Artificial Intelligence MCQ

Q: The instruction 'category must be one of: easy, medium, hard' is an example of:

Correct answer: Value constraining to prevent invalid outputs. Constraining allowed values keeps structured output predictable and machine-safe.

Q: The self-critique pattern works by:

Correct answer: Generating output, asking the model to find flaws, then regenerating improved output. Generate → critique → regenerate uses the model's own analysis to raise quality.

Q: A proper prompt evaluation compares:

Correct answer: Outputs on a fixed test set, changing one element at a time. Controlled comparison on representative test inputs turns prompt improvement into evidence-based work.

Q: Prompt injection is the risk that:

Correct answer: Untrusted pasted text contains hidden instructions the model obeys. Malicious text inside processed content can hijack the model unless treated strictly as data.

Q: Which practice violates professional prompting standards?

Correct answer: Embedding customer passwords in prompts to a third-party API. Secrets and personal data must never be sent to external services inside prompts.

AI concepts, machine learning, generative AI and prompt engineering.

Q61. The instruction 'category must be one of: easy, medium, hard' is an example of: Medium

Permalink →

Q62. The self-critique pattern works by: Medium

Permalink →

Q63. A proper prompt evaluation compares: Medium

Permalink →

Q64. Prompt injection is the risk that: Medium

Permalink →

Q65. Which practice violates professional prompting standards? Medium

Permalink →

Q66. A prompt is 'brittle' when: Medium

Permalink →

Q67. RAG reduces hallucination by: Medium

Permalink →

Q68. 'If the answer is not in the text, say so' converts: Medium

Permalink →

Q69. Requesting supporting quotes for each claim enables: Medium

Permalink →

Q70. The flip-interaction pattern: Medium

Permalink →

🎯 Want a real test?

Take a timed quiz on this topic and see your rank on the leaderboard.

Browse Quizzes