Prompt Engineering

Prompt engineering is the practice of designing and refining prompts (inputs) to optimize the
responses generated by AI language models like GPT.

Key Aspects of Prompt Engineering:

1. Instruction Clarity: Ensuring that the prompt clearly communicates the task to the
model. For example, instead of saying, "Explain photosynthesis," you might say,
"Explain the process of photosynthesis in simple terms suitable for a 10-year-old."
2. Contextual Information: Providing necessary context within the prompt to help the
model understand the scenario. For example, specifying the format or tone: "Write a
formal letter to a potential business partner."
3. Examples and Constraints: Including examples or specifying constraints can help
the model focus on what is required. For example, "Generate a list of three potential
titles for a science fiction novel set on Mars."
4. Iterative Refinement: Often, prompts need to be refined through trial and error. This
iterative process helps in identifying the best phrasing or structure that yields the
desired response from the model.
5. Task-Specific Prompts: Tailoring prompts for specific tasks such as summarization,
translation, question answering, or code generation to get the best possible output for
that particular use case.

Importance:

Prompt engineering is crucial because the quality of the output generated by an AI model is
heavily dependent on the input prompt. Well-engineered prompts can lead to more useful,
accurate, and contextually appropriate outputs, making AI models more effective and reliable
for various applications.

PROMPTING LLM

Large Language Model (LLM) involves crafting a clear and specific input to guide the
model in generating the desired response. Effective prompting is crucial because it directly
influences the quality and relevance of the output. Here’s how to approach prompting an
LLM:

1. Understand the LLM’s Capabilities

• General Knowledge: LLMs are trained on vast datasets and can provide information
on a wide range of topics.
• Language Understanding: They can comprehend and generate text in multiple
languages and styles.
• Contextual Awareness: LLMs can maintain context over a series of interactions but
may struggle with long or complex dialogues.
2. Types of Prompts

• Informative Prompts: Ask the LLM to explain or describe something.

o Example: “Explain the process of photosynthesis.”
• Instructional Prompts: Direct the LLM to perform a task or generate specific
content.
o Example: “Write a Python script to sort a list of numbers.”
• Creative Prompts: Encourage the LLM to generate creative content.
o Example: “Write a short story about a time traveler who visits ancient Egypt.”
• Interactive Prompts: Engage the LLM in a conversation or decision-making process.
o Example: “What are the pros and cons of adopting renewable energy
sources?”

3. Best Practices for Effective Prompting

• Clarity: Make your prompt as clear and unambiguous as possible.

o Example: Instead of “Tell me about space,” try “Describe the characteristics
of a black hole.”
• Specificity: Provide specific instructions or questions to narrow down the response.
o Example: “List the top three benefits of exercise for cardiovascular health.”
• Context: Provide sufficient context if the prompt requires understanding a particular
scenario.
o Example: “Given a scenario where a company is facing a data breach, what
immediate steps should be taken?”
• Open-Ended vs. Closed-Ended: Decide whether you want an open-ended response
or a specific answer.
o Open-Ended: “How can AI transform the healthcare industry?”
o Closed-Ended: “Is AI likely to improve patient outcomes in healthcare? Yes
or no.”
• Iterative Prompting: For complex tasks, break down the prompt into smaller,
manageable parts.
o Example: First, “Explain the importance of data privacy.” Then, “How can
companies ensure data privacy?”

4. Advanced Techniques

• Few-Shot Prompting: Provide a few examples of the desired output within the
prompt to guide the model.
o Example: “Translate the following phrases to French: ‘Hello’ becomes
‘Bonjour.’ ‘Goodbye’ becomes...”
• Chain of Thought Prompting: Ask the LLM to explain its reasoning process before
giving the final answer.
o Example: “How would you calculate the area of a circle? First, explain the
formula, then calculate for a radius of 5.”
• Role-Playing Prompts: Assume a role to get responses tailored to a specific
perspective.
o Example: “As a legal advisor, how would you handle a contract dispute
between two companies?”
5. Evaluating and Refining Prompts

• Test Different Variations: Experiment with different phrasings to see which yields
the best response.
• Feedback Loop: Use the output to refine the prompt iteratively.
• Avoid Ambiguity: If the response is off-target, check if the prompt could be
interpreted in multiple ways.

