FiVL: A Framework for Improved Vision-Language Alignment