{"ID":2853817,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.15202","arxiv_id":"2510.15202","title":"A Geometry-Based View of Mahalanobis OOD Detection","abstract":"Out-of-distribution (OOD) detection is critical for reliable deployment of vision models. Mahalanobis-based detectors remain strong baselines, yet their performance varies widely across modern pretrained representations, and it is unclear which properties of a feature space cause these methods to succeed or fail. We conduct a large-scale study across diverse foundation-model backbones and Mahalanobis variants. First, we show that Mahalanobis-style OOD detection is not universally reliable: performance is highly representation-dependent and can shift substantially with pretraining data and fine-tuning regimes. Second, we link this variability to in-distribution geometry and identify a two-term ID summary that consistently tracks Mahalanobis OOD behavior across detectors: within-class spectral structure and local intrinsic dimensionality. Finally, we treat normalization as a geometric control mechanism and introduce radially scaled $\\ell_2$ normalization, $φ_β(z)=z/\\|z\\|^β$, which preserves directions while contracting or expanding feature radii. Varying $β$ changes the radii while preserving directions, so the same quadratic detector sees a different ID geometry. We choose $β$ from ID-only geometry signals and typically outperform fixed normalization baselines.","short_abstract":"Out-of-distribution (OOD) detection is critical for reliable deployment of vision models. Mahalanobis-based detectors remain strong baselines, yet their performance varies widely across modern pretrained representations, and it is unclear which properties of a feature space cause these methods to succeed or fail. We co...","url_abs":"https://arxiv.org/abs/2510.15202","url_pdf":"https://arxiv.org/pdf/2510.15202v3","authors":"[\"Denis Janiak\",\"Jakub Binkowski\",\"Tomasz Kajdanowicz\"]","published":"2025-10-17T00:04:19Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.CV\"]","methods":"[]","has_code":false}