Example of Prompting an LLM

Scenario: You want the LLM to write a summary of a complex article about quantum
computing.

• Prompt: “Summarize the key points of an article on quantum computing, focusing on

how qubits differ from classical bits, the potential applications of quantum computers,
and the current challenges in developing quantum technology.”
Prompt Formatting
Prompt formatting refers to the structure and style used when crafting prompts to interact
with a Large Language Model (LLM). Proper formatting ensures clarity, enhances the
effectiveness of the prompt, and guides the LLM toward generating the desired response.

Key aspects of prompt formatting:

1. Basic Structure

• Instruction or Query: Begin with a clear command, question, or instruction.

• Context or Background: Provide necessary background information or context.
• Expected Output: Clearly state what you expect the model to produce.

Example:

Instruction: Write a summary of the following text.

Context: The text is an article on the benefits of renewable energy.
Text: [Insert article here]
Expected Output: A 3-4 sentence summary highlighting the key points.

2. Labelling Sections

• Use labels to differentiate between different parts of the prompt, such as

“Instruction,” “Context,” “Text,” “Output,” etc. This makes it easier for both the LLM
and human reviewers to understand the structure.

Example:

[Instruction]: Explain the concept of blockchain technology.

[Context]: The explanation should be suitable for someone with a basic
understanding of technology but no prior knowledge of blockchain.
[Expected Output]: A concise explanation in 2-3 paragraphs.

3. Using Delimiters

• Delimiters like brackets [ ], braces { }, or quotes " " can help clarify specific
sections of the prompt, especially when dealing with input data, expected outputs, or
examples.

Example:

[Instruction]: Translate the following English sentence to Spanish.

[Text]: "The quick brown fox jumps over the lazy dog."
[Expected Output]: Provide the translation as a complete sentence in
Spanish.
4. Bullet Points and Lists

• For prompts that require multiple inputs or steps, use bullet points or numbered lists.
This format is especially helpful for tasks involving multiple requirements or
instructions.

Example:

[Instruction]: List the following:

1. The main differences between renewable and non-renewable energy sources.
2. Three examples of renewable energy.
3. The environmental benefits of using renewable energy.

5. Examples in the Prompt

• Providing examples within the prompt can help the LLM understand the format and
style you expect. This is particularly useful for tasks like translation, code generation,
or creative writing.

Example:

[Instruction]: Generate a metaphor for each of the following emotions:

[Example]: Happiness: "Happiness is a warm summer breeze."
[Emotion 1]: Sadness:
[Emotion 2]: Fear:

6. Avoiding Ambiguity

• Be specific in your wording to avoid multiple interpretations. Ambiguity in the

prompt can lead to unexpected results.

Example:

[Ambiguous Prompt]: Write a story about a bank.

[Clear Prompt]: Write a short story about a bank heist that goes wrong.

7. Length Management

• Keep prompts concise, especially when interacting with models that have token
limits. However, include enough information to avoid vague or incomplete responses.

Example:

[Instruction]: In 2-3 sentences, summarize the impact of the Industrial

Revolution on urbanization.

8. Question Formatting

• When asking questions, be direct and ensure the question is fully formed. Use proper
punctuation and specify the type of answer expected (e.g., short answer, multiple
choice).
Example:

[Instruction]: What were the main causes of World War I?

[Expected Output]: A short paragraph listing the key causes.

9. Multimodal Prompts

• If your LLM supports multimodal input (text, images, etc.), clearly separate and label
different types of input.

Example:

[Instruction]: Analyse the image and describe the emotions it conveys.

[Image]: [Insert image here]
[Expected Output]: A description of the emotions in 2-3 sentences.

10. Iterative Prompts

• If the task is complex, use iterative prompting by breaking down the task into smaller,
sequential prompts.

Example:

[Instruction 1]: Explain the concept of neural networks.

[Instruction 2]: Now, explain how neural networks are used in image
recognition.
[Instruction 3]: Finally, discuss the challenges of training neural
networks.

