hardMultiple Choice
Generative AI Leader Practice Question: A legal team wants to use GenAI to review…
A legal team wants to use GenAI to review contracts and highlight risky clauses. They need the AI to consistently follow a specific classification taxonomy. The team has a small set of labeled examples (500 contracts). Which approach yields the BEST accuracy for this use case?
⚠ Common exam trap
Generative AI Leader often tests the confusion between RAG (good for grounding on external knowledge) and fine-tuning (good for teaching a specific output format or taxonomy), causing candidates to pick RAG when consistency of classification is the requirement.
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
✓
Fine-tune a base model using the labeled examples in Vertex AI
Fine-tuning a base model on the 500 labeled contracts teaches the model the specific classification taxonomy directly in its weights, producing consistent, high-accuracy outputs that follow the exact label set. With a small but well-labeled dataset, supervised fine-tuning is the most reliable way to enforce a fixed taxonomy rather than relying on prompt instructions.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Prompt engineer a large foundation model with few-shot examples in Vertex AI Studio
Why it's wrong here
Few-shot prompting conditions the model at inference only; it cannot durably adjust weights to a 500-example taxonomy, so classification drifts across contracts. It suits rapid prototyping or low-volume tasks where labelled data is scarce and near-perfect consistency is not required.
- ✗
Use the RAG Engine to retrieve similar clauses and ask the model to classify
Why it's wrong here
Retrieval returns semantically similar clauses but does not teach the model the taxonomy's label boundaries, so classification stays inconsistent across the 500 labelled contracts. RAG suits grounding answers in source documents, such as citing clause text. Supervised fine-tuning on the labelled examples is the approach that fixes the taxonomy.
- ✓
Fine-tune a base model using the labeled examples in Vertex AI
Why this is correct
Fine-tuning adjusts the base model's weights on the 500 labelled contracts, embedding the specific taxonomy into the model itself. This yields higher, more consistent classification accuracy than prompt engineering or retrieval alone, satisfying the small-labelled-dataset constraint.
- ✗
Use a larger foundation model without fine-tuning and rely on its pre-trained knowledge
Why it's wrong here
Pre-trained knowledge cannot enforce a bespoke taxonomy; without examples the model invents its own clause categories, so accuracy stays inconsistent. It tempts when the task is generic summarisation or drafting, where broad world knowledge suffices and no labelled data exists.
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 and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Google Cloud exam blueprint
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.