CORE Framework Enables Unified Tabular Anomaly Detection via In-Context Reconstruction

CORE introduces an in-context reconstruction approach for unified tabular anomaly detection to overcome semantic loss in distance-based feature alignment. It utilizes a decorrelated feature alignment module to detect diverse anomaly patterns across heterogeneous datasets without relying on binary classification.