1 paper
Mingyang Gao, Yuwei Wu, Shiyu Luo +8
Hyperbolic vision-language models are well suited to modeling the hierarchical structure of multimodal data, because hyperbolic volume grows exponentially with radius, enabling low…