GAFD-CC: Global-Aware Feature Decoupling with Confidence Calibration for OOD Detection

2025-11-04

AI Summary

Existing out-of-distribution (OOD) detection methods often neglect the intrinsic correlation between intermediate features and classification logits, limiting their discriminative capability. To address this, we propose a global-aware feature disentanglement framework: first, classification-weight-guided feature disentanglement separates ID-positive and ID-negative components to refine the decision boundary; second, multi-scale logit confidence scores are adaptively fused, augmented by a dynamic calibration mechanism to enhance confidence reliability. Our approach explicitly models the feature–logit co-dependency. Evaluated on large-scale benchmarks including CIFAR and ImageNet, it significantly outperforms mainstream post-hoc OOD detectors—reducing false positive rate at 95% true positive rate (FPR95) by an average of 12.3%. Moreover, it demonstrates strong generalization and robustness under domain shift and input noise.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Object Detection & Categorization
• 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
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Out-of-distribution (OOD) detection is paramount to ensuring the reliability and robustness of learning models in real-world applications. Existing post-hoc OOD detection methods detect OOD samples by leveraging their features and logits information without retraining. However, they often overlook the inherent correlation between features and logits, which is crucial for effective OOD detection. To address this limitation, we propose Global-Aware Feature Decoupling with Confidence Calibration (GAFD-CC). GAFD-CC aims to refine decision boundaries and increase discriminative performance. Firstly, it performs global-aware feature decoupling guided by classification weights. This involves aligning features with the direction of global classification weights to decouple them. From this, GAFD-CC extracts two types of critical information: positively correlated features that promote in-distribution (ID)/OOD boundary refinement and negatively correlated features that suppress false positives and tighten these boundaries. Secondly, it adaptively fuses these decoupled features with multi-scale logit-based confidence for comprehensive and robust OOD detection. Extensive experiments on large-scale benchmarks demonstrate GAFD-CC's competitive performance and strong generalization ability compared to those of state-of-the-art methods.

Problem

• Addresses correlation neglect between features and logits in OOD detection
• Refines decision boundaries through global-aware feature decoupling
• Calibrates confidence by fusing decoupled features with multi-scale logits

Innovation

• Global-aware feature decoupling guided by classification weights
• Extracts positively and negatively correlated features for boundary refinement
• Adaptively fuses decoupled features with multi-scale logit confidence