AI-102 Practice Test 33 — 15 Questions

Question 1

A developer wants to use Azure OpenAI to generate text from a prompt. Which parameter controls the diversity of the generated output?

Accepted Answer

temperature. Temperature is the parameter that directly controls the randomness or diversity of the generated output by scaling the logits before applying the softmax function. A higher temperature (e.g., 1.0) increases the probability of less likely tokens, producing more creative and varied responses, while a lower temperature (e.g., 0.1) makes the output more deterministic and focused.

Answer

presence_penalty

Answer

frequency_penalty

Answer

max_tokens

Question 2

A retail company uses Azure Computer Vision to analyze in-store camera feeds to count customers. The solution uses the Detect API on individual frames. Recently, the counts have been inaccurate due to overlapping people. Which action should you take to improve accuracy?

Accepted Answer

Train a Custom Vision object detection model with images of crowded scenes.. The Detect API in Azure Computer Vision is a general-purpose object detection model that may not be optimized for crowded scenes with significant overlap. By training a Custom Vision object detection model with images of crowded scenes, you create a specialized model that learns to distinguish partially occluded people, improving counting accuracy. This approach leverages transfer learning to adapt the model to the specific visual patterns of your store environment.

Answer

Use the Analyze API instead of Detect API.

Answer

Increase the confidence threshold in the Detect API call.

Answer

Switch to Azure Video Indexer for people counting.

Question 3

You are designing an agentic solution using Azure AI Agent Service. The agent needs to perform actions on behalf of users, such as sending emails and updating databases. The solution must use managed identities for authentication to Azure resources. Which TWO configurations are required?

Accepted Answer

Use DefaultAzureCredential in the agent's code to authenticate to Azure services. DefaultAzureCredential is the recommended authentication mechanism for Azure SDKs when using managed identities. It automatically chains multiple credential sources, including environment variables, managed identity endpoints, and Visual Studio credentials, allowing the agent to authenticate to Azure services without hardcoding secrets. This aligns with the requirement to use managed identities for authentication.

Answer

Store connection strings in Azure Key Vault and reference them in the agent's configuration

Answer

Create a service principal in Microsoft Entra ID and assign RBAC roles to the agent's resource

Answer

Configure the agent to use an API key for each external service

Question 4

Your organization is building a knowledge base from technical manuals stored in multiple formats (PDF, Word, HTML). You need to extract text and images from these documents and create a searchable index. The solution must handle tables and preserve their structure. Which approach should you use?

Accepted Answer

Use Azure AI Document Intelligence layout model as a custom skill. Azure AI Document Intelligence Layout model extracts text, tables, and their structure from documents, making it suitable for preserving table structure and building a searchable knowledge base. It can be used as a custom skill in an Azure AI Search enrichment pipeline. Option A is incorrect because uploading documents directly to Azure AI Search only indexes raw text without extraction or table structure preservation. Option B is incorrect because Custom Entity Extraction in Azure AI Language focuses on identifying entities, not extracting tables or preserving structure. Option D is incorrect because Azure AI Vision OCR extracts text but not table structure, and is not designed for document layout analysis.

Answer

Upload documents directly to Azure AI Search

Answer

Use Azure AI Language custom entity extraction

Answer

Use Azure AI Vision OCR skill in the skillset

Question 5

You have an Azure AI Search index defined as shown in the exhibit. Users want to filter search results by author and by a date range, and also see a count of documents per tag. However, the filter on author is not working. What is the most likely reason?

Accepted Answer

The filter expression uses incorrect OData syntax.. The most likely reason the filter on author is not working is an incorrect OData syntax in the filter expression. Since the exhibit shows the 'author' field is marked as filterable, the index definition itself is correct. However, if the filter uses invalid syntax (e.g., incorrect operator, missing quotes, or malformed date comparison), the filter will fail. Option B is incorrect because the 'id' field's role as a key is unrelated to filtering on 'author'. Option C is incorrect because the 'author' field is filterable as shown in the exhibit. Option D is incorrect because the '$orderby' parameter does not conflict with filters; they can be used together.

Answer

The 'id' field is not used as the key.

Answer

The 'author' field is not set as filterable in the index definition.

