Latent space is a learned representation space in which a model organizes data according to patterns that may not be directly visible in the original input.
Latent spaces are used in embeddings, generative models, clustering, compression, and similarity search.