Latent Space

Latent space is the internal mathematical representation a model uses to encode meaning, relationships, and concepts. Understanding latent space helps explain how models generalise, how embeddings work, and why certain outputs cluster together.