Environment-Conditioned Tail Reweighting for Total Variation Invariant Risk Minimization

TL;DR

This paper introduces ECTR, a novel framework that enhances invariant risk minimization by combining environment-conditioned tail reweighting with TV-based learning, improving out-of-distribution generalization across diverse tasks.

Contribution

It proposes a unified approach that jointly addresses environment-level and sample-level distribution shifts, including scenarios without explicit environment labels.

Findings

Consistent improvements in worst-environment OOD performance.

Enhanced average OOD generalization across multiple benchmarks.

Effective handling of mixed distribution shifts in various data modalities.

Abstract

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.