New GLASS framework aligns graph and language for anomaly detection

Researchers have developed GLASS, a novel framework for graph-level anomaly detection that leverages graph-language alignment on a hypersphere to achieve robust cross-domain transferability. The system creates a unified representation space by connecting a graph encoder with a text embedding, using a multi-slice soft cosine objective. This approach serializes graph properties into a "Graph Descriptor Prompt" (GraphDP), acting as a text bridge for domain-agnostic anomaly scoring. GLASS employs Matryoshka representations for multi-scale consistency and Spherical Multi-Modal Scoring (SMS) using von Mises-Fisher kernel density estimators for a principled fusion of structural and semantic anomaly signals. The framework demonstrates effective zero-shot and few-shot anomaly detection across twelve benchmarks, outperforming existing methods. AI

IMPACT This research introduces a novel approach to anomaly detection by aligning graph and language representations, potentially improving cross-domain transferability in AI systems.

RANK_REASON The item is an academic paper detailing a new framework and methodology for graph-level anomaly detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Graph Descriptor Prompt
• Hugging Face
• IArxiv Recommender
• Influence Flower
• Matryoshka
• ScienceCast
• Spherical Multi-Modal Scoring

• paper
• other

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