All terms
Data
Embedding Model
A model that turns text or other inputs into vectors so similar items sit close together.
Definition
An embedding model encodes inputs such as text passages, queries, or images into fixed-length dense vectors, arranged so that semantically similar items have nearby representations. Text embedding models are typically trained by showing them many matching and non-matching pairs so they learn to place related items close together. They are the backbone of semantic search, the document-fetching stage of RAG (a method that looks up relevant documents and feeds them to a model), and clustering and classification pipelines.