Research map / Label-efficient and robust learning / Domain adaptation and generalization
On Calibration and Out-of-Domain Generalization
Yoav Wald, Amir Feder, Daniel Greenfeld, Uri Shalit
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 |
|---|---|---|---|
| Fine-tune top layers with CLOvE | ERM · FMoW OOD test | worst unseen region accuracy: 32.63 -> 44.16 | Improved |
| Use robust isotonic calibration | ERM · FMoW OOD test | worst unseen region accuracy: 32.63 -> 37.19 | Improved |
| Fine-tune top layers with CLOvE | DeepCORAL · Camelyon17 OOD test | accuracy: 72.44 -> 79.96 | Improved |
| Use robust isotonic calibration | DeepCORAL · Camelyon17 OOD test | accuracy: 72.44 -> 76.8 | Improved |
| Use naive isotonic calibration | ERM · PovertyMap OOD test | Pearson correlation r: 0.832 -> 0.827 | Hurt |
| Use robust isotonic calibration | IRM · PovertyMap OOD test | Pearson correlation r: 0.735 -> 0.815 | Improved |
| Calibrate using CLOvE | ERM · CivilComments test | average group accuracy: 63.65 -> 80.39 | Improved |
| Calibrate using CLOvE | IRM · CivilComments test | average group accuracy: 40.61 -> 68.45 | Mixed |
| Select models by average ECE across training environments | Colored MNIST models · Colored MNIST OOD evaluation | Pearson correlation with OOD accuracy: -0.92 | Improved |
| Use the IRMv1 penalty rather than average ECE for model selection | Colored MNIST models · Colored MNIST OOD evaluation | Pearson correlation with OOD accuracy: -0.59 | Mixed |
| Optimize multi-domain calibration with CLOvE instead of IRMv1 | Two-Bit environments · Two-Bit OOD test | invariant classifier recovery | Improved |
Closest papers by meaning
- In Search of Forgotten Domain Generalization (ICLR 2025 · similarity 0.88)
- Lost Domain Generalization Is a Natural Consequence of Lack of Training Domains (AAAI 2024 · similarity 0.88)
- Reassessing How to Compare and Improve the Calibration of Machine Learning Models (ICLR 2025 · similarity 0.88)
- Towards Robust Out-of-Distribution Generalization Bounds via Sharpness (ICLR 2024 · similarity 0.88)
- CRoFT: Robust Fine-Tuning with Concurrent Optimization for OOD Generalization and Open-Set OOD Detection (ICML 2024 · similarity 0.87)
- Rethinking the Evaluation Protocol of Domain Generalization (CVPR 2024 · similarity 0.87)
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.