Answer

The query uses a $orderby parameter that conflicts with the filter.

Question 6

You are using Azure OpenAI Service to summarize customer emails. The summaries must be concise and contain only key information. Which prompt engineering technique should you apply?

Accepted Answer

Use few-shot prompting with examples of desired summaries. Few-shot prompting is the correct technique because it provides the model with explicit examples of desired input-output pairs (e.g., a verbose email and its concise summary). This guides the model to learn the exact format, tone, and level of detail required for the summaries, which is critical for consistency in a production summarization pipeline. Without examples, the model may default to its training distribution and produce overly verbose or irrelevant output.

Answer

Use chain-of-thought prompting

Answer

Use zero-shot prompting with a one-sentence instruction

Answer

Use negative prompting to avoid verbose output

Question 7

You are reviewing the skillset definition for an Azure AI Search indexer. The SplitSkill splits the document content into pages of 5000 characters. The SentimentSkill is set to run on each page. However, the sentiment analysis is not producing correct results. What is the most likely cause?

Accepted Answer

The context of the SentimentSkill is set to an array, which is not supported. Setting the context of the SentimentSkill to an array (e.g., '/document/pages') causes the skill to run once on the entire array of pages, rather than iterating over each page individually. This results in incorrect sentiment analysis because the skill receives a collection of strings as a single input, which is not the intended per-page processing. The correct configuration is to set the context to '/document/pages/*' so that the skill executes on each page separately.

Answer

The maximumPageLength of 5000 is too high for sentiment analysis

Answer

The input source for SentimentSkill should be '/document/pages/*' but the SplitSkill output is named 'pages', so the input should be '/document/pages'

Answer

The SentimentSkill uses an incorrect @odata.type version

Question 8

A company is using Azure Form Recognizer to extract data from invoices. The prebuilt model does not correctly extract a custom field that is specific to the company's invoices. What is the most appropriate action to improve extraction accuracy for this field?

Accepted Answer

Train a custom model using labeled invoices that include the custom field.. The prebuilt Form Recognizer model is designed for common invoice layouts and may not recognize company-specific fields. Training a custom model with labeled invoices that include the custom field allows the model to learn the field's location and semantics, significantly improving extraction accuracy for that specific field.

Answer

Use the prebuilt model with a custom field mapping.

Answer

Adjust the confidence threshold for the prebuilt model.

Answer

Retrain the prebuilt model with additional invoices.

Question 9

You are using Azure AI Language to analyze social media comments. You need to identify the language of each comment and then extract key phrases. Which TWO features should you use? (Select TWO.)

Accepted Answer

Language detection. Language detection is the correct feature because it identifies the language of each comment, which is a prerequisite for further analysis. Key phrase extraction is the second correct feature because it extracts important terms from the text, directly addressing the requirement to 'extract key phrases' after language identification.

Answer

Sentiment analysis

Answer

Summarization

Answer

Entity recognition

Question 10

Which TWO actions can you take to mitigate the risk of generating harmful content when using Azure OpenAI Service? (Choose two.)

Accepted Answer

Set a system message that instructs the model to avoid harmful outputs.. Setting a system message instructs the model to avoid harmful outputs, guiding its behavior. Option D is correct because Azure AI Content Safety filters automatically detect and block harmful content. Option B is wrong because fine-tuning on safe examples may not fully prevent harmful generations; content safety is more reliable. Option C is wrong because deploying in multiple regions does not directly affect content safety. Option E is wrong because increasing maxTokens does not mitigate harmful content risk; it only controls response length.

Answer

Fine-tune the model on a dataset of safe examples.

Answer

Deploy the model in multiple regions.

Answer

Increase the maxTokens parameter to allow longer responses.

Question 11

You need to create a chatbot that uses Azure OpenAI to answer questions about your company's internal policies. The responses must be based only on the provided policy documents. Which approach should you use?

