Research map / Label-efficient and robust learning / OOD and anomaly detection
ID and OOD Performance Are Sometimes Inversely Correlated on Real-world Datasets
Damien Teney, LIN Yong, Seong Joon Oh, Ehsan Abbasnejad
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 |
|---|---|---|---|
| train for more epochs (1 to 10) when re-training the last layer (ablation) | DenseNet-121 linear probe (ERM) · WILDS-Camelyon17 | OOD accuracy vs ID accuracy | Mixed |
| add gradient-diversity loss to the linear classifiers (ablation) | DenseNet-121 linear probe · WILDS-Camelyon17 | range of ID/OOD accuracy | Improved |
| keep only models trained for a fixed 10 epochs and pool pretraining seeds (ablation) | ERM models · WILDS-Camelyon17 | visible ID/OOD pattern | Hurt |
| vary the re-training seed (ablation) | ERM linear probe · WILDS-Camelyon17 | ID/OOD accuracy | No clear effect |
| vary the pretraining seed (ablation) | DenseNet-121 · WILDS-Camelyon17 | OOD accuracy | Mixed |
| select epochs by ID validation accuracy instead of OOD | ERM, mixup, class balancing methods · WildTime-arXiv and waterbirds | OOD accuracy | Hurt |
Closest papers by meaning
- Aggregation Hides Out-of-Distribution Generalization Failures from Spurious Correlations (NeurIPS 2025 · similarity 0.91)
- Bridging OOD Detection and Generalization: A Graph-Theoretic View (NeurIPS 2024 · similarity 0.88)
- OOD-Chameleon: Is Algorithm Selection for OOD Generalization Learnable? (ICML 2025 · similarity 0.87)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.87)
- LCA-on-the-Line: Benchmarking Out of Distribution Generalization with Class Taxonomies (ICML 2024 · similarity 0.86)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 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.
