Q1. What makes a prompt clear versus ambiguous? Provide examples.
A clear prompt specifies the task, format, and constraints unambiguously.
Ambiguous prompt: "Tell me about dogs."
Clear prompt: "List three interesting facts about Golden Retrievers in bullet points."
Clear prompts leave no room for interpretation – the model knows exactly what you expect.
Ambiguous prompts cause the model to guess, leading to inconsistent or irrelevant answers.
Always ask yourself: Could someone else (without domain knowledge) follow these instructions exactly?
Ambiguous prompt: "Tell me about dogs."
Clear prompt: "List three interesting facts about Golden Retrievers in bullet points."
Clear prompts leave no room for interpretation – the model knows exactly what you expect.
Ambiguous prompts cause the model to guess, leading to inconsistent or irrelevant answers.
Always ask yourself: Could someone else (without domain knowledge) follow these instructions exactly?
Q2. How does ambiguity affect the quality of LLM responses?
Ambiguity leads to:
• Vague or generic answers (e.g., "Dogs are great companions")
• Unexpected output formats (sometimes a list, sometimes a paragraph)
• Missing important details
• Hallucinations (the model fills gaps with incorrect information)
• Inconsistent responses to the same prompt across different runs
For business-critical applications, ambiguity is unacceptable.
Engineered prompts must be precise to ensure reliability and user trust.
• Vague or generic answers (e.g., "Dogs are great companions")
• Unexpected output formats (sometimes a list, sometimes a paragraph)
• Missing important details
• Hallucinations (the model fills gaps with incorrect information)
• Inconsistent responses to the same prompt across different runs
For business-critical applications, ambiguity is unacceptable.
Engineered prompts must be precise to ensure reliability and user trust.
Q3. What are common sources of ambiguity in prompts?
Common sources include:
• Vague verbs: "discuss", "explain", "review" – be specific: "list", "summarize in one sentence", "compare"
• Unclear scope: "recent events" – define: "events from the last 2 years"
• Missing output format: specify if you want bullet points, JSON, a table, or plain text
• Pronouns without clear antecedents: "They said it was good" – who said what?
• Double negatives or complex conditionals
• Assuming the model knows implicit context (e.g., company-specific jargon)
Each of these can be fixed by adding explicit details.
• Vague verbs: "discuss", "explain", "review" – be specific: "list", "summarize in one sentence", "compare"
• Unclear scope: "recent events" – define: "events from the last 2 years"
• Missing output format: specify if you want bullet points, JSON, a table, or plain text
• Pronouns without clear antecedents: "They said it was good" – who said what?
• Double negatives or complex conditionals
• Assuming the model knows implicit context (e.g., company-specific jargon)
Each of these can be fixed by adding explicit details.
Q4. How can you test whether a prompt is clear enough?
Use these techniques:
• Read the prompt out loud to a colleague – if they hesitate, it's not clear.
• Ask the model to repeat the instruction back to you before answering (e.g., "First, restate the task in your own words, then answer.")
• Run the prompt multiple times – if outputs vary wildly, it's ambiguous.
• Add a validation step: ask the model to output in a structured format (JSON) and programmatically check if fields are present.
• A/B test variations: change one word at a time and see if output changes significantly.
Iterative testing is key to removing ambiguity.
• Read the prompt out loud to a colleague – if they hesitate, it's not clear.
• Ask the model to repeat the instruction back to you before answering (e.g., "First, restate the task in your own words, then answer.")
• Run the prompt multiple times – if outputs vary wildly, it's ambiguous.
• Add a validation step: ask the model to output in a structured format (JSON) and programmatically check if fields are present.
• A/B test variations: change one word at a time and see if output changes significantly.
Iterative testing is key to removing ambiguity.
Q5. What is the difference between implicit and explicit instructions? Give an example.
Implicit instruction: the model infers what you want from context but it's not stated.
Example: "Write an email to a customer about a delayed shipment."
The model decides the tone, length, and content, which may not match your needs.
Explicit instruction: you state every requirement clearly.
Example: "Write a short, apologetic email to a customer about a 3-day shipping delay. Keep it professional and offer a 10% discount code. Use no more than 100 words."
Explicit instructions give you control; implicit ones rely on the model's guess.
Example: "Write an email to a customer about a delayed shipment."
The model decides the tone, length, and content, which may not match your needs.
Explicit instruction: you state every requirement clearly.
Example: "Write a short, apologetic email to a customer about a 3-day shipping delay. Keep it professional and offer a 10% discount code. Use no more than 100 words."
Explicit instructions give you control; implicit ones rely on the model's guess.