mediumMultiple Choice
Generative AI Leader Practice Question: A data scientist is using Vertex AI to fine-tune…
A data scientist is using Vertex AI to fine-tune a Gemini model for a specialized legal document summarization task. They have a small set of labeled examples (200 pairs). Which fine-tuning method is MOST cost-effective and likely to perform well?
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
✓
Adapter-based fine-tuning (e.g., LoRA)
Adapter-based fine-tuning (like LoRA) updates only a small fraction of parameters, making it efficient with small datasets and low cost, while still adapting the model to the task.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Full fine-tuning of all model parameters
Why it's wrong here
Full fine-tuning updates every parameter, demanding substantial compute and risking overfitting on only 200 labelled pairs. Parameter-efficient tuning, such as LoRA, freezes the base weights and trains small adapter matrices, cutting cost while suiting limited data. Full fine-tuning becomes the right choice when large, diverse datasets and maximum task-specific accuracy justify the expense.
- ✓
Adapter-based fine-tuning (e.g., LoRA)
Why this is correct
Adapter-based fine-tuning such as LoRA freezes the base weights and trains small injected matrices, so only a fraction of parameters update. With just 200 labelled pairs, this avoids overfitting and full fine-tuning's cost, satisfying the stem's small-dataset, cost-effective constraint.
- ✗
Training a small custom model from scratch
Why it's wrong here
Training from scratch discards Gemini's pretrained language capability and needs far more than 200 pairs to reach usable quality, making it costly and unlikely to converge. It is appropriate only when large domain-specific corpora exist and no suitable foundation model is available.
- ✗
Prompt engineering with few-shot examples only
Why it's wrong here
Few-shot prompting changes only the input context; no weights are updated, so the model cannot absorb the specialised legal vocabulary and summarisation style present in the 200 labelled pairs. It is the right low-cost choice when examples are too few or the task is already within the model's capability.
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.