179 questions
No questions match those filters.
← All questions
RAGEasy· RAG & Retrieval
When would you NOT use RAG?
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansThree cases where reaching for RAG by default is the wrong call:
- The problem is behavior, not knowledge — tone, format, domain-specific style. Fine-tuning fixes that; retrieval doesn’t.
- The knowledge is small and stable — if it fits comfortably in a system prompt and doesn’t change often, just prompt it. Retrieval adds latency and failure surface for no benefit.
- The answer requires reasoning across relationships spanning many documents, not locating a single passage — at that point a graph representation of the entities and relationships usually serves better than similarity search over chunks.
RAG earns its place specifically for knowledge that changes over time and where the answer needs to be traceable back to a source.