MLLMs-Augmented Visual-Language Representation Learning