{"ID":2859278,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.05782","arxiv_id":"2510.05782","title":"Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection","abstract":"Out-of-distribution (OOD) detection is essential for reliably deploying machine learning models in the wild. Yet, most methods treat large pre-trained models as monolithic encoders and rely solely on their final-layer representations for detection. We challenge this wisdom. We reveal the \\textit{intermediate layers} of pre-trained models, shaped by residual connections that subtly transform input projections, \\textit{can} encode \\textit{surprisingly rich and diverse signals} for detecting distributional shifts. Importantly, to exploit latent representation diversity across layers, we introduce an entropy-based criterion to \\textit{automatically} identify layers offering the most complementary information in a training-free setting -- \\textit{without access to OOD data}. We show that selectively incorporating these intermediate representations can increase the accuracy of OOD detection by up to \\textbf{$10\\%$} in far-OOD and over \\textbf{$7\\%$} in near-OOD benchmarks compared to state-of-the-art training-free methods across various model architectures and training objectives. Our findings reveal a new avenue for OOD detection research and uncover the impact of various training objectives and model architectures on confidence-based OOD detection methods.","short_abstract":"Out-of-distribution (OOD) detection is essential for reliably deploying machine learning models in the wild. Yet, most methods treat large pre-trained models as monolithic encoders and rely solely on their final-layer representations for detection. We challenge this wisdom. We reveal the \\textit{intermediate layers} of...","url_abs":"https://arxiv.org/abs/2510.05782","url_pdf":"https://arxiv.org/pdf/2510.05782v2","authors":"[\"I. M. De la Jara\",\"C. Rodriguez-Opazo\",\"D. Teney\",\"D. Ranasinghe\",\"E. Abbasnejad\"]","published":"2025-10-07T10:55:47Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[]","has_code":false}