CCAR-P Advanced Agentic Architecture Practice Question
A team wants an agent to answer questions about a large internal corpus. They notice the agent invents details when the retrieved chunks are only loosely related to the question. Which change most directly reduces fabricated answers grounded in weak evidence?
⚠ Common exam trap
The trap here is treating every hallucination as a retrieval problem and reaching for embeddings or more chunks, when the described behavior is generation overreach best fixed by grounding instructions.
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
✓
Instruct the agent in the system prompt to answer only from provided documents and to state when evidence is insufficient.
The reported failure is generation overreach under weak retrieval, so the highest-leverage fix is a grounding instruction that scopes answers to supplied documents and explicitly permits declining. This directly constrains what the model may assert and gives it a safe alternative to invention, independent of any retrieval tuning that might follow.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Instruct the agent in the system prompt to answer only from provided documents and to state when evidence is insufficient.
Why this is correct
Grounding instructions tie generation to the supplied context and give the model an explicit escape hatch when retrieval is weak. By authorizing an 'insufficient evidence' response, you remove the pressure to fabricate a plausible answer. This is the most direct lever because it changes behavior at generation time regardless of retrieval quality, and it composes with retrieval improvements later.
- ✗
Raise the temperature setting so the model explores more diverse phrasings of the answer.
Why it's wrong here
Higher temperature increases randomness, which magnifies hallucination rather than reducing it. Diversity in phrasing is irrelevant when the core problem is unsupported claims. For factual grounding tasks you generally want lower temperature, and no sampling change can substitute for instructing the model to rely on retrieved evidence and to decline when that evidence is inadequate.
- ✗
Increase the number of retrieved chunks returned by the vector search to fifty per query.
Why it's wrong here
Dumping more loosely related chunks adds noise and can bury the relevant passage, giving the model more weak evidence to over-generalize from. Recall is not the issue described; precision and grounding are. More context also consumes the window and can dilute attention, so this change can worsen fabrication even though it feels like it improves coverage.
- ✗
Switch the embedding model to one with a larger vector dimension for finer similarity scoring.
Why it's wrong here
A larger embedding dimension may sharpen ranking, but the failure occurs after retrieval, when the model fills gaps with invented detail. Better similarity does not stop the model from answering beyond what the chunks support. The described symptom is a generation-grounding problem, so changing the embedding geometry addresses retrieval quality, not the tendency to fabricate under weak evidence.
About these practice questions
Courseiva writes every CCAR-P question from scratch — 262 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. 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 CCAR-P 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 CCAR-P exam.