GLASS: Graph-Language Alignment with Spherical Scoring for Transferable Graph-Level Anomaly Detection
GLASS uses graph-language alignment to make graph-level anomaly detection transfer across domains with little or no target data.
The paper aligns a structure-aware graph encoder with instruction-aware text embeddings on a unit hypersphere. It turns graph properties into a compact Graph Descriptor Prompt, then scores anomalies with density estimation in both graph and text embedding spaces. The authors report the best average AUROC and rank across twelve benchmarks and three meta-domains against recent GLAD baselines. ArXiv · AI/CL/LG's note
The paper aligns a structure-aware graph encoder with instruction-aware text embeddings on a unit hypersphere. It turns graph properties into a compact Graph Descriptor Prompt, then scores anomalies with density estimation in both graph and text embedding spaces. The authors report the best average AUROC and rank across twelve benchmarks and three meta-domains against recent GLAD baselines. ArXiv · AI/CL/LG's note
score 4