Abstract: Anomaly detection in multivariate time series is crucial for safeguarding the reliability of industrial, financial, and cybersecurity systems. However, conventional deep learning approaches ...