A Comparative Analysis of Contrastive and Generative Vision-Language Models for Zero-Shot Behavior Recognition in Surveillance Videos. ESIT- International Journal. 2026;3(1):23-37. doi:10.66279/vcth9h10