NCA-GENL Core Machine Learning and AI Knowledge Practice Question
A developer is building a text summarization assistant that must produce concise, faithful summaries of long support tickets. The team wants to fine-tune a pre-trained large language model on a small labeled dataset of ticket-summary pairs. Which training approach best matches this goal?
⚠ Common exam trap
The trap here is assuming that any method using the ticket data, such as retrieval or pre-training, will teach the model to summarize, when only supervised fine-tuning uses the labeled summary targets directly.
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
✓
Supervised fine-tuning (SFT) on the ticket-summary pairs
The team has labeled input-output pairs for a specific generation task, which is exactly what supervised fine-tuning is designed for. SFT updates the pre-trained model's weights to map support tickets to concise summaries, leveraging existing language knowledge while learning the task. Other approaches either ignore the labels, require far more data and compute, or do not modify the model to perform the target 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.
- ✓
Supervised fine-tuning (SFT) on the ticket-summary pairs
Why this is correct
SFT directly optimizes the model on input-output pairs where the ticket is the prompt and the desired summary is the target. Because the team already has labeled summaries, this approach teaches the model the exact task and style required. It is the standard method for adapting a pre-trained LLM to a specific downstream generation task with limited labeled data.
- ✗
Unsupervised pre-training from scratch on the ticket corpus
Why it's wrong here
Pre-training from scratch on a small ticket corpus would require massive compute and data, and it would not teach the model the summary format. The model already has broad language knowledge from its original pre-training. Starting over discards that knowledge and is impractical for a small labeled dataset, so it does not fit the scenario.
- ✗
Reinforcement learning from human feedback (RLHF) using pairwise summary preferences
Why it's wrong here
RLHF is typically used to align a model with subjective human preferences after an initial SFT stage. It requires a reward model trained on many preference comparisons, which the team does not have. For a straightforward summarization task with labeled target summaries, RLHF adds unnecessary complexity and is not the most direct approach.
- ✗
Retrieval-augmented generation (RAG) with no model weight updates
Why it's wrong here
RAG retrieves external context at inference time to ground responses, but it does not change the model's behavior for producing concise summaries. The scenario asks how to train the model on labeled ticket-summary pairs. RAG could complement fine-tuning, but alone it does not adapt the model to the summarization task or its desired output style.
About these practice questions
One of 367 original NCA-GENL 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 NVIDIA exam blueprint
This NCA-GENL practice question is part of Courseiva's free NVIDIA 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 NCA-GENL exam.