AI Associate [AI-102] 5 — Questions and Answers
Question 1: Which Azure OpenAI feature allows you to extend the model's capabilities by letting it call external APIs or run code?
- Prompt engineering
- Function calling (tools) (Correct answer)
- Fine-tuning
- Embeddings
Correct answer: Function calling (tools)
Function calling enables the model to generate structured JSON arguments to invoke developer-defined functions or external APIs as part of a conversation.
Question 2: You must ensure your Azure AI solution logs all inputs and outputs for compliance auditing. Which Azure service provides this capability when integrated with Azure OpenAI?
- Azure Monitor with diagnostic settings (Correct answer)
- Azure Defender for Cloud
- Azure Service Bus
- Azure Traffic Manager
Correct answer: Azure Monitor with diagnostic settings
Azure Monitor diagnostic settings can capture Azure OpenAI request and response logs, storing them in Log Analytics or a storage account for auditing.
Question 3: What does 'few-shot prompting' mean when working with large language models?
- Limiting the model to generate only a few tokens
- Providing a small number of input-output examples in the prompt to guide model behavior (Correct answer)
- Fine-tuning the model on a small dataset
- Reducing the model's temperature to near zero
Correct answer: Providing a small number of input-output examples in the prompt to guide model behavior
Few-shot prompting includes a handful of example input-output pairs in the prompt itself, showing the model the desired behavior without retraining.
Question 4: Which Azure AI service would you use to automatically extract and recognize handwritten text from scanned documents?
- Azure AI Language
- Azure AI Vision with OCR (Correct answer)
- Azure Custom Vision
- Azure Video Indexer
Correct answer: Azure AI Vision with OCR
Azure AI Vision's OCR capability extracts printed and handwritten text from images and documents using deep learning-based recognition models.
Question 5: When fine-tuning an Azure OpenAI model, what data format is required for the training file?
- CSV with columns 'prompt' and 'completion'
- JSON Lines (JSONL) with 'messages' in chat format or prompt/completion pairs (Correct answer)
- Plain text with examples separated by newlines
- XML with <example> tags for each training pair
Correct answer: JSON Lines (JSONL) with 'messages' in chat format or prompt/completion pairs
Azure OpenAI fine-tuning requires JSONL files where each line is a JSON object with 'messages' (chat format) or 'prompt'/'completion' fields.
Question 6: A chatbot built with Azure Bot Service needs to hand off complex queries to a human agent. Which Bot Framework feature supports this pattern?
- Adaptive Cards
- Handoff protocol (activity-based handoff) (Correct answer)
- QnA Maker integration
- LUIS entity recognition
Correct answer: Handoff protocol (activity-based handoff)
The Bot Framework handoff protocol allows a bot to signal that a conversation should be transferred to a human agent using standardized handoff activity events.
Question 7: In Azure AI Studio, what is the primary purpose of the 'prompt flow' tool?
- Monitoring token usage across Azure OpenAI deployments
- Building, testing, and deploying LLM-based workflows as executable flows (Correct answer)
- Auto-generating system prompts from user requirements
- Managing API keys and access policies for AI services
Correct answer: Building, testing, and deploying LLM-based workflows as executable flows
Prompt flow provides a visual IDE for creating, evaluating, and deploying complex LLM-driven workflows that chain prompts, tools, and data sources.
AI Associate [AI-102] 5
Which Azure OpenAI feature allows you to extend the model's capabilities by letting it call external APIs or run code?