New Jev-GAD framework uses pretrained models for generalist graph anomaly detection

Researchers have introduced Jev-GAD, a novel framework for training-free generalist graph anomaly detection. This approach addresses challenges posed by heterogeneous anomaly patterns and distributional differences across graphs by constructing complementary structural, attribute, and neighborhood evidence. Jev-GAD transforms these statistics into graph-relative anomaly percentiles and utilizes a single, frozen pretrained language model for evidence-conditioned anomaly assessment. Experiments on seven benchmarks show Jev-GAD achieves strong performance, with average AUROC and AUPRC scores of 89.18% and 52.40%, respectively. AI

IMPACT Introduces a novel approach to graph anomaly detection by leveraging pretrained language models, potentially improving performance on complex graph datasets.

RANK_REASON Academic paper detailing a new method for graph anomaly detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Graph-to-State (G2S)
• Hugging Face
• IArxiv Recommender
• Influence Flower
• Jev-GAD
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →