Prompt Engineering: Getting the Best From Ai

• A. A password that unlocks the model's full capabilities
• B. The input text that instructs the model on what task to perform and how to respond ✓
• C. A training dataset used to fine-tune the model's weights
• D. An error message generated when the model cannot process a request

Why: A prompt is the input text you give the model, including instructions, context, and any examples, that guides the output it produces. It is not a password, a training dataset, or an error message.

• A. Write about dogs.
• B. Tell me something about pets.
• C. Write a 150-word blog introduction about adopting an adult dog, aimed at first-time pet owners and written in a warm, encouraging tone. ✓
• D. Dogs?

Why: Specific prompts that define the task, length, audience, and tone give the model clear constraints, leading to more relevant and useful output. Vague prompts leave too much open to interpretation.

• A. Repeating your question several times to emphasize it
• B. Including background information the model needs to understand the task, such as the purpose, audience, or relevant details ✓
• C. Asking the model to recall its previous training data word for word
• D. Setting the maximum length of the model's response

Why: Context is the background information that helps the model understand your situation, such as who the output is for, why it is needed, and what relevant facts apply. Repetition, recall of training data, and length settings are different things.

• A. It makes the model run faster on the same hardware
• B. It guarantees the content is factually correct
• C. It makes the response easier to read, parse, or integrate into other tools and workflows ✓
• D. It prevents the model from making any errors

Why: Specifying a format shapes the response into a predictable structure, which is valuable for human readability and for programmatic use. It does not affect speed, guarantee accuracy, or eliminate errors.

• A. Zero-shot prompting
• B. Few-shot prompting ✓
• C. Temperature scaling
• D. Token truncation

Why: Few-shot prompting includes a small number of examples in the prompt to demonstrate the expected pattern. Zero-shot prompting gives no examples, while temperature scaling and token truncation are unrelated to example-based instruction.

• A. Summarize this report.
• B. Summarize this report in 3 bullet points for a non-technical manager, focusing on costs and deadlines. ✓
• C. Make this shorter and nicer.
• D. Do something useful with this report.

Why: Clear instructions specify the task, the output format, the audience, and the focus. Option B gives all of these, so the model knows exactly what a good response looks like.

• A. It makes the AI respond more slowly.
• B. It reduces ambiguity so the response is easier to use directly. ✓
• C. It guarantees the information is factually correct.
• D. It forces the AI to use more words.

Why: Stating the format (a table, a list, JSON, a word limit) removes guesswork about structure, so the output can be used without extensive reformatting. It does not guarantee accuracy or length.

• A. Write something about dogs.
• B. Write a few nice sentences about dogs.
• C. Write a 100-word product description for a waterproof dog leash aimed at hikers. ✓
• D. Tell me stuff about dogs that's interesting.

Why: Option C defines the length, content, and audience precisely. Words like 'something,' 'nice,' 'a few,' and 'interesting' leave the model to guess what you mean.

• A. Give up and assume the AI cannot help.
• B. Repeat the exact same prompt several times without changes.
• C. Refine the prompt by adding specific constraints or examples of what you want. ✓
• D. Switch to typing in all capital letters.

Why: Iterating on the prompt by adding missing details, constraints, or examples usually improves results. Repeating an identical prompt rarely changes much, since the model has no new information to work with.

• A. Don't make the answer too long or use jargon.
• B. Don't be confusing.
• C. Don't include any unnecessary details.
• D. Keep the answer under 150 words and explain each term in plain English. ✓

Why: Positive instructions state the desired behavior directly, which is usually easier for a model to follow. Option D tells the model exactly what to do, while the other options only describe what to avoid without specifying the alternative.

• A. It makes the prompt longer, which forces the AI to think for more time
• B. It gives the AI background about the purpose, audience, and constraints, so it can tailor its answer appropriately ✓
• C. It causes the AI to memorize the information permanently for future conversations
• D. It reduces the number of tokens the AI needs to generate a response

Why: AI models produce responses based on the information in the prompt. Context such as the audience, purpose, and constraints narrows the range of reasonable answers, helping the model produce output that fits the user's actual need rather than a generic response.

• A. Write a poem about the ocean.
• B. Write a poem about the ocean in a formal tone, at least 12 lines long, and for a high school audience.
• C. Classify these reviews as positive or negative. Example: 'Great service!' -> Positive. 'Food was cold.' -> Negative. Now classify: 'The room was spotless.' ✓
• D. Please, I really need you to be as helpful as possible and answer carefully.

Why: Few-shot prompting includes a small number of worked examples of the input-output pattern you want before presenting the new case. Option C shows two labeled examples followed by a new input, which lets the model infer the task format and labeling rules.

• A. It guarantees the AI will never make factual errors
• B. It removes the need to explain the task at all
• C. It makes the response structure more consistent and easier to parse or use ✓
• D. It forces the AI to respond in fewer than 50 words

Why: Showing the desired format, such as a specific table layout, JSON structure, or bullet pattern, helps the model reproduce that structure reliably. This is especially useful when the output will be read by another program or compared across many responses. It does not guarantee accuracy or length.

• A. Make it sound better
• B. Write an email
• C. Write an apology email to a customer whose order shipped 5 days late, acknowledging the delay, offering a 10% discount code, and keeping a warm but professional tone ✓
• D. Write something nice

Why: The strongest prompt supplies specifics: the situation, the required content elements, and the tone. These details let the model produce a usable draft on the first attempt instead of guessing at what the user needs.

• A. Only when the user does not know how to describe the task in words
• B. When the desired style, format, or pattern is hard to describe precisely in words ✓
• C. Examples should never be used because they confuse the model
• D. Only for tasks involving mathematics

Why: Some qualities, such as a particular writing voice, a nuanced classification scheme, or an unusual formatting pattern, are easier to show than to explain. Examples communicate these patterns directly, often more effectively than lengthy verbal descriptions.

• A. The role or persona assigned to the AI
• B. An explicit output format instruction, such as requesting JSON or a bulleted list ✓
• C. The temperature setting used by the model
• D. The length of the user's question

Why: Explicit format instructions, such as naming JSON or a specific structure, are the most direct way to control output format. Personas influence tone and perspective, temperature affects randomness, and question length does not dictate structure.

• A. It makes the prompt longer, which always improves quality
• B. It shows the model the exact structure, keys, and style to follow ✓
• C. It forces the model to memorize the example for future conversations
• D. It disables the model's ability to make mistakes

Why: A concrete example demonstrates the expected structure, field names, and style, reducing ambiguity. Models learn the pattern from the example far more reliably than from a verbal description alone.

• A. Please write something nice about the topic
• B. Return only valid JSON with no surrounding text or markdown ✓
• C. Give me your best thoughts when you have time
• D. Try to be creative with how you present the answer

Why: Asking for only valid JSON with no surrounding text or markdown produces output a parser can read directly. Vague or open-ended instructions often add prose or code fences that break automated processing.

• A. The model is malfunctioning and must be reset
• B. The prompt did not explicitly prohibit conversational preamble ✓
• C. JSON format requires an introduction to be valid
• D. The model always adds introductions regardless of instructions

Why: Models default to conversational framing unless told otherwise. Stating that the output should contain only the JSON, with no introduction or explanation, prevents this extra text.

• A. Using a different format name each time to keep answers varied
• B. Specifying the exact field names, data types, and nesting you expect ✓
• C. Asking for the answer in as many formats as possible
• D. Leaving the structure open so the model can improve it

Why: Specifying exact field names, data types, and nesting constrains the output to a predictable schema. This consistency is essential when the output feeds into other systems or is compared across runs.

• A. To make the AI respond faster by shortening its answers
• B. To encourage the AI to show intermediate reasoning steps before giving a final answer ✓
• C. To translate prompts into multiple languages
• D. To limit the AI's responses to a single word

Why: Chain-of-thought prompting asks the model to work through a problem step by step, which often improves accuracy on tasks requiring multi-step reasoning.

• A. Adding the phrase 'Let's think step by step' to the prompt ✓
• B. Writing the prompt in all capital letters
• C. Asking the model to respond with only yes or no
• D. Removing all context from the question

Why: The phrase 'Let's think step by step' is a widely used zero-shot trigger that prompts the model to reason through intermediate steps before answering.

• A. Models cannot read numbers in text
• B. Jumping straight to an answer skips the intermediate reasoning that helps catch errors ✓
• C. Models only work with numbers written as digits
• D. Word problems are always longer than the model's context window

Why: When a model commits to an answer immediately, it has no space to break the problem down, so errors in intermediate calculations go uncorrected.

• A. A prompt that asks for only a few words in response
• B. A prompt that includes a few worked examples showing step-by-step reasoning before the new question ✓
• C. A prompt that is sent to the model only a few times
• D. A prompt that uses a small number of tokens

Why: Few-shot chain-of-thought prompts demonstrate the desired reasoning pattern through example question-and-solution pairs, guiding the model to reason similarly on the new question.

• A. Accepting the answer as correct if the reasoning sounds confident
• B. Checking each intermediate step for logical and factual errors ✓
• C. Asking the model to repeat its final answer three times
• D. Shortening the prompt to reduce the chance of errors

Why: Reasoning can sound convincing while containing errors, so checking each step is essential to confirm that the final answer is trustworthy.

• A. Abandon the conversation and start over with a new chat
• B. Identify specifically what is wrong with the response and refine your prompt accordingly
• C. Accept the response as-is because the AI's first answer is usually the best
• D. Ask the AI to make the response longer without changing anything else

• A. Good, now do something else
• B. Make it better
• C. Rewrite this as a three-bullet summary aimed at a non-technical manager, focusing on cost impacts
• D. Try again

• A. Criteria make the AI respond faster to your prompts
• B. Clear criteria, such as accuracy, tone, length, and relevance, help you pinpoint exactly what to change in your next prompt
• C. Criteria remove the need to read the response carefully
• D. The AI only responds to prompts that include a scoring rubric

• A. Assume it is correct since the AI sounds confident
• B. Ask the AI to cite its sources or explain its reasoning, then verify the claim independently
• C. Delete the claim and never revisit the topic
• D. Ask the AI to repeat the same claim in a different format

• A. Keeping only the final version of the prompt and discarding earlier attempts
• B. Changing one element of the prompt at a time and noting how each change affected the output
• C. Changing the entire prompt, topic, and tone with every attempt
• D. Relying on memory to recall which wording worked earlier

• A. Zero-shot prompting
• B. Few-shot prompting ✓
• C. Chain-of-thought prompting
• D. Role prompting

Why: Few-shot prompting includes several worked examples in the prompt so the model can infer the desired pattern, format, and style before handling the new input.

• A. It makes the response shorter and more concise
• B. It guarantees the answer is factually correct
• C. It helps the model break down multi-step problems, often improving accuracy on reasoning tasks ✓
• D. It prevents the model from using any external information

Why: Chain-of-thought prompting encourages the model to lay out intermediate reasoning steps, which often improves performance on math, logic, and multi-step problems, though it does not guarantee correctness.

• A. Asking the model to list five facts about photosynthesis
• B. Telling the model to act as an experienced nutritionist when answering diet questions ✓
• C. Providing a list of example questions and answers
• D. Specifying that the response must be under 100 words

Why: Role prompting assigns the model a persona or expertise, such as a nutritionist, to shape the tone, knowledge focus, and perspective of its responses.

• A. It makes the model slower but more thorough
• B. It removes the need to check the model's output
• C. It makes responses easier to use and parse, reducing the need for reformatting ✓
• D. It forces the model to use more tokens for no benefit

Why: Explicit format instructions produce consistent, usable output that can be integrated into workflows or processed by other software, saving time on manual reformatting.

• A. Give up and assume the model cannot perform the task
• B. Repeat the exact same prompt until a different result appears
• C. Refine the prompt by adding clearer context, constraints, or examples, then try again ✓
• D. Switch to writing the entire response yourself without further attempts

Why: Iterative refinement, adjusting the prompt with more context, constraints, or examples, is a core prompt engineering practice that typically yields better results than repeating the same prompt unchanged.