Example of a Well-Formatted Prompt

[Instruction]: Write a short, persuasive essay on why renewable energy is
essential for future sustainability.
[Context]: Assume the reader has some basic knowledge of energy sources but
is skeptical about the transition to renewable energy.
[Requirements]:
- Start with a compelling introduction.
- Include at least three key arguments supporting renewable energy.
- Conclude with a strong call to action.
[Expected Output]: An essay of approximately 300 words.
Elements of a prompt
The effectiveness of a prompt largely depends on its elements, which guide the model's
response.

Key elements of a well-structured prompt:

1. Instruction or Task

• Definition: The core directive that tells the model what to do.
• Purpose: This element defines the action you want the LLM to take, whether it’s
generating text, answering a question, summarizing information, or performing
another task.
• Example: "Summarize the following text."

2. Context or Background

• Definition: Provides the necessary background or setting for the task.

• Purpose: Helps the model understand the scenario or the specific angle from which
the task should be approached.
• Example: "The text is an article about the benefits of renewable energy."

3. Input Data

• Definition: The data or content that the model should process or analyze.
• Purpose: Specifies the information the LLM needs to work with, such as a text
passage, code snippet, or list of items.
• Example: "The quick brown fox jumps over the lazy dog."

4. Output Specification

• Definition: Describes what the final result should look like.

• Purpose: Guides the model in formatting its response, whether it's a short answer, a
detailed explanation, or a creative piece.
• Example: "Provide a 2-3 sentence summary."

5. Examples or References

• Definition: Sample outputs that demonstrate the desired response.

• Purpose: Helps the model understand the format, style, or content expected,
particularly useful in complex or creative tasks.
• Example: "Translate the following: 'Hello' becomes 'Hola.'"

6. Constraints or Guidelines

• Definition: Specific rules or boundaries for the task.

• Purpose: Limits the scope of the response, defining things like word count, tone, or
content requirements.
• Example: "The summary should not exceed 100 words."
7. Tone and Style Instructions

• Definition: Describes the desired tone, style, or register of the output.

• Purpose: Ensures the response matches the intended audience or use case, whether
formal, informal, persuasive, etc.
• Example: "Write the essay in a formal tone suitable for academic readers."

8. Target Audience

• Definition: The intended readers or recipients of the output.

• Purpose: Helps tailor the response to be appropriate for a specific group, such as
beginners, experts, children, etc.
• Example: "Explain the concept of blockchain in simple terms suitable for high school
students."

9. Feedback or Iteration Points

• Definition: Steps or instructions for refining the output based on initial responses.
• Purpose: Guides the model through a process of refining or improving its response
through iterations.
• Example: "If the first explanation is too technical, simplify it further."

10. Conditional Instructions

• Definition: Provides different paths based on certain conditions or responses.

• Purpose: Allows for dynamic responses based on how the task unfolds, enabling
complex interactions.
• Example: "If the user asks for more details, provide a detailed explanation.
Otherwise, give a brief summary."

Example Prompt Incorporating All Elements

[Instruction]: Write a persuasive essay on why renewable energy is crucial
for future sustainability.
[Context]: Assume the reader is skeptical about the transition to renewable
energy.
[Input Data]: Consider recent advancements in solar and wind energy
technologies.
[Output Specification]: The essay should be approximately 300 words long.
[Examples]: "For example, the adoption of solar energy has increased by 30%
in the last decade."
[Constraints]: Use at least three sources to support your arguments.
[Tone]: The essay should be formal and authoritative.
[Target Audience]: The essay is intended for policymakers and environmental
experts.
[Feedback]: After writing, review the essay for clarity and coherence.
[Conditional Instructions]: If the essay exceeds 300 words, summarize the
main points in a final paragraph.
General Tips for Designing Prompts
Designing effective prompts for interacting with a Large Language Model (LLM) is crucial
for obtaining accurate and relevant responses.

1. Be Clear and Specific

• Tip: Make your prompt as clear and unambiguous as possible. Avoid vague or overly
broad questions.
• Example: Instead of asking, “Tell me about energy,” specify, “Explain the difference
between renewable and non-renewable energy.”

