1 paper
Geonmo Gu, Byeongho Heo, Jaemyung Yu +7
Universal Multimodal embedding models built on Multimodal Large Language Models (MLLMs) have traditionally employed contrastive learning, which aligns representations of query-targ…