Leveraging Large Language Models and Vision-Language Models for Training-free Video Anomaly Detection

A novel training-free method for video anomaly detection that leverages pre-trained large language models and vision-language models to detect anomalies without any task-specific training or data collection.