Bridging the Digital Divide: Performance Variation across Socio-Economic Factors in Vision-Language Models