What is query rewriting in RAG?

Query rewriting in RAG (Retrieval-Augmented Generation) is the process of transforming the user's original query into a more precise or contextually enriched form before sending it to the relevant information retrieval stage. The goal is to improve the quality of the retrieved data, which is then used by the generative model to formulate a response.

For example, if a user asks a short or ambiguous question, query rewriting can add context, clarify terminology, or rephrase the query so that the search system better understands the intent and provides more relevant documents.

This is especially important in RAG, where the quality of the retrieved information directly impacts the quality of the final response generated by the model.