ModelRefs / HyDE (Hypothetical Document Embeddings) — AI Glossary

HyDE (Hypothetical Document Embeddings) — AI Glossary

A query-expansion technique generating a hypothetical answer document and using its embedding to retrieve real documents.

Overview

HyDE (Gao et al. 2022) prompts the LLM to generate a hypothetical (potentially hallucinated) answer, then embeds that answer rather than the original query. The hypothesis embedding is closer to relevant documents than the short sparse query. Particularly effective for zero-shot dense retrieval where queries are short and retrieval docs are long.

Reference details

Topicrag
Also known ashypothetical document embedding
Last reviewed2026-06-24

Commonly confused with

A query-expansion technique with a counter-intuitive mechanism: it asks the model to write a plausible answer, then embeds that answer to search with. It works because a hypothetical answer sits closer in embedding space to real answers than a question does. The generated document is deliberately discarded and its factual accuracy does not matter — a point that is easy to miss, since it looks like the model is answering.

Primary source

Continue your research

Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to HyDE (Hypothetical Document Embeddings) — AI Glossary.

Frequently asked questions

What is HyDE (Hypothetical Document Embeddings)?

A query-expansion technique generating a hypothetical answer document and using its embedding to retrieve real documents.

Is HyDE (Hypothetical Document Embeddings) the same as hypothetical document embedding?

Yes — hypothetical document embedding are common aliases for HyDE (Hypothetical Document Embeddings).

What concepts are related to HyDE (Hypothetical Document Embeddings)?

Closely related concepts include query expansion, dense retrieval, naive rag.