SKILLUMEN
Try for free
Interview Questions  ›   RAG and retrieval

RAG

6 interview questions you should be able to answer.
3 worked answers   •   3 practice questions   •   Applied
Don't recall.
Read the page.
  Then answer.↘
Updated Sep 2026
01 CORE
1 / 6

What problem does RAG solve about a model's knowledge,
and how does it reduce hallucination?

Think about your answer before revealing.
Give it a try!
Interview answer
◷  ~ 30 seconds

A model's knowledge is frozen at its training cutoff and never included your private data, so it can't answer about recent events or your own docs, and when pushed it tends to hallucinate a confident guess.

RAG (retrieval-augmented generation) retrieves the most relevant real passages at question time and pastes them into the prompt, so the model answers from the page in front of it instead of from memory. Grounding it on real text sharply cuts made-up answers and lets it cite sources.

It also helps to instruct the model to answer only from that context, or to say "I don't know".

WITHOUT RAG LLM ✗ confident guess WITH RAG retrieved passages LLM ✓ grounded + cited
What the interviewer is testing
▤
Knowledgefrozen at cutoff
⌕
Retrievalreal passages in
▥
Hallucinationdrops sharply
Follow-up they may ask
02 PIPELINE
2 / 6

Walk through the RAG pipeline from documents at rest to an answer,
naming the retriever's role.

Sketch the two phases before revealing.
Interview answer
◷  ~ 35 seconds

Offline: split each document into chunks, embed each chunk into a vector, and store them in a vector store. This runs once, and re-runs when docs change.

Online: embed the incoming question the same way. The retriever compares that vector to the stored ones, usually by cosine similarity, and returns the top-k nearest chunks.

Those chunks are pasted into the prompt as context, and the generative model writes the final answer grounded in them. The retriever's job is the make-or-break middle step: it decides which passages the model gets to read.

OFFLINE · PREP THE KNOWLEDGE, ONCE Docs Chunk Embed Vector store the same store serves both lanes ONLINE · EVERY QUESTION Question Embed Retrieve top-k LLM ✓ grounded answer
03 DEEPER
3 / 6

Why does RAG let a model use private or very recent information
it was never trained on?

Think about your answer before revealing.
Interview answer
◷  ~ 25 seconds

RAG never asks the model to remember that information. It supplies it at question time.

The model's weights stay fixed, and the private or fresh knowledge lives in your document index, which you can update the moment a file changes.

Re-index the new doc and the assistant is current the same day, with no re-training and no waiting for a new model. That's why RAG is the go-to when an assistant needs to know your internal docs or answer about things that happened after the cutoff.

Model weights
Trained once
Frozen at the
cutoff
Never saw
your docs
Document index
Private + fresh
docs live here
Re-index when
a file changes
Current the
same day
More questions
04If a RAG system returns a bad answer, where is the failure most likely, and what three fixes address it?
05Why is the 'open-book exam' analogy apt? What does the model still do, and what does RAG hand it?
06Does RAG fully eliminate hallucination?
These run inside Skillumen with graded answers.

This is a small sample.
600+ interview questions live inside Skillumen.

Every concept card carries its own interview questions, graded as you answer.
Free to start.

See more interview questions
←  HallucinationRAG vs Fine-tune vs Long-Context  →