Introduction-to-LLM-and-RAG

Lab: Exploring Prompt Engineering Techniques

Objectives:

• Understand the principles of prompt engineering.
• Discover and practice techniques for crafting effective prompts.
• Observe the impact of prompt variations on generated responses.

---

General Instructions:

1. Read each section carefully.
2. Follow the steps in the activities and record your observations.
3. Test your prompts in a text-generation tool powered by an LLM (e.g., Copilot).
4. Compare the responses obtained to better understand the effects of adjustments in your prompts.

---

Activity 1: Define a Personality and an Audience

Objective:

Explore how personality and target audience influence the model’s responses.

Steps:

1. Write a prompt where the model acts as an art history expert.
2. Test these variations:
   • A pedagogical expert.
   • A popularizer.
   • A formal lecturer.

Sample Prompt:

“You are an expert in art history. Explain the Italian Renaissance to a teenage audience.”

Observation Questions:

• How do tone and structure change across variations?
• Which variation seems most suitable for a teenage audience? Why?

---

Activity 2: Specify Output Format

Objective:

Learn to guide the model in structuring its responses.

Steps:

1. Ask the model to provide “5 steps to resolve a workplace conflict.”
2. Test the following variations:
   • Present the steps as a numbered list.
   • Explain the steps in a narrative paragraph.
   • Add a concluding summary of the key points.

Sample Prompt:

“Provide 5 steps to resolve a workplace conflict in a concise list format.”

Observation Questions:

• Does the model follow the formatting instructions?
• Which format do you find the clearest and most useful? Why?

---

Activity 3: Targeting the Action

Objective:

Learn to direct the model’s responses toward a specific goal.

Steps:

1. Ask the model to explain “supervised learning.”
2. Test the following variations:
   • Add a request for a concrete example.
   • Limit the response to under 50 words.
   • Tailor the explanation for high school students.

Sample Prompt:

“Explain supervised learning using a simple analogy for high school students.”

Observation Questions:

• Did the model follow your specific instructions?
• Which variation produced the most accessible and comprehensible response?

---

Activity 4: Handling Specific Scenarios

Objective:

Anticipate and specify how the model should respond in specific situations.

Steps:

1. Ask the model to recommend books on creativity.
2. Add clauses to handle the following cases:
   • If the model doesn’t know of any books, suggest blogs or videos.
   • If the request concerns a child, recommend age-appropriate books.

Sample Prompt:

“Recommend books on creativity. If none are available, suggest blogs or online videos. If the request is for a child, recommend age-appropriate materials.”

Observation Questions:

• Does the model correctly address the scenarios you specified?
• What additional instructions could improve the response?

---

Conclusion

1. Quick Summary:
   • Did you observe significant differences in responses based on prompt variations?
   • Which techniques do you find most useful? Why?
2. Further Applications:
   • Consider other situations where these techniques could be applied, such as drafting emails, creating educational content, or solving specific problems.

---

Optional: Extra Exercise

Reformulate this initial, vague prompt:

“Explain artificial intelligence.”

To make it more precise and effective, applying the techniques learned in this lab.

You may use Copilot or the self-training platform for prompt engineering Training.