Incorporating Dense Knowledge Alignment into Unified Multimodal Representation Models | Read Paper on Bytez