RAG & Retrieval · Ground answers in evidence
What is HyDE and when does it help?
Short answer
HyDE asks a model to draft a hypothetical answer, then embeds that answer for retrieval. It helps when questions and documents use different language, but adds latency and can retrieve plausible fiction.
Hypothetical Document Embeddings generate a fake answer and search with its embedding. The intuition is that an answer looks more like the source documents than a question does, closing the vocabulary gap. It is a query-expansion technique with real failure modes.
- Useful when users ask short questions against long explanatory documents.
- Costs an extra generation call before retrieval, increasing latency.
- A wrong hypothetical answer can pull retrieval toward the wrong topic.
- Combine with the original query rather than replacing it outright.
Follow-ups you should expect
Interviewers use these to check whether the first answer was understanding or memorisation.
- 1What happens if the hypothetical answer is confidently wrong?
- 2Would you use HyDE for exact identifier searches?
- 3How would you combine HyDE with hybrid retrieval?
Say it in the interview
HyDE generates a hypothetical answer and embeds that instead of the raw question, because answers often look more like the source documents. It can help when the vocabulary gap is large, but it adds a generation call and a failure mode: a wrong draft steers retrieval to the wrong topic. I would run it alongside the original query and only keep it if it improves recall on my own eval set.
Use this as a model, not a script. Saying it in your own words is what makes it survive follow-up pressure.
Related questions
Reading is step one
Now say it out loud.
Understanding an answer and delivering it under pressure are different skills. Rehearse this topic in a live RAG & Retrieval mock interview and get scored feedback on depth, clarity and confidence.