Anomaly Detection (AD) aims to identify patterns that deviate from normal behavior. In this talk, we will first present recent advancements in AD, focusing on diverse scenarios, including one-class, image-based, and video-based AD. We will discuss the proposed core methodologies behind each task and showcase the flexibility of our techniques by extending them to medical image analysis. Additionally, this talk highlights our latest work on Vision-Language Models, addressing complex challenges like text-to-image generation and trajectory prediction.