Compare RAG Retrieval Approaches

HardGenerative AI & LLMs00:00

Practice interviewer

In session

5 left

00:00

Your question is Compare RAG Retrieval Approaches. Take a moment with it on the right.

Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).

You need to log in / sign up to chat or submit.

Problem

Scenario

You're building a retrieval layer for a RAG system and need to choose between semantic search, keyword search, and a hybrid approach. The corpus includes both natural language documents and text with exact identifiers like product names, error codes, and API fields.

Question

How would you compare semantic search, keyword search, and hybrid search for a RAG retriever? When would you prefer each one, and how would you evaluate the choice?

What this tests

Understanding of BM25 keyword retrieval versus dense vector retrieval
Ability to reason about mixed query types in RAG
Evaluation of retrievers before generation quality
Awareness of grounding, hallucination, and retrieval safety