Module assessment

1.

What is the main purpose of retrieval-augmented generation (RAG)?

To retrain a language model whenever source information changes.

To ground generated responses in relevant information retrieved from a data source.

To replace a language model with a search engine.

2.

Why are large documents commonly divided into chunks before they're indexed?

So the solution can retrieve focused passages that fit efficiently into a prompt.

So every chunk can be used to train a separate language model.

So the original documents no longer need to be stored.

3.

What does an embedding represent in a RAG solution?

A numerical vector that captures semantic characteristics of content.

A set of instructions that defines the model's behavior.

A final answer generated for a user.

4.

Which search approach can find a passage about 'company travel costs' when the query asks about 'business trip expenses,' even if the words differ?

Semantic vector search.

An exact filename match.

Random document selection.

5.

A RAG response is fluent but contains an incorrect policy limit. What should be examined first?

Whether the correct and current policy passage was retrieved and supplied as context.

Whether the language model can produce a longer response.

Whether more unrelated documents can be added to the prompt.

Feedback