AI RESEARCH
Improving Multimodal Learning with Dispersive and Anchoring Regularization
arXiv CS.LG
•
ArXi:2601.21670v2 Announce Type: replace-cross Multimodal learning aims to integrate complementary information from heterogeneous modalities, yet strong optimization alone does not guaranty well-structured representations. Even under carefully balanced