2. Define the Task Explicitly

• Tip: Clearly state what you want the LLM to do. Whether it’s summarizing, listing,
comparing, or generating content, the task should be explicit.
• Example: “List three benefits of renewable energy” is more effective than simply
saying, “Talk about renewable energy.”

3. Provide Context

• Tip: Give the necessary background information or context to help the model
understand the scenario.
• Example: “Assume the reader is a high school student learning about climate change.
Explain the greenhouse effect.”

4. Set Output Expectations

• Tip: Specify the desired format, length, or style of the output. This helps in getting a
response that matches your needs.
• Example: “Write a 200-word summary of the provided article.”

5. Use Examples

• Tip: Providing examples can guide the model in understanding the desired output
format or style, especially in complex tasks.
• Example: “Translate the sentence, ‘Hello, how are you?’ into Spanish. Example:
‘Good morning’ becomes ‘Buenos días.’”

6. Start Simple, Then Iterate

• Tip: If the task is complex, break it down into smaller, simpler steps. Use iterative
prompting to refine the response.
• Example: First, ask, “What is blockchain?” and then follow up with, “Explain how
blockchain is used in cryptocurrencies.”
7. Avoid Ambiguity

• Tip: Avoid prompts that can be interpreted in multiple ways. Clarify any potentially
confusing language.
• Example: Instead of “What is Python?” specify, “What is Python, the programming
language?”

8. Be Concise

• Tip: While providing enough detail is important, being concise helps in keeping the
model focused on the task at hand.
• Example: “Summarize the main points of this article in 3-4 sentences.”

9. Consider the Model’s Limitations

• Tip: Keep in mind the token limit, context retention, and potential biases of the
model. Avoid overly long or complex prompts that might exceed the model’s
capabilities.
• Example: If discussing a complex topic, split it into smaller prompts instead of one
long one.

10. Test and Refine Your Prompts

• Tip: Experiment with different phrasings or structures to see which yields the best
results. Refine the prompt based on the output.
• Example: If the initial response is off-target, try rephrasing or adding more context.

11. Use Role-Playing for Specific Perspectives

• Tip: Frame the prompt from a particular perspective or role to get a tailored response.
• Example: “As a legal advisor, how would you advise a client in a breach of contract
case?”

12. Utilize Conditional Instructions

• Tip: Provide instructions on how the model should proceed based on certain
conditions or responses.
• Example: “If the user asks for more details, provide a longer explanation. Otherwise,
keep it brief.”

13. Set Constraints

• Tip: Define any constraints, such as word count, format, or style, to ensure the output
meets your requirements.
• Example: “Write a formal letter of recommendation in less than 150 words.”

14. Leverage Chain of Thought Prompting

 • Tip: Ask the model to explain its reasoning step-by-step before providing a final
answer to complex questions.
• Example: “Explain the steps you would take to solve this math problem: 3x + 5 =
20.”

15. Consider the Target Audience

• Tip: Tailor the prompt to generate responses appropriate for the intended audience,
whether it’s beginners, experts, children, etc.
• Example: “Explain quantum computing in simple terms that a high school student
could understand.”

Specificity is key when designing prompts for Large Language Models (LLMs) because it
directly influences the accuracy, relevance, and usefulness of the model's output. Here’s how
you can incorporate specificity into your prompts:

1. Clearly Define the Task

• Tip: Be explicit about what you want the model to do. Avoid general instructions that
could be interpreted in multiple ways.
• Specific Example: Instead of saying, “Describe AI,” use “Describe the difference
between supervised and unsupervised learning in AI.”

2. Limit the Scope

• Tip: Narrow down the topic to avoid broad or overly general responses. Define
exactly what aspect of the subject you want the model to focus on.
• Specific Example: Instead of “Explain climate change,” try “Explain how greenhouse
gases contribute to global warming.”

3. Specify the Output Format

• Tip: Indicate the format in which you want the information presented—whether it’s a
list, a paragraph, a step-by-step guide, or a comparison.
• Specific Example: “List three key features of Python programming in bullet points.”

4. Set Clear Length Expectations

• Tip: Provide clear instructions on the length of the response, whether it’s in terms of
word count, sentences, or paragraphs.
• Specific Example: “Summarize the article in 150-200 words.”