Accepted Answer

Use Retrieval-Augmented Generation (RAG) with an Azure AI Search index of the documents.. Retrieval-Augmented Generation (RAG) uses an Azure AI Search index to retrieve relevant policy documents and then generates answers based solely on that content, ensuring responses are grounded in the provided documents. Option A is wrong because the model's pre‑existing knowledge is generic and not specific to the company's internal policies, so answers may be inaccurate or unsupported. Option B is wrong because fine‑tuning a GPT model on the policy documents can still lead to hallucination; the model might generate plausible‑sounding but ungrounded content not directly from the documents. Option C is wrong because prompt engineering alone cannot force the model to ignore its training data; without retrieval, the model may incorporate external knowledge or fabricate information.

Answer

Use the model's pre-existing knowledge about common policies.

Answer

Fine-tune a GPT model on the policy documents.

Answer

Use prompt engineering to instruct the model to only use policy knowledge.

Question 12

Your company is building a multi-tenant SaaS application using Azure AI Foundry. Each tenant (customer) should have isolated model deployments and data, but you want to share the base models across tenants to reduce costs. Which approach should you use?

Accepted Answer

Deploy one model per base model and use separate endpoints with routing logic per tenant.. Deploying one model per base model and using separate endpoints with routing logic per tenant provides tenant isolation while sharing the base model, reducing costs. Option A is incorrect because using a single deployment with a tenant ID in the header does not provide sufficient isolation and may lead to data leakage. Option B is incorrect because creating separate fine-tuned models per tenant increases costs and does not share the base model. Option C is incorrect because deploying separate base models for each tenant also increases costs unnecessarily.

Answer

Use a single model deployment and a shared endpoint with tenant ID in the header.

Answer

Create separate fine-tuned models for each tenant.

Answer

Deploy separate base models for each tenant.

Question 13

Your company uses Azure Bot Service with QnA Maker to answer HR questions. You need to migrate to Microsoft Copilot Studio to leverage generative AI capabilities. You have an existing QnA Maker knowledge base. What is the recommended migration path?

Accepted Answer

In Copilot Studio, use the 'Import from QnA Maker' option to directly bring in the knowledge base.. Copilot Studio provides a direct 'Import from QnA Maker' option that migrates the entire knowledge base automatically, preserving Q&A pairs, metadata, and synonyms. Option B is incorrect because manually recreating each Q&A pair is unnecessary and inefficient when the native import tool exists. Option C is incorrect because Azure PowerShell does not have a specific cmdlet to export a QnA Maker knowledge base for direct import into Copilot Studio; the import feature handles the migration. Option D is incorrect because exporting to TSV and importing into Azure AI Language custom question answering is an intermediate step that doesn't directly move the KB into Copilot Studio; the simplest path is the built-in import.

Answer

Create a new Copilot and manually recreate each Q&A pair.

Answer

Use Azure PowerShell to export the knowledge base and then upload to Copilot Studio.

Answer

Export the QnA Maker KB to a TSV file, then import into Azure AI Language custom question answering.

Question 14

You are deploying a computer vision model using Azure AI Custom Vision with a small dataset of 200 images per class. The model shows high accuracy on training data but low accuracy on test data. Which action should you take to reduce overfitting?

Accepted Answer

Increase the training dataset size with more varied images. Increase the training dataset size with more varied images. Overfitting occurs when the model learns noise from a small dataset. Adding more varied images helps the model generalize. Option A (increase learning rate) is wrong because it may cause divergence or unstable training, not directly reduce overfitting. Option B (reduce image size) is wrong because it can lose important features and may even increase overfitting. Option C (increase training epochs) is wrong because it can actually increase overfitting by allowing the model to memorize more noise.

Answer

Increase the learning rate

Answer

Reduce the image size to lower resolution

Answer

Increase the number of training epochs

Question 15

Which THREE Azure AI services can be used to extract text from images?

Accepted Answer

Azure AI Search. The question asks for three Azure AI services that can extract text from images. Azure AI Document Intelligence layout model (C) and Azure AI Vision OCR (D) are direct text extraction services. Azure AI Search (B) can also extract text from images by using OCR as a built-in cognitive skill in an enrichment pipeline. Option A (Speech) processes audio, not images. Option E (Language custom NER) works only on pre-extracted text. Therefore, the correct options are B, C, and D.

Answer

Azure AI Speech

Answer

Azure AI Language custom NER