Lesson 7 Quiz — RAG, Vector DBs & Customization Tradeoffs (10 easy + 10 hard)

What does RAG stand for?

What does Amazon Bedrock Knowledge Bases primarily provide?

Which AWS service is a dedicated search/vector engine popular for RAG at scale?

Which database adds vector search to a relational database via the pgvector extension?

Lower temperature in inference parameters generally produces:

What is prompt caching primarily used for?

Which is the cheapest/fastest customization approach to try first?

Which vector database option is best when relationships between entities matter alongside similarity?

What distinguishes an AI agent from a plain foundation model call?

Which customization approach is generally the MOST expensive and time-consuming?

A team needs a chatbot to answer questions grounded in their internal, frequently-updated policy documents, without retraining. Best approach?

Which of these are legitimate AWS vector-database options mentioned for RAG use cases?

Order these customization approaches from cheapest/fastest to most expensive/slowest.

Match each RAG-pipeline step to its description.

A high-volume production feature needs a much cheaper model that mimics a larger model's behavior on a narrow task. Best fit?

A factual customer-support Q&A bot should generally use which temperature setting?

Which factors are part of practical FM selection at the application-design level (beyond business-level factors)?

Why does RAG specifically help reduce hallucinations compared to relying on the base model alone?

A team repeatedly sends the same long system prompt with each new user question. What technique cuts the repeated cost/latency of reprocessing that prefix?

A customer-service agent needs to not just answer a billing question but actually issue a refund through an internal API. What capability does this require beyond a plain FM call?