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. https://doi.org/10.66279/vcth9h10.