Long captions in language-image pre-training enhance model performance across various downstream tasks.
Long captions in language-image pre-training enhance model performance across various tasks.