Breaking Semantic Hegemony: Decoupling Principal and Residual Subspaces for Generalized OOD Detection

TL;DR

This paper identifies a bias in deep feature spaces that hampers detection of structurally distinct OOD samples and proposes a geometric decoupling method, D-KNN, to improve sensitivity and establish new state-of-the-art results.

Contribution

It introduces D-KNN, a training-free geometric decoupling framework that separates semantic and residual features to enhance OOD detection performance.

Findings

Reduces FPR95 from 31.3% to 2.3% on CIFAR benchmarks.

Improves AUROC from 79.7% to 94.9% under sensor noise.

Establishes new state-of-the-art results in OOD detection.

Abstract

While feature-based post-hoc methods have made significant strides in Out-of-Distribution (OOD) detection, we uncover a counter-intuitive Simplicity Paradox in existing state-of-the-art (SOTA) models: these models exhibit keen sensitivity in distinguishing semantically subtle OOD samples but suffer from severe Geometric Blindness when confronting structurally distinct yet semantically simple samples or high-frequency sensor noise. We attribute this phenomenon to Semantic Hegemony within the deep feature space and reveal its mathematical essence through the lens of Neural Collapse. Theoretical analysis demonstrates that the spectral concentration bias, induced by the high variance of the principal subspace, numerically masks the structural distribution shift signals that should be significant in the residual subspace. To address this issue, we propose D-KNN, a training-free, plug-and-play geometric decoupling framework. This method utilizes orthogonal decomposition to explicitly separate semantic components from structural residuals and introduces a dual-space calibration mechanism to reactivate the model's sensitivity to weak residual signals. Extensive experiments demonstrate that D-KNN effectively breaks Semantic Hegemony, establishing new SOTA performance on both CIFAR and ImageNet benchmarks. Notably, in resolving the Simplicity Paradox, it reduces the FPR95 from 31.3% to 2.3%; when addressing sensor failures such as Gaussian noise, it boosts the detection performance (AUROC) from a baseline of 79.7% to 94.9%.