Final check — Prompt Engineering & Generative AI at Work

Answer the 10 questions. You need 80% or more to pass. You can retry as many times as you like.

Question 1What is the basic unit of text an LLM reads and predicts, one step at a time?A tokenA pixelA database rowA web page

Question 2Which term describes an LLM producing fluent, confident text that is factually wrong?CompressionEncryptionIndexingHallucination

Question 3A good structured prompt typically specifies the role, task, context, format and what else?The model's training dateExplicit constraintsThe GPU typeThe server location

Question 4You want the model to imitate a specific output style by giving it a few input→output examples. Which pattern is this?Zero-shot promptingFine-tuningFew-shot promptingTokenization

Question 5Asking the model to "think step by step" before answering is best described as which technique?Retrieval-augmented generationChain-of-thought promptingEmbedding searchTemperature scaling

Question 6For factual extraction or code where you need consistent, reproducible answers, the temperature should be set to:As high as possible (near 1)Exactly 0.5 alwaysAs low as possible (near 0)It has no effect

Question 7Where should persistent rules like persona, tone and forbidden actions usually be placed?In the system promptIn the temperature settingIn the file nameNowhere — it cannot be controlled

Question 8A chatbot must answer from a company knowledge base that changes every week. What is usually the best approach?Train a new model from scratch weeklyRaise the temperature to 1.0Remove the system promptRetrieval-augmented generation (RAG)

Question 9A user crafts hidden text in a document so the model ignores its original instructions. This security risk is called:OverfittingPrompt injectionTokenizationCaching

Question 10Which practice best reduces the risk of leaking confidential information when using a public AI tool at work?Paste full customer records to get better answersIncrease the max tokensDisable the system promptNever enter secrets or confidential data, and follow the AI-use policy

Final check — Prompt Engineering & Generative AI at Work

Already paid?