Research map / Graph learning / Graph self-supervised learning
Towards Generalist Graph-Level Anomaly Detection
Junjun Pan, Yixin Liu, Yu Zheng, Fuyi Li, Alan Wee-Chung Liew, Shirui Pan
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a generalist approach to graph-level anomaly detection that aims to work across datasets and domains rather than training a separate model per dataset.
Problem
Graph-level anomaly detection (GLAD) typically follows a dataset-specific paradigm, requiring substantial in-domain data and training cost when applied to new application scenarios.
Approach
Shifts GLAD from a dataset-specific paradigm toward a generalist setting that transfers to new scenarios without per-dataset training; the abstract is truncated so the specific mechanism is not stated.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add node-level pseudo-anomaly loss during pre-training (ablation) | GenGLAD · IMDB-BINARY | AUROC: 58.03 -> 74.11 | Improved |
| add L1 sparsity regularization on the subgraph mask (ablation) | GenGLAD · IMDB-BINARY | AUROC: 60.06 -> 74.11 | Improved |
| add class-level discrepancy score (ablation) | GenGLAD · DHFR | AUROC: 45.63 -> 74.96 | Improved |
| add domain discrepancy score (ablation) | GenGLAD · IMDB-BINARY | AUROC: 51.65 -> 74.11 | Improved |
| add prior discrepancy score (ablation) | GenGLAD · IMDB-BINARY | AUROC: 71.15 -> 74.11 | Improved |
| add instance-level discrepancy score (ablation) | GenGLAD · PROTEINS | AUROC: 73.75 -> 73.63 | No clear effect |
| use learned linear scoring module instead of mean/max/min aggregation (ablation) | GenGLAD · DHFR | AUROC: 28.92 -> 74.96 | Improved |
| add cross-domain pre-training before few-shot fine-tuning for conventional GLAD baselines | GLAD baselines · 10 test datasets | AUROC | No clear effect |
| increase number of pre-training datasets (ablation) | GenGLAD · test datasets | AUROC | Improved |
| increase number of few-shot normal graphs for adaptation (ablation) | GenGLAD · social networks | AUROC | Improved |
| increase number of retrieved memory tokens K beyond a threshold (ablation) | GenGLAD · science graphs and social networks | AUROC | Hurt |
Closest papers by meaning
- ARC: A Generalist Graph Anomaly Detector with In-Context Learning (NeurIPS 2024 · similarity 0.91)
- Leveraging Diffusion Model as Pseudo-Anomalous Graph Generator for Graph-Level Anomaly Detection (ICML 2025 · similarity 0.90)
- UniFORM: Towards Unified Framework for Anomaly Detection on Graphs (AAAI 2025 · similarity 0.90)
- Self-Discriminative Modeling for Anomalous Graph Detection (ICML 2025 · similarity 0.90)
- ADA-GAD: Anomaly-Denoised Autoencoders for Graph Anomaly Detection (AAAI 2024 · similarity 0.89)
- UniGAD: Unifying Multi-level Graph Anomaly Detection (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
