Likelihood-Aware Semantic Alignment for Full-Spectrum Out-of-Distribution Detection

Abstract

Full-spectrum out-of-distribution (F-OOD) detection aims to accurately recognize in-distribution (ID) samples while encountering semantic and covariate shifts simultaneously. However, existing out-of-distribution (OOD) detectors tend to overfit the covariance information and ignore intrinsic semantic correlation, inadequate for adapting to complex domain transformations. To address this issue, we propose a Likelihood-Aware Semantic Alignment (LSA) framework to promote the image-text alignment into semantically high-likelihood regions. LSA consists of an offline Gaussian sampling strategy which efficiently samples semantic-relevant visual embeddings from the class-conditional Gaussian distribution, and a bidirectional prompt customization mechanism that adjusts ID-related and negative context for a discriminative ID/OOD boundary. Extensive experiments demonstrate the remarkable OOD detection performance of our proposed LSA, especially on the intractable Near-OOD setting, surpassing existing methods by a margin of 15.26% and 18.88% on two F-OOD benchmarks, respectively.