![[2602.18793] From Few-Shot to Zero-Shot: Towards Generalist Graph Anomaly Detection](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18793] From Few-Shot to Zero-Shot: Towards Generalist Graph Anomaly Detection
Summary
This paper presents a novel approach to graph anomaly detection (GAD) that transitions from few-shot to zero-shot learning, enabling effective anomaly detection across multiple unseen datasets without extensive retraining.
Why It Matters
Graph anomaly detection is crucial in various fields, including cybersecurity and social networks. Traditional methods require extensive dataset-specific training, which is costly and limits generalization. This research proposes a unified model that enhances efficiency and adaptability, addressing significant challenges in the field.
Key Takeaways
- Introduces a generalist GAD paradigm for anomaly detection.
- Presents ARC and ARC_zero methods for few-shot and zero-shot learning.
- Demonstrates strong generalization ability across 17 real-world datasets.
- Reduces computational costs and enhances transferability of models.
- Addresses privacy concerns by minimizing the need for extensive labeled data.
Computer Science > Machine Learning arXiv:2602.18793 (cs) [Submitted on 21 Feb 2026] Title:From Few-Shot to Zero-Shot: Towards Generalist Graph Anomaly Detection Authors:Yixin Liu, Shiyuan Li, Yu Zheng, Qingfeng Chen, Chengqi Zhang, Philip S. Yu, Shirui Pan View a PDF of the paper titled From Few-Shot to Zero-Shot: Towards Generalist Graph Anomaly Detection, by Yixin Liu and 6 other authors View PDF HTML (experimental) Abstract:Graph anomaly detection (GAD) is critical for identifying abnormal nodes in graph-structured data from diverse domains, including cybersecurity and social networks. The existing GAD methods often focus on the learning paradigms of "one-model-for-one-dataset", requiring dataset-specific training for each dataset to achieve optimal performance. However, this paradigm suffers from significant limitations, such as high computational and data costs, limited generalization and transferability to new datasets, and challenges in privacy-sensitive scenarios where access to full datasets or sufficient labels is restricted. To address these limitations, we propose a novel generalist GAD paradigm that aims to develop a unified model capable of detecting anomalies on multiple unseen datasets without extensive retraining/fine-tuning or dataset-specific customization. To this end, we propose ARC, a few-shot generalist GAD method that leverages in-context learning and requires only a few labeled normal samples at inference time. Specifically, ARC consists of three ...