Skip to main content

Final Exam - Prompt Engineering Complete

This comprehensive exam covers all 10 modules with 50 questions. You should complete all module quizzes before attempting the final exam.

Exam Guidelines
  • Time: No limit (recommended 60-90 minutes)
  • Passing score: 70% (35/50)
  • Expert level: 90%+ (45/50)

Section A: Foundations (Questions 1-10)​

Question 1​

Which component of a prompt specifies the desired structure of the AI's response?

  • A) Instruction
  • B) Context
  • C) Input
  • D) Output Format

Answer: D - Output Format specifies how the response should be structured.

Question 2​

What is the primary benefit of few-shot prompting over zero-shot?

  • A) Lower token cost
  • B) Higher consistency through demonstrated patterns
  • C) Faster responses
  • D) Simpler prompts

Answer: B - Few-shot provides examples that help ensure consistent, pattern-following outputs.

Question 3-10 Answers​

Q3: B (CoT improves reasoning), Q4: A (Tokens are ~4 chars), Q5: C (Context window limits total text), Q6: B (Personas direct expertise/tone), Q7: A (CREATE has 6 components), Q8: C (Claude excels with XML), Q9: B (Temperature controls randomness), Q10: A (Zero-shot = no examples)

Section B: Techniques (Questions 11-25)​

Question 11​

When should you use Chain-of-Thought prompting?

  • A) Simple factual questions
  • B) Complex reasoning and math problems
  • C) Creative writing
  • D) Quick summaries

Answer: B - CoT is most effective for complex reasoning, math, and multi-step problems.

Question 12​

What is prompt injection?

  • A) Adding more examples
  • B) Malicious input that manipulates AI behavior
  • C) Improving prompt quality
  • D) A formatting technique

Answer: B - Prompt injection is when malicious input attempts to override AI instructions.

Question 13-25 Answers​

Q13: C (RTF = Role, Task, Format), Q14: A (Variables enable template reuse), Q15: B (Validate between chain steps), Q16: C (RACE for business prompts), Q17: A (Specify language for code), Q18: B (Diverse examples reduce bias), Q19: C (Human oversight required), Q20: A (Iterate one change at a time), Q21: B (Test with diverse cases), Q22: C (Neutral wording reduces bias), Q23: A (Strong instruction hierarchy), Q24: B (Document prompt versions), Q25: C (Measure with metrics)

Section C: Models and Applications (Questions 26-40)​

Question 26​

Which model family is known for the largest context window (200K tokens)?

  • A) GPT-4
  • B) Llama
  • C) Claude 3
  • D) Mistral

Answer: C - Claude 3 models support up to 200K tokens of context.

Question 27​

What is RAG?

  • A) A type of prompt pattern
  • B) Retrieval-Augmented Generation
  • C) A rating system
  • D) Random Answer Generation

Answer: B - RAG combines AI with information retrieval from external sources.

Question 28-40 Answers​

Q28: A (Open source for privacy), Q29: C (System prompts for GPT), Q30: B (Llama format INST), Q31: A (Multimodal = text+images+audio), Q32: C (Agents take actions), Q33: B (Calculate ROI for business), Q34: A (Chain for complex workflows), Q35: C (Error handling in pipelines), Q36: B (Marketing: audience+platform), Q37: A (Code: specify inputs/outputs), Q38: C (Analysis: structured questions), Q39: B (Document processing pipeline), Q40: A (API integration patterns)

Section D: Ethics and Best Practices (Questions 41-50)​

Question 41​

What is the most important principle for responsible AI use?

  • A) Maximum automation
  • B) Transparency about AI involvement
  • C) Avoiding all AI use
  • D) Only using free models

Answer: B - Transparency about when and how AI is used is fundamental to responsible AI practices.

Question 42​

How should you handle sensitive/personal data in prompts?

  • A) Include everything for context
  • B) Avoid sharing sensitive data, protect privacy
  • C) Encrypt it in the prompt
  • D) Only use it with GPT-4

Answer: B - Protect privacy by not sharing sensitive personal data in prompts.

Question 43-50 Answers​

Q43: C (Verify claims for accuracy), Q44: A (Human oversight for decisions), Q45: B (Debiasing techniques), Q46: C (Input validation for security), Q47: A (Separate trusted/untrusted input), Q48: B (Output filtering), Q49: C (Regular security testing), Q50: A (Accountability with humans)

Scoring​

Count your correct answers:

ScoreLevelCertificate
45-50ExpertWith Distinction
40-44AdvancedStandard
35-39ProficientStandard
30-34DevelopingReview Recommended
Below 30BeginningPlease Review Course

Congratulations!​

If you've passed the final exam, you've demonstrated comprehensive knowledge of prompt engineering!

Next Steps​

  1. Build a portfolio of prompts you've created
  2. Practice daily with different AI models
  3. Stay current with new developments
  4. Share knowledge with others
  5. Apply skills in real projects

Resources​

← Back to Course Home