Generalist Graph Anomaly Detection via Prototype-Based Distillation

2026-05-26

AI Summary

This work proposes ProMoS, the first unsupervised, zero-shot framework for general-purpose graph anomaly detection. Existing methods often rely on labeled data or few-shot samples during inference, limiting their ability to generalize to unseen anomaly patterns. ProMoS addresses this by distilling normality priors from a frozen self-supervised GNN teacher model via knowledge distillation. It employs a hybrid student architecture featuring a shared global branch and lightweight personalized branches, along with a prototype-guided soft-label distillation mechanism to enable cross-graph generalization. Requiring no annotations, ProMoS detects anomalies in previously unseen graphs in a zero-shot manner, significantly enhancing out-of-distribution generalization. Extensive experiments demonstrate that ProMoS achieves both high efficiency and superior performance across multiple benchmarks, establishing a practical new paradigm for label-free, general graph anomaly detection.

Abstract

Driven by the pressing demand for graph anomaly detection (GAD) in high-stakes domains, the generalist GAD paradigm, which trains a single detector transferable across new graphs, has recently gained growing attention. However, existing methods often rely on scarce and costly annotations for training and sometimes even require few-shot support at inference, which limits their robustness to diverse and unseen anomaly patterns. To address this limitation, we introduce ProMoS, the first unsupervised generalist GAD framework, which detects anomalies by modeling the abundant normality in unlabeled data. ProMoS adopts a knowledge-distillation paradigm to distill normality priors from a frozen self-supervised graph neural network (GNN) teacher to a mixture-of-students model with shared global and lightweight personalized branches, enabling efficient and expressive normality modeling without learning from scratch. We further propose prototype-guided soft-label distillation to align teacher and student in a shared prototype space, enhancing cross-graph generalizability. During inference, ProMoS performs zero-shot anomaly detection on unseen graphs via distillation bias and prototype geometric deviation. Extensive experiments show the effectiveness and efficiency of ProMoS, charting a practical path toward label-free, zero-shot generalist GAD.

Problem

• graph anomaly detection
• generalist GAD
• unsupervised learning
• zero-shot detection
• cross-graph generalization

Innovation

• unsupervised generalist GAD
• prototype-based distillation
• zero-shot anomaly detection
• graph neural networks
• normality modeling