VADER: Towards Causal Video Anomaly Understanding with Relation-Aware Large Language Models - Explained Simply | ArXiv Explained