mediumMultiple Choice
Generative AI Leader Practice Question: A company has a Gemini-based application that…
A company has a Gemini-based application that sometimes produces factually incorrect answers. They want to improve accuracy without retraining the model. Which technique should they implement?
⚠ Common exam trap
This question tests the misconception that prompt engineering alone can fix factual accuracy issues, when in reality, without external knowledge grounding, the model remains reliant on its parametric memory which is prone to hallucination.
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
✓
Implement Retrieval-Augmented Generation (RAG) with a curated knowledge base
Retrieval-Augmented Generation (RAG) is the correct technique because it grounds the model's responses in a curated, external knowledge base, providing factual context that reduces hallucinations without modifying the model's weights. This directly addresses the need for improved accuracy in a Gemini-based application while avoiding costly retraining.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Reduce the top-k value to 10
Why it's wrong here
Lowering top-k to 10 restricts sampling to the ten highest-probability tokens, which can sharpen fluency but does nothing to ground answers in verified facts, so hallucinations persist. It is tempting because top-k tuning shapes output diversity, yet accuracy requires retrieval-augmented generation supplying authoritative context.
- ✓
Implement Retrieval-Augmented Generation (RAG) with a curated knowledge base
Why this is correct
Retrieval-Augmented Generation grounds responses in a curated knowledge base, retrieving relevant documents at inference time so the model cites verified content rather than relying on parametric memory. This directly addresses the factual inaccuracy constraint without retraining, since the Gemini weights stay frozen while accuracy improves through external retrieval.
- ✗
Use prompt engineering to instruct the model to 'be more accurate'
Why it's wrong here
Instructing the model to 'be more accurate' is a vague directive that does not supply the factual grounding needed to prevent hallucinations. It is tempting because prompt engineering is a low-effort, no-retraining technique, but accuracy improves only when prompts include retrieved, verifiable context via retrieval-augmented generation.
- ✗
Increase the temperature to 1.0 for more diverse outputs
Why it's wrong here
Raising temperature to 1.0 increases sampling randomness, producing more varied and often less factual outputs, worsening hallucination. It is tempting because temperature tuning controls creativity, but reducing factual errors without retraining requires retrieval-augmented generation grounding responses in verified source documents.
Go deeper
Related to this question
About these practice questions
One of 1,008 original Generative AI Leader 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 by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This Generative AI Leader practice question is part of Courseiva's free Google Cloud 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 Generative AI Leader exam.