CCDV-F Prompt and Context Engineering Practice Question
Which THREE strategies are effective for reducing hallucinations when Claude is asked to answer questions based on a large provided context?
⚠ Common exam trap
Candidates frequently rely on vague phrases like 'be accurate,' ignoring the necessity of explicit verification loops like extraction steps and 'I don't know' permissions to curb hallucinations.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Instructing the model to say 'I don't know' if the answer is not in the text.
Reducing hallucinations requires a combination of structural constraints and behavioral guidance. By allowing the model to express uncertainty and encouraging it to ground its answers in direct quotes, developers create a verification loop. These techniques ensure the model prioritizes accuracy over the helpfulness of providing an answer even when the information is missing.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Instructing the model to say 'I don't know' if the answer is not in the text.
Why this is correct
Giving Claude an 'out' is one of the most effective ways to prevent it from making up information. By explicitly permitting the model to admit a lack of information, you reduce the pressure for it to be helpful at the expense of being truthful, which is a common cause of hallucinations.
- ✗
Setting the temperature to 0.7 to allow for more creative synthesis of the facts.
Why it's wrong here
Higher temperature increases the variability of the output, which directly increases the risk of hallucinations. For fact-based question answering, a temperature of 0.0 is almost always preferred to ensure that the model stays as close to the provided text as possible without introducing creative fabrications.
- ✓
Asking the model to provide direct quotes from the text to support its answer.
Why this is correct
Requiring citations or direct quotes forces the model to locate and verify the information within the provided context before answering. This grounding mechanism makes it much harder for the model to hallucinate because it must find a literal string match to satisfy the instruction, increasing overall reliability.
- ✓
Using a two-step process: first extract relevant snippets, then answer the question.
Why this is correct
Breaking the task into two steps—retrieval followed by synthesis—improves focus. In the first step, Claude identifies the exact parts of the context that are relevant, which reduces the amount of noise it has to deal with when finally formulating the answer, significantly lowering the chance of error.
- ✗
Increasing the frequency_penalty to prevent the model from repeating context words.
Why it's wrong here
Frequency penalties discourage the repetition of tokens, which is actually counter-productive for grounded question answering. In this scenario, you want the model to use the exact words from the source text. A high frequency penalty might force the model to use synonyms or paraphrases, which can introduce inaccuracies.
About these practice questions
One of 257 original CCDV-F practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCDV-F practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCDV-F exam.