Out-of-Distribution Detection using Counterfactual Distance

2025-08-13

AI Summary

This paper addresses the dual challenges of accuracy and interpretability in out-of-distribution (OOD) detection for machine learning systems. We propose a novel embedding-space distance method grounded in counterfactual explanations. Our core contribution is the first integration of counterfactual reasoning into OOD detection: instead of optimizing in high-dimensional input space, we directly generate counterfactual samples near the decision boundary within the feature embedding space—yielding substantial gains in computational efficiency and scalability. We further introduce a lightweight post-processing strategy to accelerate counterfactual generation. The resulting method achieves both strong interpretability—by providing intuitive, instance-level anomaly attributions—and state-of-the-art performance: AUROC scores of 93.50%, 97.05%, and 92.55% on CIFAR-10, CIFAR-100, and ImageNet-200, respectively—surpassing all existing SOTA approaches.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Accurate and explainable out-of-distribution (OOD) detection is required to use machine learning systems safely. Previous work has shown that feature distance to decision boundaries can be used to identify OOD data effectively. In this paper, we build on this intuition and propose a post-hoc OOD detection method that, given an input, calculates the distance to decision boundaries by leveraging counterfactual explanations. Since computing explanations can be expensive for large architectures, we also propose strategies to improve scalability by computing counterfactuals directly in embedding space. Crucially, as the method employs counterfactual explanations, we can seamlessly use them to help interpret the results of our detector. We show that our method is in line with the state of the art on CIFAR-10, achieving 93.50% AUROC and 25.80% FPR95. Our method outperforms these methods on CIFAR-100 with 97.05% AUROC and 13.79% FPR95 and on ImageNet-200 with 92.55% AUROC and 33.55% FPR95 across four OOD datasets

Problem

• Detecting out-of-distribution data using counterfactual distance
• Improving scalability by computing counterfactuals in embedding space
• Providing explainable OOD detection through counterfactual explanations

Innovation

• Leverages counterfactual explanations for distance calculation
• Computes counterfactuals in embedding space for scalability
• Uses decision boundary distance for OOD detection