Query Optimization Techniques in Retrieval-Augmented Generation (RAG) Systems

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Retrieval-Augmented Generation (RAG) helps language models give more accurate answers by using information from external documents. Instead of depending only on what the model learned during training, it can search for relevant information when needed. The quality of the retrieved information is important for getting good results. This paper reviews different ways to improve retrieval, such as rewriting queries, adding related terms, breaking complex questions into smaller parts, generating a sample answer before searching, and refining queries through multiple steps. A simple example is included to show how these methods can improve search results. The paper also compares the techniques and discusses their advantages and disadvantages. Finally, it highlights some common challenges, including slower response times, difficulties in measuring performance, and the possibility of changing the user's original meaning