Skip to content

Quiz 4: Prompt Engineering

Test your understanding of prompt engineering techniques and strategies.

Questions

Question 1 (Remember)

What is zero-shot prompting?

  1. Asking the model to perform a task without providing examples
  2. Using no system prompt
  3. Generating zero tokens
  4. A technique that always fails
Answer

A) Asking the model to perform a task without providing examples - Zero-shot prompting relies on the model's pre-trained knowledge without task-specific examples. The model must understand the task from instructions alone.

Question 2 (Remember)

What is chain-of-thought prompting?

  1. Linking multiple prompts together
  2. Encouraging the model to show step-by-step reasoning
  3. Using multiple models in sequence
  4. A type of system prompt
Answer

B) Encouraging the model to show step-by-step reasoning - Chain-of-thought prompting asks the model to "think step by step," which improves accuracy on complex reasoning tasks by making intermediate steps explicit.

Question 3 (Understand)

Why does few-shot prompting often outperform zero-shot for specialized tasks?

  1. It uses more tokens
  2. It costs more
  3. Examples demonstrate the desired pattern and format
  4. It always produces longer outputs
Answer

C) Examples demonstrate the desired pattern and format - Few-shot examples show the model exactly what input/output patterns you expect, reducing ambiguity and improving consistency for specialized tasks.

Question 4 (Understand)

What is the purpose of a system prompt?

  1. To increase token count
  2. To define the AI's behavior, persona, and constraints
  3. To bypass safety features
  4. To reduce costs
Answer

B) To define the AI's behavior, persona, and constraints - System prompts establish context, role, communication style, and boundaries that persist throughout the conversation.

Question 5 (Apply)

You need consistent JSON output from an LLM. Which technique is most effective?

  1. Ask politely for JSON
  2. Provide JSON schema and examples, specify "respond only with valid JSON"
  3. Use maximum temperature
  4. Avoid all instructions
Answer

B) Provide JSON schema and examples, specify "respond only with valid JSON" - Structured output requires clear format specification, examples of the expected structure, and explicit instructions to output only the specified format.

Question 6 (Apply)

A user complains that AI responses are too verbose. What prompt modification would help?

  • "Give me more details"
  • "Respond in 2-3 sentences maximum"
  1. Remove the system prompt
  2. Increase temperature
Answer

B) "Respond in 2-3 sentences maximum" - Explicit length constraints in the prompt directly address verbosity. Being specific about desired output length helps control response size.

Question 7 (Analyze)

Compare tree-of-thought with chain-of-thought prompting:

  1. They are identical techniques
  2. Tree-of-thought is simpler
  3. Tree-of-thought explores multiple reasoning branches before selecting
  4. Chain-of-thought explores more options
Answer

C) Tree-of-thought explores multiple reasoning branches before selecting - Chain-of-thought follows a single reasoning path, while tree-of-thought explores multiple branches and evaluates alternatives before choosing the best approach.

Question 8 (Analyze)

Why might a prompt template with placeholders be preferable to writing prompts from scratch?

  1. Templates are always cheaper
  2. Templates require no customization
  3. Templates ensure consistency and encode best practices
  4. Templates eliminate the need for prompt engineering
Answer

C) Templates ensure consistency and encode best practices - Templates standardize effective prompt patterns, making them reusable, consistent, and incorporating lessons learned from iteration.

Question 9 (Evaluate)

Which scenario would benefit MOST from self-consistency prompting?

  1. Simple factual questions
  2. Creative writing
  3. Complex reasoning with multiple valid approaches
  4. Generating random content
Answer

C) Complex reasoning with multiple valid approaches - Self-consistency generates multiple reasoning paths and selects the most common answer, which is most valuable when problems have multiple solution approaches.

Question 10 (Create)

Design a prompt strategy for a customer service bot that must: (1) maintain a friendly tone, (2) provide accurate product information, (3) escalate complex issues.

  1. Just use zero-shot with "be helpful"
  2. Maximum temperature for creativity
  3. System prompt defining persona and escalation rules, few-shot examples of good responses, structured output for escalation flags
  4. No prompt engineering needed
Answer

C) System prompt defining persona and escalation rules, few-shot examples of good responses, structured output for escalation flags - This combines multiple techniques: system prompt for behavior, few-shot for quality patterns, and structured output for actionable escalation.

Score Interpretation

  • 9-10 correct: Excellent understanding of prompt engineering
  • 7-8 correct: Good grasp, review missed concepts
  • 5-6 correct: Fair understanding, revisit chapter sections
  • Below 5: Re-read Chapter 4 before proceeding

Previous: Quiz 3 | Back to Quizzes | Next: Quiz 5 - Custom GPTs & RAG