RAG (Retrieval-Augmented Generation)

Question 1

What are the three core steps of RAG?

Accepted Answer

Retrieve, augment, generate — RAG retrieves relevant chunks, augments the prompt with them as context, and generates an answer grounded in that context.

Question 2

RAG stands for…

Accepted Answer

Retrieval-Augmented Generation — RAG retrieves relevant external documents and provides them as context so the LLM generates grounded answers.

Question 3

The core idea of RAG is to…

Accepted Answer

Retrieve info, add it to the prompt — It augments the model's fixed knowledge with retrieved, query-relevant context — without changing the model's weights.

Question 4

RAG primarily helps with which problem?

Accepted Answer

Accessing facts not in model training — LLM weights are static and lack private/recent data; RAG injects that knowledge at query time so answers stay accurate and current.

Question 5

RAG reduces HALLUCINATION by…

Accepted Answer

Grounding answers in retrieved sources — Supplying authoritative context lets the model cite/base answers on real data, curbing fabricated content.

Question 6

The two main components of a RAG system are the…

Accepted Answer

Retriever and generator — A retriever fetches context from a knowledge store; the generator (LLM) synthesizes an answer using that context.

Question 7

Retrieval in RAG is typically done via…

Accepted Answer

Embedding query and matching chunks — The query is embedded and matched against a vector index of document chunks to retrieve the most relevant context.

Question 8

Documents are split into CHUNKS before indexing because…

Accepted Answer

Smaller pieces retrieve more precisely — Chunking yields granular, relevant retrieval units; chunk size/overlap is tuned to balance precision and preserved context.

Question 9

The offline 'indexing' phase of RAG involves…

Accepted Answer

Loading, chunking, embedding docs — Before serving, the corpus is processed into embeddings/index so retrieval is fast at query time.

Question 10

Compared to fine-tuning, RAG is better when knowledge…

Accepted Answer

Knowledge changes often or grows large — Updating a document index is cheap and immediate; re-fine-tuning for every knowledge change is expensive and slow.

Key insight

Trace a RAG request

Test your knowledge

What are the three core steps of RAG?

The short version

In more depth

RAG

Interview Questions

References & Further Reading

Related topics

Love this visualization?