Abstract: High-performance Multimodal Large Language Models (MLLMs) rely heavily on data quality. This study introduces a novel data synthesis method, leveraging insights from contrastive learning and ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results