Jul 2026
The Hyperspherical Geometry of CLIP Latent Space: A Semantic Mixture Model
A principled density model for the CLIP latent space based on Mixtures of von Mises-Fisher distributions defined on the unit hypersphere is proposed, establishing a simple and geometrically consistent probabilistic framework for modeling and understanding multimodal representations.
Zijie Yu, Gaowen Liu, R. Kompella et al.
· arXiv.org · 0 citations