5. Target the Audience

• Tip: Specify who the response is intended for, which can influence the tone,
complexity, and content of the response.
• Specific Example: “Explain the concept of blockchain to a high school student.”
6. Contextualize the Prompt

• Tip: Provide any necessary background or context that the model needs to understand
to produce a relevant response.
• Specific Example: “Given that the company’s revenue has decreased by 20% over
the last year, suggest three strategies to improve sales.”

7. Use Direct and Specific Language

• Tip: Avoid vague terms and use precise language to guide the model towards the
exact response you need.
• Specific Example: Instead of “What are the benefits of exercise?” use “What are the
cardiovascular benefits of regular aerobic exercise?”

8. Include Examples

• Tip: Provide examples within the prompt to clarify the expected response, especially
if the task is complex or abstract.
• Specific Example: “Write a metaphor for sadness. For example, ‘Sadness is a dark
cloud that lingers in the sky.’ Now, create your own metaphor.”

9. Incorporate Conditional Statements

• Tip: Use “if-then” statements to guide the model’s response based on different
scenarios or to handle multiple parts of a task.
• Specific Example: “If the customer asks about the return policy, explain the standard
procedure. If they ask about warranty coverage, explain that instead.”

10. Break Down Complex Tasks

• Tip: If the task is complicated, break it down into smaller, more manageable parts and
specify each step clearly.
• Specific Example: “First, define what a neural network is. Then, explain how neural
networks are used in image recognition. Finally, discuss the challenges of training
neural networks.”

11. Avoid Ambiguity

• Tip: Remove any potential for multiple interpretations by being as precise as possible.
• Specific Example: Instead of “Write about renewable energy,” use “Write about the
advantages of solar power over fossil fuels in energy production.”

12. Ask Specific Questions

• Tip: Formulate questions that are direct and have a clear focus to avoid broad or
unfocused answers.
• Specific Example: “What were the three main causes of the 2008 financial crisis?”

13. Guide the Tone and Style

 • Tip: Indicate the tone, style, or formality level you want the response to have, based
on the audience or context.
• Specific Example: “Write a formal email to a potential client introducing your
company’s services.”

14. Provide Clear Instructions for Complex Queries

• Tip: For tasks requiring specific methodologies or frameworks, guide the model on
how to approach the query.
• Specific Example: “Using the SWOT analysis framework, evaluate the strengths and
weaknesses of Company X in the current market.”

15. Iterate and Refine

• Tip: If the initial prompt doesn’t yield the desired result, refine the prompt with more
specific instructions.
• Specific Example: “If the summary is too long, shorten it to include only the key
points about the financial impact.”
Examples of prompt:
1. Text Summarization: "Summarize the following article in 3-4 sentences."
o Article: "The Industrial Revolution began in the late 18th century and had a
profound impact on society. It introduced new manufacturing processes,
leading to significant economic and social changes. Cities grew rapidly as
people moved from rural areas to urban centers in search of work, which also
led to the development of new technologies and infrastructure. However, it
also resulted in challenging living conditions and environmental issues."

2. Information Extraction: "Extract the key details from this text, including names,
dates, and locations."
o Text: "Marie Curie was born on November 7, 1867, in Warsaw, Poland. She
conducted much of her groundbreaking research on radioactivity at the
University of Paris."

3. Question Answering: "What is the capital of France?"

o Answer: "The capital of France is Paris."

4. Text Classification: "Classify the following text as either positive, neutral, or

negative sentiment."
o Text: "I really enjoyed the movie; it was fantastic and kept me entertained
throughout."
o Classification: Positive

5. Conversation: "Engage in a conversation with me about the benefits of exercise."

o Response: "Exercise has numerous benefits, including improving
cardiovascular health, boosting mood, and increasing overall energy levels. It
can also help with weight management and reduce the risk of chronic diseases.
What kind of exercise do you enjoy?"

6. Code Generation: "Generate a Python function that calculates the factorial of a

number."
o Code:

def factorial(n):
if n == 0 or n == 1:
return 1
else:
return n * factorial(n - 1)
These prompts demonstrate different ways in which AI can be used to process and interact
with text.