Tracking the Discriminative Axis: Dual Prototypes for Test-Time OOD Detection Under Covariate Shift

2026-03-16

AI Summary

This work addresses the challenge of out-of-distribution (OOD) detection under covariate shift, where the test distribution dynamically changes and in-distribution (ID) and OOD samples are intermixed. Existing methods suffer significant performance degradation due to their reliance on a static ID distribution assumption. To overcome this limitation, we propose DART, which exploits the newly identified property that ID and OOD samples remain separable along a fixed discriminative axis even under covariate shift. DART introduces a training-free, fully online dual-prototype tracking mechanism that dynamically maintains ID and OOD prototypes, enhanced by multi-layer feature fusion and flip correction for improved stability. Evaluated on strong shift benchmarks such as ImageNet-C versus Textures-C, DART achieves a 15.32 percentage point improvement in AUROC and reduces FPR@95TPR by 49.15 percentage points, substantially outperforming current state-of-the-art methods.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

For reliable deployment of deep-learning systems, out-of-distribution (OOD) detection is indispensable. In the real world, where test-time inputs often arrive as streaming mixtures of in-distribution (ID) and OOD samples under evolving covariate shifts, OOD samples are domain-constrained and bounded by the environment, and both ID and OOD are jointly affected by the same covariate factors. Existing methods typically assume a stationary ID distribution, but this assumption breaks down in such settings, leading to severe performance degradation. We empirically discover that, even under covariate shift, covariate-shifted ID (csID) and OOD (csOOD) samples remain separable along a discriminative axis in feature space. Building on this observation, we propose DART, a test-time, online OOD detection method that dynamically tracks dual prototypes -- one for ID and the other for OOD -- to recover the drifting discriminative axis, augmented with multi-layer fusion and flip correction for robustness. Extensive experiments on a wide range of challenging benchmarks, where all datasets are subjected to 15 common corruption types at severity level 5, demonstrate that our method significantly improves performance, yielding 15.32 percentage points (pp) AUROC gain and 49.15 pp FPR@95TPR reduction on ImageNet-C vs. Textures-C compared to established baselines. These results highlight the potential of the test-time discriminative axis tracking for dependable OOD detection in dynamically changing environments.

Problem

• out-of-distribution detection
• covariate shift
• test-time adaptation
• dynamic environments
• distribution shift

Innovation

• test-time OOD detection
• covariate shift
• dual prototypes
• discriminative axis
• online adaptation