Courseiva

AI-900 Practice Question: Describe features of generative AI workloads on Azure

What is a 'hallucination' in the context of large language models?

⚠ Common exam trap

Many exam-takers confuse a model's refusal to answer (safety guardrails) with a hallucination, or think that running out of context window is a type of hallucination, when in fact hallucination is specifically about generating confident but false content.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

When a model generates plausible-sounding but factually incorrect information

In the context of large language models (LLMs), a hallucination occurs when the model generates text that is fluent, coherent, and plausible-sounding but is factually incorrect or nonsensical. This happens because LLMs are trained to predict the next token based on statistical patterns in their training data, not to verify facts against a ground truth. Option B correctly identifies this behavior.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

  • When a model refuses to answer a question

    Why it's wrong here

    Refusing to answer is an intentional safety-guardrail behavior, typically implemented through instruction tuning, RLHF, or classifier-based filters, so the model declines requests that could cause harm. A refusal is an abstinence decision: the model says it cannot or should not comply. Hallucination is the opposite failure — the model complies eagerly but produces plausible misinformation, so it is a truthfulness problem rather than a compliance or safety mechanism.

  • When a model generates plausible-sounding but factually incorrect information

    Why this is correct

    Hallucination arises because an autoregressive transformer samples the next token from a probability distribution learned over human text, optimizing for linguistic fluency and coherence rather than verifiable truth. Without an external knowledge base or retrieval verification, the model can confidently emit well-formed statements that are entirely fabricated. This is the key reliability limitation of purely statistical language models.

  • When a model processes images instead of text

    Why it's wrong here

    Processing images instead of text refers to multimodal input handling, such as when a vision-language model accepts an image and produces a caption or answers visual questions. That is a separate capability and not an accuracy failure. Hallucination, by contrast, is a text-generation defect in which the model invents facts while producing natural-sounding language, regardless of whether the input was text or an image.

  • When a model runs out of context window space

    Why it's wrong here

    Running out of context window space is a hard technical constraint: the model's attention mechanism can only attend to a fixed maximum number of tokens, so longer inputs are truncated or must be summarized. This causes loss of earlier information and degraded performance, but it is not the same failure mode as hallucination. Hallucination is not about forgetting due to an exhausted buffer; it is about generating false content that was never in the input at all.

About these practice questions

This AI-900 question is part of Courseiva's 985-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →

How Courseiva writes practice questions · Editorial policy

JA

Written by Johnson Ajibi, MSc IT Security

Senior Network & Security Engineer · founder of Courseiva

This AI-900 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI-900 exam.