Prompt Engineering Fundamentals
Last reviewed
Correct answer: A. The model picking up a pattern you never intended to teach it
Explanation
The principle — A model generalises from whatever your examples have in common. If they share something accidental, it learns that too.
Why the key is correct — Anthropic's guidance is to make examples diverse enough that the model does not pick up unintended patterns, and to cover edge cases. Three examples that all happen to be one sentence long teach "answer in one sentence" whether you meant it or not.
Why the others are wrong — Varied examples cost no more room than repetitive ones. Inconsistent formatting is a real but separate fault — you want varied content and identical structure. And a model will attempt a task with no matching example at all.
Remember this — Vary the content of your examples. Keep their shape identical.
Sources — Anthropic, Prompting best practices.
Sources
“Diverse: Cover edge cases and vary enough that Claude doesn't pick up unintended patterns.”
Practise 10 questions on this topic
Take Prompt Engineering Fundamentals — Timed Test 1 (10 questions) — scored instantly, explanation for every question, no login.