The author revisits node-level graph anomaly detection methods, focusing on outlier injection techniques, message passing impact, and hyperbolic neural networks to enhance performance.
