A basic RAG system usually relies on vector similarity search to find relevant document chunks for a user query. While this works well for simple cases, real-world queries are often more complex. Users may use different wording, ask multi-part questions, or search for very specific information that simple vector search may fail to retrieve.

To improve retrieval quality, modern systems use a variety of advanced retrieval techniques like hybrid search (combining keyword and vector search), query rewriting, re-ranking models, multi-query retrieval, and context-aware retrieval strategies. Each technique helps the system better understand user intent and surface more relevant information.

These methods are what transform a basic RAG prototype into a high-quality search and question-answering system that users can trust.

In this chapter, we will explore several advanced retrieval techniques and learn how they can significantly improve the accuracy, relevance, and reliability of RAG-based applications.