Environment-Conditioned Tail Reweighting for Total Variation Invariant Risk Minimization
This work addresses out-of-distribution generalization for machine learning models, particularly under mixed distribution shifts, representing an incremental advancement over existing invariant risk minimization methods.
The paper tackles the challenge of out-of-distribution generalization under mixed correlation and diversity shifts by proposing ECTR, a framework that combines environment-level invariance with sample-level robustness, resulting in consistent improvements in worst-environment and average OOD performance across multiple benchmarks.
Out-of-distribution (OOD) generalization remains challenging when models simultaneously encounter correlation shifts across environments and diversity shifts driven by rare or hard samples. Existing invariant risk minimization (IRM) methods primarily address spurious correlations at the environment level, but often overlook sample-level heterogeneity within environments, which can critically impact OOD performance. In this work, we propose Environment-Conditioned Tail Reweighting for Total Variation Invariant Risk Minimization (ECTR), a unified framework that augments TV-based invariant learning with environment-conditioned tail reweighting to jointly address both types of distribution shift. By integrating environment-level invariance with within-environment robustness, the proposed approach makes these two mechanisms complementary under mixed distribution shifts. We further extend the framework to scenarios without explicit environment annotations by inferring latent environments through a minimax formulation. Experiments across regression, tabular, time-series, and image classification benchmarks under mixed distribution shifts demonstrate consistent improvements in both worst-environment and average OOD performance.