Prompt Engineering Practice Test

Prompt Engineering Practice Test: all 15 questions with answers

Try each question first, then open “Show answer” to check the correct option and explanation.

1. Q1.A model keeps giving vague answers. What usually helps most?

   • A.Writing the prompt in capital letters
   • B.Giving specific instructions, relevant context and the output format you want
   • C.Making the prompt as short as possible
   • D.Asking the same question again without changes
   Show answer

   Correct answer: B. Giving specific instructions, relevant context and the output format you want

   Models perform best when the task, context, audience and expected format are stated clearly.

2. Q2.What is few-shot prompting?

   • A.Limiting the model to a few words
   • B.Including a few worked examples of input and desired output in the prompt
   • C.Running the prompt a few times
   • D.Using a smaller model
   Show answer

   Correct answer: B. Including a few worked examples of input and desired output in the prompt

   Examples show the model the exact pattern, tone and format to follow.

3. Q3.What is zero-shot prompting?

   • A.Asking the model to do a task without giving any examples
   • B.A prompt with zero words
   • C.A prompt that always fails
   • D.Training a model from scratch
   Show answer

   Correct answer: A. Asking the model to do a task without giving any examples

   Zero-shot means the model relies only on the instructions, with no examples.

4. Q4.What is the main purpose of a system prompt?

   • A.To store the user's password
   • B.To set the model's role, rules and behaviour for the whole conversation
   • C.To speed up the API
   • D.To translate the user's message
   Show answer

   Correct answer: B. To set the model's role, rules and behaviour for the whole conversation

   System prompts give persistent instructions, such as role, tone, constraints and output rules.

5. Q5.When does asking the model to reason step by step help most?

   • A.Simple greetings
   • B.Multi-step problems such as maths, logic or planning
   • C.Spelling a single word
   • D.Choosing a random number
   Show answer

   Correct answer: B. Multi-step problems such as maths, logic or planning

   Step-by-step (chain-of-thought) reasoning gives the model room to work through intermediate steps before answering.

6. Q6.Your code must parse the model's reply. What is the best approach?

   • A.Hope the model formats it correctly
   • B.Ask for an exact format such as JSON with a defined schema, and validate the result
   • C.Ask for a friendly paragraph
   • D.Use a higher temperature
   Show answer

   Correct answer: B. Ask for an exact format such as JSON with a defined schema, and validate the result

   A defined schema (or a structured-output feature) plus validation makes output reliable for code.

7. Q7.What is an AI 'hallucination'?

   • A.The model refusing to answer
   • B.The model stating false or unsupported information confidently
   • C.A very slow response
   • D.An image generated by AI
   Show answer

   Correct answer: B. The model stating false or unsupported information confidently

   Hallucinations are fluent but incorrect or made-up statements.

8. Q8.Which prompt change best reduces hallucinations in a factual Q&A app?

   • A.Ask the model to sound more confident
   • B.Provide the source documents and tell the model to answer only from them and say when it does not know
   • C.Raise the temperature
   • D.Remove all context
   Show answer

   Correct answer: B. Provide the source documents and tell the model to answer only from them and say when it does not know

   Grounding the model in sources and allowing 'I don't know' cuts down on invented answers.

9. Q9.Why wrap pasted content in delimiters such as XML tags or triple quotes?

   • A.It makes the prompt look nicer
   • B.It clearly separates instructions from the data the model should work on
   • C.It reduces the token count to zero
   • D.It is required by every API
   Show answer

   Correct answer: B. It clearly separates instructions from the data the model should work on

   Delimiters stop the model from confusing your instructions with the document or user input.

10. Q10.What does a higher temperature setting usually do?

    • A.Makes outputs more random and varied
    • B.Makes outputs always identical
    • C.Makes the model faster
    • D.Increases the context window
    Show answer

    Correct answer: A. Makes outputs more random and varied

    Higher temperature samples less likely tokens more often, giving more varied output. Lower is more deterministic.

11. Q11.What is prompt injection?

    • A.A way to make prompts shorter
    • B.Untrusted input that contains instructions trying to override the application's own instructions
    • C.Adding examples to a prompt
    • D.A method for fine-tuning
    Show answer

    Correct answer: B. Untrusted input that contains instructions trying to override the application's own instructions

    For example, a web page or email that says 'ignore previous instructions'. Treat such content as data, never as commands.

12. Q12.A complex task gives poor results in one long prompt. What is a good fix?

    • A.Break it into smaller steps and chain prompts, passing each result to the next
    • B.Add more exclamation marks
    • C.Remove all instructions
    • D.Use the same prompt with a random seed
    Show answer

    Correct answer: A. Break it into smaller steps and chain prompts, passing each result to the next

    Prompt chaining lets each step focus on one job and makes failures easier to debug.

13. Q13.Why give the model a role such as 'You are an experienced tax accountant'?

    • A.It changes the model's training data
    • B.It focuses tone, vocabulary and level of expertise for the task
    • C.It makes answers guaranteed to be correct
    • D.It is required for the API to work
    Show answer

    Correct answer: B. It focuses tone, vocabulary and level of expertise for the task

    A role frames how the model should approach the task. It does not add new knowledge or guarantee accuracy.

14. Q14.A prompt works on some inputs but not others. What is the best next step?

    • A.Try it once more and ship it
    • B.Build a small test set of example inputs and compare outputs as you change the prompt
    • C.Delete the prompt
    • D.Only test with your favourite example
    Show answer

    Correct answer: B. Build a small test set of example inputs and compare outputs as you change the prompt

    Evaluating against a fixed set of examples turns prompt tuning into measurable work instead of guesswork.

15. Q15.Which instruction is usually more effective?

    • A."Don't use markdown."
    • B."Write your answer as plain prose paragraphs."
    • C."No bad formatting!!!"
    • D."Format it however."
    Show answer

    Correct answer: B. "Write your answer as plain prose paragraphs."

    Telling the model what to do works better than only telling it what not to do.