AI-300 Practice Question 80: RAG vs Fine Tuning
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 8 · question 10 of 10
A company's product documentation changes every week. Its assistant must answer with current information and point to the source documents. The team is considering fine-tuning a model on the documentation. Which approach is best?
Show answer and explanation
Answer:
Fine-tuning changes model behavior and style, but it is a poor way to add facts that change often, and it gives no source references. RAG retrieves current passages at query time and passes them to the model, so answers stay current and can cite sources, and the index only needs a refresh when documents change (option C).
- A is wrong because answers would be out of date between training runs, and a fine-tuned model cannot cite the document it drew a fact from. - B is wrong because this is very costly per request and quickly exceeds the context window as the documentation grows. - D is wrong because distillation transfers behavior at a point in time. It does not add current knowledge or source citations.
Reference: Fine-tuning considerations