1 paper · 1 filter
Pedro Domingos, João Pereira, Vasco Lopes +2
Automated video surveillance with Large Vision-Language Models is limited by their inherent bias towards normality, often failing to detect crimes. While Chain-of-Thought reasoning…