Revisiting Logit Distributions for Reliable Out-of-Distribution Detection

Hong Chang (Institute of Computing Technology, Chinese Academy of Sciences) · RuiBing Hou (Chinese Academy of Sciences) · Shiguang Shan (Chinese Academy of Sciences) · Xilin Chen (Institute of Computing Technology, Chinese Academy of Sciences) · Jiachen Liang (Institute of Computing Technology, Chinese Academy of Sciences) · Minyang Hu (Huawei Technologies Ltd.)

deep learning models empirical evidence in-distribution samplesinformative logitslogitgaplogits spacemaximum logitood samplesout-of-distribution detectionpost-hoc methodsseparability state-of-the-art performance theoretical analysistraining-free strategyvision-language models

Out-of-distribution (OOD) detection is critical for ensuring the reliability of deep learning models in open-world applications. While post-hoc methods are favored for their efficiency and ease of deployment, existing approaches often underexploit the rich information embedded in the model’s logits space. In this paper, we propose LogitGap, a novel post-hoc OOD detection method that explicitly exploits the relationship between the maximum logit and the remaining logits to enhance the separability between in-distribution (ID) and OOD samples. To further improve its effectiveness, we refine LogitGap by focusing on a more compact and informative subset of the logit space. Specifically, we introduce a training-free strategy that automatically identifies the most informative logits for scoring. We provide both theoretical analysis and empirical evidence to validate the effectiveness of our approach. Extensive experiments on both vision-language and vision-only models demonstrate that LogitGap consistently achieves state-of-the-art performance across diverse OOD detection scenarios and benchmarks.