Representation Potentials of Foundation Models for Multimodal Alignment: A Survey

Jianglin Lu, Hailing Wang, Yi Xu +3

Foundation models learn highly transferable representations through large-scale pretraining on diverse data. An increasing body of research indicates that these representations exh…