (NeurIPS 2021 · similarity 0.90)
- Fairness under Covariate Shift: Improving Fairness-Accuracy Tradeoff with Few Unlabeled Test Samples (AAAI 2024 · similarity 0.88)
- Beyond the Training Distribution: Evaluating Predictions Under Distribution Shift and Selection Bias (NeurIPS 2026 · similarity 0.86)
- Sequential Harmful Shift Detection Without Labels (NeurIPS 2024 · similarity 0.85)
- Computational Efficiency under Covariate Shift in Kernel Ridge Regression (NeurIPS 2025 · similarity 0.85)
- Not all distributional shifts are equal: Fine-grained robust conformal inference (ICML 2024 · similarity 0.85)
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.