Implement generative AI solutions →hardMultiple Choice

AI-102 Implement generative AI solutions Practice Question

Exhibit

Refer to the exhibit.
{
  "role": "system",
  "content": "You are a helpful assistant. Use the following sources to answer questions: [source1.pdf, source2.pdf]. If you cannot find the answer, say 'I don't know'."
}

You have deployed a chatbot using Azure OpenAI with a system message as shown. The chatbot sometimes provides incorrect answers that are not supported by the sources. What is the most likely cause?

⚠ Common exam trap

Many candidates assume a system message is sufficient to enforce factual accuracy, confusing instruction-following with grounded generation, and overlook the need for retrieval-augmented generation or explicit source constraints.

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

✓

The system message does not guarantee grounding; the model may still hallucinate.

A system message in Azure OpenAI provides instructions and context but does not enforce factual grounding. The model can still generate responses that are not supported by the provided sources (hallucination), especially if the system message is not explicitly designed to restrict the model to only use the given data. Grounding requires additional techniques like retrieval-augmented generation (RAG) or explicit constraints in the prompt.

Answer analysis

Option-by-option breakdown

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

✗
Content filters are incorrectly configured, allowing harmful content.
Why it's wrong here
Content filters screen prompts and completions for harmful categories such as violence or hate; they do not verify factual grounding against retrieved sources. Misconfigured filters would surface toxic text, not unsupported claims. Filters are the right control when the requirement is blocking harmful or prohibited content in a compliant deployment.
✗
The data ingestion pipeline has errors, so the sources are not available.
Why it's wrong here
If ingestion failed, retrieval would return no source chunks, so the model would refuse or state it lacks information rather than fabricate grounded-sounding answers. Ingestion monitoring is the correct fix when responses are empty or cite missing documents, not when they contradict available sources.
✗
The temperature parameter is too low, causing repetitive answers.
Why it's wrong here
Low temperature reduces randomness, producing deterministic, repetitive output; it does not introduce unsupported claims. Fabrication stems from the model answering without grounding constraints. Raising temperature is appropriate when you need varied creative phrasing, not factual accuracy against sources.
✓
The system message does not guarantee grounding; the model may still hallucinate.
Why this is correct
System messages steer tone and boundaries but cannot enforce factual grounding; the model still generates tokens probabilistically, so unsupported claims remain possible. Since the stem's constraint is answers not supported by sources, hallucination persists regardless of prompt wording. Retrieval augmentation or grounding data is required to constrain outputs to source content.

Go deeper

AI-102 Implement generative AI solutions Practice Question

Exhibit

Why each option matters

Option-by-option breakdown

Related to this question