You want to implement "Long-Term Memory" by storing user preferences across months. Which tool is best for this?
check_circle
Correct AnswerB
While Threads are for sessions, a Vector Store can store and retrieve historical user data for long-term personalization.
Question 22
You notice that your RAG application's costs are high because the system retrieves 20 large document chunks for every query. Which technique would best reduce token usage while maintaining quality?
check_circle
Correct AnswerA
Small-to-Large retrieval fetches a small sentence for matching but only sends the surrounding context to the LLM if needed, saving prompt tokens.
Question 23
In the "Azure AI Agent Service", what is the primary benefit of using "Code Interpreter" for data analysis over a standard LLM prompt?
check_circle
Correct AnswerB
LLMs are prone to calculation errors; Code Interpreter ensures accuracy by executing actual code to perform the math.
Question 24
Which metric in AI Foundry Evaluation measures the "Similarity" between two vectors to determine how well the search result matches the query?
check_circle
Correct AnswerC
Cosine Similarity is the standard metric for measuring the orientation (semantic similarity) between two high-dimensional embeddings.
Question 25
In a "Multi-Agent" environment, what is a "Supervisor" agent responsible for?
check_circle
Correct AnswerD
A Supervisor or Orchestrator agent manages the lifecycle of a complex task by delegating work to specialized sub-agents.
Question 26
You are building an agent that needs to create a summary of a recorded 20-minute phone call. Which strategy is most efficient?
check_circle
Correct AnswerC
Converting audio to text first is the standard architecture for allowing an LLM to "read" and summarize spoken content.
Question 27
Which Azure AI Foundry tool allows you to run a "Benchmarking" job to compare the latency and throughput of GPT-4o versus a Llama-3-70B model?
check_circle
Correct AnswerD
The Benchmarking feature in the Model Catalog allows developers to compare performance metrics of different models using standardized datasets.
Question 28
Which Azure AI Search feature helps resolve the "Empty Result Set" problem by suggesting alternate terms when a user's query returns 0 hits?
check_circle
Correct AnswerC
Synonym maps expand the query to include related terms, increasing the likelihood of finding a match even if the exact keyword is missing.
Question 29
You want to give your agent the ability to read and describe a complex architectural diagram in an image. Which model is best?
check_circle
Correct AnswerA
GPT-4o’s vision capabilities allow it to perform high-level reasoning and data extraction from visual inputs like diagrams.
Question 30
When monitoring an AI agent in production, you see a high number of 429 Too Many Requests errors. What is the most likely cause?
check_circle
Correct AnswerB
429 errors specifically indicate that the request rate has exceeded the provisioned or default quota for the deployment.
