Home / Retrieval models
Retrieval models guides
11 published guides. Start with Retrieval models explained.
The model families that retrieve documents, how they differ and how to choose between them.
Read the guideLearning path
All guides
Understand it3 guides
Start hereRetrieval models explainedThe model families that retrieve documents, how they differ and how to choose between them.ExplainerBi-encodersHow bi-encoders embed queries and documents independently, which is what makes semantic search fast.ExplainerCross-encodersHow cross-encoders read query and document together for precise scores, and why they are too slow for first-stage search.
How it works6 guides
ExplainerDense retrievalHow dense retrievers encode queries and passages separately and match them by vector similarity.ExplainerSparse retrievalSparse retrieval matches on weighted terms. How classic and learned sparse models differ.ExplainerSPLADE and learned sparse retrievalHow SPLADE expands documents with learned terms to get semantic recall from an inverted index.ExplainerColBERT and late interactionHow late interaction keeps one vector per token for near cross-encoder accuracy at retrieval speed.ExplainerMulti-vector searchWhen one vector per document is not enough and how multi-vector search stores and queries several.ExplainerAgentic retrieval and agentic searchHow agentic search plans queries, picks tools and iterates, and what it costs compared with one-shot retrieval.