Spurious-Aware Prototype Refinement for Reliable Out-of-Distribution Detection

2025-06-30

AI Summary

In real-world scenarios, machine learning models often suffer degraded out-of-distribution (OOD) detection performance due to spurious correlations present in training data. To address this, we propose SPROD—a post-hoc, parameter-free, and data-agnostic robust OOD detection framework. Our method identifies latent spurious features via feature disentanglement and adaptively refines class prototypes in the prototype space to mitigate bias induced by spurious correlations. SPROD is architecture- and paradigm-agnostic, seamlessly integrating with diverse backbone networks and OOD detection strategies. Extensive evaluations on benchmark datasets—including CelebA, Waterbirds, and Spurious ImageNet—demonstrate that SPROD consistently outperforms state-of-the-art methods, achieving an average AUROC improvement of 4.7% and a 9.3% reduction in FPR@95. By eliminating reliance on auxiliary data or hyperparameter tuning, SPROD provides an efficient, generalizable, and interpretable solution for enhancing OOD detection robustness.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Out-of-distribution (OOD) detection is crucial for ensuring the reliability and safety of machine learning models in real-world applications, where they frequently face data distributions unseen during training. Despite progress, existing methods are often vulnerable to spurious correlations that mislead models and compromise robustness. To address this, we propose SPROD, a novel prototype-based OOD detection approach that explicitly addresses the challenge posed by unknown spurious correlations. Our post-hoc method refines class prototypes to mitigate bias from spurious features without additional data or hyperparameter tuning, and is broadly applicable across diverse backbones and OOD detection settings. We conduct a comprehensive spurious correlation OOD detection benchmarking, comparing our method against existing approaches and demonstrating its superior performance across challenging OOD datasets, such as CelebA, Waterbirds, UrbanCars, Spurious Imagenet, and the newly introduced Animals MetaCoCo. On average, SPROD improves AUROC by 4.7% and FPR@95 by 9.3% over the second best.

Problem

• Addressing spurious correlations in OOD detection
• Refining prototypes to mitigate feature bias
• Improving robustness across diverse OOD datasets

Innovation

• Refines class prototypes to mitigate spurious bias
• Works without extra data or hyperparameter tuning
• Applicable across diverse backbones and settings