AB-410 Exam Questions

Implicit sharing lets the app creator share the AI capability without giving each user direct access to the underlying service.

Category Classification lets you train a model to assign text to custom groups based on your specific business requirements.

Inclusiveness means technology should be usable by everyone, regardless of physical or cognitive abilities.

Azure APIM lets you secure, throttle, and monitor the APIs consumed by intelligent applications.

Prompt Shields are designed to detect and block malicious content concealed in user inputs or retrieved data.

PII detection identifies and masks sensitive data points found in a text string.

ACS provides APIs for voice, video, and SMS, enabling seamless transitions from digital bots to live voice.

The System Message sets the AI session‘s core behavior, persona, and language constraints.

Custom Connectors enable Power Platform tools to communicate with any service that exposes a RESTful API.

The embedding model changes the text query into a numerical vector so it can be compared with indexed data.

Formatting rules such as “output as a bulleted list” are core behavioral instructions defined in the System Message.

Plugins enable agents to take actions and retrieve live data from external systems by using Power Automate connectors.

Entity Extraction is designed to identify and extract structured data elements from natural language text.

Regular testing against a “Golden Dataset” of current data helps keep accuracy high as real-world inputs change.

Prompt engineering involves writing precise prompts that guide an LLM toward a specific, high-quality response.

The Alpha parameter, ranging from 0 to 1, controls the extent to which the search engine uses semantic vector scores compared with traditional keyword scores.

Power Automate flows connect Copilot Studio to external systems by using connectors or custom APIs.

The Model Catalog in Azure AI Studio serves as the central place to discover and deploy both frontier models and open-source models.

In the Power Platform, solutions serve as containers for packaging and moving apps, flows, and bots between environments.

OCR is a cognitive skill in the AI Search enrichment pipeline that extracts text from images and PDFs so it can be indexed.

Grounding anchors the AI’s responses to your specific business data instead of relying only on general training knowledge.

A lower temperature, close to 0, makes the model more deterministic and factual, while a higher temperature makes responses more creative and varied.

Phi-3 models are optimized for efficiency and can run in environments where low latency and low power consumption are critical.

Stop sequences prevent the model from continuing unnecessarily and can force generation to end after a specific output pattern, such as a closing bracket.

Max Tokens sets the maximum output length, which helps control both response size and token costs.

The Prebuilt Receipt model is specifically trained to identify and extract key financial fields from different receipt formats.

Image Classification tags or categorizes a complete image using predefined labels.

Frequency penalty lowers the chance that the model will repeat tokens that have already appeared often in the output.

User-scoped variables let the bot remember a specific person’s preferences or history when that person returns to the bot later.

Reliability and Safety make sure the application behaves predictably and avoids harm in edge-case scenarios.

Azure AI Vision can extract text from images and video frames in real time, including when motion is involved.

Brand Detection locates thousands of well-known logos and brands within video frames.

Prediction models are designed to forecast binary or multiple outcomes from historical patterns in your data.

Language Detection identifies the source language, and AI Translator converts the text into the target language for summarization.

Dataverse supports geospatial data, including Latitude and Longitude, and supports queries that calculate distances and locate nearby places.

This trigger is important for tailoring the beginning of a session or checking authentication before the user says anything.

Vector search uses the distance between high-dimensional numeric representations of text to find semantic matches.

Prompt Flow in Azure AI Studio lets developers build executable graphs that connect LLMs, prompts, and Python code.

Document Translation is a specialized Translator service feature that preserves the layout and structure of the original file.

A temperature of 0.0, or close to zero, makes the model deterministic, which supports consistent and focused responses.

This principle makes sure the system is resilient, robust, and avoids causing harm during failures or edge cases.

AI Builder identifies the object, while Dataverse stores the relationship between that object and the manual’s location.

Copilot Studio is the centralized platform for designing, publishing, and managing intelligent conversational agents.

Tokenization converts raw text into numerical tokens, which serve as the basic data units for transformer-based models.

Azure AI Content Safety provides Prompt Shields, which help protect against malicious inputs intended to override or undermine model instructions.

System topics are built-in templates that handle the basic conversational flow for every bot.

Trigger phrases and intent recognition connect the user’s natural language to the correct logic for refunds or exchanges.

Zero-shot prompting depends on the model‘s existing knowledge to complete a request without task-specific examples.

Security Roles and Business Units establish horizontal data access boundaries within an organization.

Speech-to-Text transcribes spoken language into a machine-readable format.

Zero-shot prompting depends on the model’s internal general knowledge to complete a request without specific examples for the task.

Transparency makes the logic and data behind AI decisions clear and understandable for stakeholders and users.

PII detection, a feature of Azure AI Language, identifies and masks sensitive data.

Formula columns support real-time calculations in Dataverse, giving an AI agent pre-processed logic to consume.

Content filtering is a core safety capability that helps the AI follow responsible AI standards by blocking unsafe content.

The Unstructured Document model is built for documents where text and data points appear in different positions across files.

Reliability and Safety ensure AI applications are robust and avoid causing harm or producing dangerous errors in edge-case scenarios.

Azure AI Content Safety is a dedicated service for monitoring and filtering harmful content in AI applications.

Correct: B. Asking an AI to perform a task without providing any examples in the prompt.
Explanation:
In the context of AI and LLM (Large Language Model) prompting techniques, Zero-shot prompting is a method where you provide the model with a task, question, or instruction without supplying any prior examples (shots) of the expected input-output behavior. You rely entirely on the model‘s pre-trained knowledge to understand the intent and generate the correct response.

How it works: You simply ask the model to perform a task directly (e.g., “Classify this sentiment: ‘I love this product‘“).

Contrast: This is distinct from Few-shot prompting, where you provide a few examples in the prompt to guide the model‘s output format or tone.

Incorrect:
A. A prompt with an error: This is incorrect. A prompt with an error is simply a poorly constructed or invalid prompt, which has nothing to do with the “Zero-shot“ classification technique.

C. A prompt for image generation only: This is incorrect. While zero-shot prompting can be used with image generation models (like asking a model to “Draw a futuristic city“ without showing it existing city sketches), the term is a general prompting methodology applicable to text, code, image, and multi-modal models alike.

D. A prompt that costs $0: This is incorrect. The “zero“ in “zero-shot“ refers to the number of examples provided, not the financial cost of executing the prompt.

Security Roles and Business Units set the horizontal and vertical data access boundaries within an organization.