Research map / Label-efficient and robust learning / OOD and anomaly detection
Understanding and Improving Feature Learning for Out-of-Distribution Generalization
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 FeAT-learned features instead of ERM-learned features | GroupDRO on Camelyon17 · Camelyon17 | Avg. acc. (%): 76.09 ± 6.46 -> 80.41 ± 3.30 | Helped |
| Use FeAT-learned features instead of ERM-learned features | GroupDRO on CivilComments · CivilComments | Worst acc. (%): 69.50 ± 0.15 -> 71.29 ± 0.46 | Helped |
| Use FeAT-learned features instead of ERM-learned features | IRMv1 on FMoW · FMoW | Worst acc. (%): 33.45 ± 1.07 -> 34.04 ± 0.70 | Helped |
| Use FeAT-learned features instead of ERM-learned features | IRMv1 on iWildCam · iWildCam | Macro F1: 28.76 ± 0.45 -> 29.66 ± 1.52 | Helped |
| Use FeAT-learned features instead of ERM-learned features | ERM on RxRx1 · RxRx1 | Avg. acc. (%): 30.21 ± 0.09 -> 30.66 ± 0.42 | Helped |
| Use FeAT-learned features instead of RFC-learned features | IRMv1 on ColoredMNIST-01 · ColoredMNIST-01 | OOD accuracy: 76.71 ± 4.10 -> 82.33 ± 1.77 | Helped |
| Use FeAT-learned features instead of RFC-learned features | IRMv1 on ColoredMNIST-025 · ColoredMNIST-025 | OOD accuracy: 70.28 ± 0.72 -> 70.57 ± 0.68 | No clear effect |
| Continue training after FeAT Round-2 (ablation) | FeAT on ColoredMNIST-025 · ColoredMNIST-025 | OOD accuracy: 70.64± 0.62 -> 10.07± 0.26 | Hurt |
| Continue training after FeAT Round-2 (ablation) | FeAT on ColoredMNIST-025 · ColoredMNIST-025 | Retention accuracy: 88.11± 4.28 -> 43.82± 0.59 | Hurt |
Closest papers by meaning
- Towards a Theoretical Framework of Out-of-Distribution Generalization (NeurIPS 2021 · similarity 0.90)
- Towards Robust Out-of-Distribution Generalization Bounds via Sharpness (ICLR 2024 · similarity 0.88)
- Feature Contamination: Neural Networks Learn Uncorrelated Features and Fail to Generalize (ICML 2024 · similarity 0.87)
- Towards Optimal Feature-Shaping Methods for Out-of-Distribution Detection (ICLR 2024 · similarity 0.87)
- Spurious Feature Diversification Improves Out-of-distribution Generalization (ICLR 2024 · similarity 0.87)
- Controlling Neural Collapse Enhances Out-of-Distribution Detection and Transfer Learning (ICML 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.