**Dense vectors** represent different dimensions, with mostly non-zero values, of the data that are essential for understanding patterns, relationships, and underlying structures (i.e., its semantic information). They are more efficient at capturing complex relationships and nuances in language data than [[Sparse Vectors ("Embeddings")]]. They are used in [[Semantic Search]] and are typically stored in a [[Vector Database]]. They are generated by a [[Bi-encoder]]. Dense retrievers employ deep neural networks to identify semantic relationships between queries and documents by encoding them into high-dimensional embeddings and measuring the cosine similarity. Ref. [[Composite Embedding]]