Research map / Label-efficient and robust learning / OOD and anomaly detection
Pareto Invariant Risk Minimization: Towards Mitigating the Optimization Dilemma in 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 the preference-aware PAIR optimizer with IRMx objectives instead of linear weighting | IRMx · WILDS | Avg. Rank (↓): 4.00 -> 2.17 | Helped |
| Use the preference-aware PAIR optimizer with IRMx objectives instead of linear weighting | IRMx · CMNIST-m | accuracy: 81.6 ± 2.0 -> 83.7 ± 1.2 | Helped |
| Use the preference-aware PAIR optimizer with IRMx objectives instead of linear weighting | IRMx · Camelyon17 | Avg. acc. (%): 67.0 ± 6.6 -> 74.0 ± 7.0 | Mixed |
| Use the preference-aware PAIR optimizer with IRMx objectives instead of linear weighting | IRMx · CivilComments | Worst acc. (%): 74.3 ± 0.8 -> 75.2 ± 0.7 | Helped |
| Use classifier-only gradients to determine objective weights rather than full-model gradients (ablation) | PAIR optimizer · CMNIST | accuracy: 68.6 ± 0.9 -> 69.2 ± 0.7 | No clear effect |
| Use classifier-only gradients to determine objective weights rather than featurizer gradients (ablation) | PAIR optimizer · CMNIST-m | accuracy: 83.7 ± 1.2 -> 83.7 ± 1.2 | No clear effect |
| Select models by jointly trading off ERM and OOD performance (ablation) | DANN · CMNIST | Δ wr.: +0.9 | Helped |
| Select models by jointly trading off ERM and OOD performance (ablation) | GroupDRO · CMNIST | Δ wr.: +3.1 | Helped |
| Select models by jointly trading off ERM and OOD performance (ablation) | IRMv1 · CMNIST | Δ wr.: +14.3 | Helped |
| Select models by jointly trading off ERM and OOD performance (ablation) | Fishr · CMNIST | Δ wr.: +12.7 | Helped |
| Select models by jointly trading off ERM and OOD performance (ablation) | GroupDRO · PACS | Δ wr.: +1.8 | Helped |
| Select models by jointly trading off ERM and OOD performance (ablation) | GroupDRO · Terra | Δ wr.: +0.0 | No clear effect |
Closest papers by meaning
- Regularization Penalty Optimization for Addressing Data Quality Variance in OoD Algorithms (AAAI 2022 · similarity 0.90)
- Out-of-distribution Generalization for Total Variation based Invariant Risk Minimization (ICLR 2025 · similarity 0.89)
- Towards Robust Out-of-Distribution Generalization Bounds via Sharpness (ICLR 2024 · similarity 0.87)
- Structure-informed Risk Minimization for Robust Ensemble Learning (ICML 2025 · similarity 0.87)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.86)
- Double Descent Meets Out-of-Distribution Detection: Theoretical Insights and Empirical Analysis on the Role of Model Complexity (NeurIPS 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.