VMF-GOS: Geometry-guided virtual Outlier Synthesis for Long-Tailed OOD Detection

2026-02-05

AI Summary

This work addresses the challenge of ambiguous decision boundaries in out-of-distribution (OOD) detection under long-tailed data distributions, where tail classes suffer from severe sample scarcity. To circumvent the reliance on external real anomalous data, the authors propose a virtual anomaly synthesis method that operates on the feature hypersphere. By modeling class-conditional von Mises–Fisher distributions, the approach identifies low-likelihood annular regions and performs directional sampling to generate synthetic anomalies. These synthetic outliers are integrated into a contrastive learning framework via a dual-granularity semantic loss, enhancing the separability between in-distribution samples and synthesized anomalies. The resulting geometry-guided synthesis strategy eliminates the need for external anomaly data entirely and achieves state-of-the-art OOD detection performance on benchmarks such as CIFAR-LT, surpassing existing methods that depend on real external anomalies, while offering improved efficiency and privacy preservation.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Other Foundations of Computer Vision

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Out-of-Distribution (OOD) detection under long-tailed distributions is a highly challenging task because the scarcity of samples in tail classes leads to blurred decision boundaries in the feature space. Current state-of-the-art (sota) methods typically employ Outlier Exposure (OE) strategies, relying on large-scale real external datasets (such as 80 Million Tiny Images) to regularize the feature space. However, this dependence on external data often becomes infeasible in practical deployment due to high data acquisition costs and privacy sensitivity. To this end, we propose a novel data-free framework aimed at completely eliminating reliance on external datasets while maintaining superior detection performance. We introduce a Geometry-guided virtual Outlier Synthesis (GOS) strategy that models statistical properties using the von Mises-Fisher (vMF) distribution on a hypersphere. Specifically, we locate a low-likelihood annulus in the feature space and perform directional sampling of virtual outliers in this region. Simultaneously, we introduce a new Dual-Granularity Semantic Loss (DGS) that utilizes contrastive learning to maximize the distinction between in-distribution (ID) features and these synthesized boundary outliers. Extensive experiments on benchmarks such as CIFAR-LT demonstrate that our method outperforms sota approaches that utilize external real images.

Problem

• Out-of-Distribution detection
• long-tailed distribution
• Outlier Exposure
• data-free
• feature space

Innovation

• Geometry-guided virtual Outlier Synthesis
• von Mises-Fisher distribution
• Dual-Granularity Semantic Loss
• data-free OOD detection
• long-tailed distribution