Modular Prompt Learning Improves Vision-Language ModelsZhenhan HuangTejaswini Pedapatiet al.2025ICASSP 2025