3.3 Identify Grounding Techniques and Their Use Cases. GENERATIVE-AI-LEADER Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A retail company wants to deploy an internal assistant that answers questions about its product catalog, pricing, and inventory policies. The catalog is stored in a structured database, and the policies are in PDF documents. The team has no ML engineers and wants a solution that works out of the box. Which approach is most appropriate?

   • ABuild a custom RAG pipeline using the RAG APIs.
   • BUse Google Search grounding to retrieve the catalog and policies.
   • CUse Agent Search to index the catalog and policy documents and ground the assistant's responses.
   • DFine-tune a foundation model on the catalog and policies.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   How are safety settings typically configured in a model's API request?

   • ABy specifying thresholds for different content categories
   • BBy providing a list of banned words
   • CBy setting a maximum token count
   • DBy adjusting the temperature parameter
   Reveal Solution Discussion Select an answer first
4. 28application · easy

   A legal document summarization tool must produce consistent, factual summaries of contracts. The team notices that the same contract sometimes yields slightly different summaries on repeated runs. Which configuration change is most likely to make the outputs more consistent?

   • ADecrease temperature to 0.1 and top-p to 0.2.
   • BDisable safety settings.
   • CIncrease temperature to 1.0 and top-p to 1.0.
   • DIncrease the max output token count.
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A travel agency wants to build a trip-planning assistant. The assistant must answer questions about real-time flight delays, weather conditions at destinations, and local events happening this week. The agency has no internal data for these topics. Which grounding data source should the team use?

   • AThe model's internal parametric knowledge from its training corpus.
   • BFirst-party enterprise data from the agency's booking system.
   • CThird-party data from a static industry report on travel trends.
   • DWorld data via grounding with Google Search.
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A mid-sized insurance company wants to deploy an internal assistant that answers employee questions using the company's policy manuals, HR documents, and IT runbooks. The team has limited ML engineering resources and wants the fastest path to a working solution. Which approach should they choose?

   • AFine-tune a foundation model on all internal documents.
   • BBuild a custom RAG pipeline from scratch using the RAG APIs and a vector database.
   • CUse Google Cloud's pre-built RAG solution with Agent Search.
   • DGround the model with Google Search so it can find the documents online.
   Reveal Solution Discussion Select an answer first