mediumMultiple Select
Generative AI Leader Practice Question: Using Vertex AI to build a language model for…
A company is using Vertex AI to build a language model for generating legal documents. They need to ensure the model's outputs are accurate and verifiable. Which TWO features should they use?
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
✓
Chain-of-thought reasoning
Chain-of-thought reasoning (C) is correct because it makes the model's step-by-step reasoning explicit, which improves the accuracy of complex legal drafting and lets reviewers trace how a conclusion or clause was derived, supporting verifiability. Grounding with citations to relevant legal texts (D) is correct because it anchors the model's outputs in authoritative sources and returns citations, so generated legal content can be checked against the referenced statutes, cases, or documents. The other options do not meet the requirement: confidence indicators (A) only give a score and do not make outputs verifiable; safety filters for legal content (B) restrict harmful or disallowed content but do not improve factual accuracy or traceability; and Model Cards (E) are documentation artifacts describing a model's intended use and limitations, not runtime features that verify generated outputs.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Confidence indicators
Why it's wrong here
Confidence indicators surface a model's own certainty scores, which are uncalibrated in large language models and cannot verify whether a cited clause or statute actually exists. They are tempting because confidence scores genuinely help triage predictions in classification tasks, but legal accuracy demands grounding against authoritative sources, not self-reported certainty.
- ✗
Safety filters for legal content
Why it's wrong here
Safety filters block harmful or toxic content categories, not factual errors; they cannot detect a fabricated case citation or an incorrect statutory deadline. They are tempting because legal content carries real compliance risk, so filtering seems prudent, but the requirement is verifiability of outputs, which grounding and citation features address instead.
- ✓
Chain-of-thought reasoning
Why this is correct
Chain-of-thought reasoning forces the model to expose intermediate legal reasoning steps, making each conclusion traceable and auditable. This satisfies the accuracy and verifiability constraint by letting reviewers inspect the derivation rather than trusting a bare output.
- ✓
Grounding with citations to relevant legal texts
Why this is correct
Grounding with citations anchors generated text to retrieved legal sources, returning references alongside each claim. This directly satisfies the accuracy and verifiability requirement: reviewers can trace assertions back to authoritative texts, and unsupported statements become detectable rather than silently fabricated.
- ✗
Model Cards
Why it's wrong here
Model Cards document a model's intended use, training data and evaluation metrics; they describe the model itself rather than verifying any individual generated document. They are tempting because governance and transparency artefacts are genuinely required for compliance reviews, but they operate at model level, not per-output accuracy, so they cannot satisfy the verifiability requirement.
Go deeper
Related to this question
About these practice questions
Courseiva writes every Generative AI Leader question from scratch — 1,008 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 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.