8.8 Advanced RAG
You can fix a RAG system that returns plausible nonsense.
Before:07. Natural Language ProcessingUnlocks:09. Agentic AI12. Frontier Topics
Advanced RAG is the repair kit for plausible nonsense: hybrid sparse-plus-dense search, query rewriting, HyDE, graph structures, contextual retrieval — each fixing a specific failure, none fixing RAG in general. It follows the base RAG topic as its debugging course. The method matters more than the menu: diagnose whether queries, chunks or ranking are failing first, because stacking techniques without a diagnosis adds cost and complexity to a system nobody understands.
Work through these
Hybrid search: BM25 plus dense
Combining classical keyword scoring with vector similarity catches both exact terms and paraphrases. It is usually the first improvement worth making.
Query rewriting, HyDE, multi-query
Rewriting the user's question, generating a hypothetical answer to search with, or issuing several queries at once. Each addresses questions that are poorly phrased for retrieval.
Graph RAG and structured knowledge
Retrieving over a structured graph of entities rather than loose text, which suits questions that require joining facts. It costs more to build and answers questions plain retrieval cannot.
Contextual retrieval and late chunking
Attaching surrounding context to each piece before it is stored, and deferring the cutting until after understanding. Both address the same weakness in naive chunking.
Sign in to keep your progress.
Free resources
We haven't checked most of these for screen reader use yet.
Links last checked 29 Aug 2026.
Stuck here?
Ask a mentor. A real person answers, and they can see exactly which topic you're on. Usually within a couple of working days.
Checking your session…
Topics shown in module order.