NCA-GENL Core Machine Learning and AI Knowledge Practice Question
A team is evaluating a generative large language model for a customer support chatbot. They need to ensure the model produces factually accurate and contextually appropriate responses while avoiding harmful or biased outputs. Which two techniques are most effective for aligning the model's behavior with these safety and quality requirements? (Choose two.)
⚠ Common exam trap
The trap here is assuming that scaling model size or tuning inference randomness can enforce safety, when alignment requires targeted training with human feedback or curated data.
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 on curated, high-quality demonstrations
RLHF and supervised fine-tuning on curated demonstrations are central alignment techniques. RLHF uses human preferences to optimize for safety and helpfulness, while supervised fine-tuning directly teaches desired responses. Temperature, parameter count, and batch size do not specifically improve factual accuracy or reduce harmful 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.
- ✗
Using a larger batch size during pretraining
Why it's wrong here
Batch size during pretraining influences optimization dynamics and can affect convergence, but it does not directly address safety, bias, or factual accuracy. Pretraining on raw web data still exposes the model to harmful content regardless of batch size. This is not a targeted alignment technique for the described requirements.
- ✓
Supervised fine-tuning on curated, high-quality demonstrations
Why this is correct
Supervised fine-tuning on carefully curated examples of desired responses teaches the model to mimic safe, accurate, and contextually appropriate behavior. It provides direct signal on how to handle sensitive topics and maintain factual grounding. Combined with human oversight, it is a core alignment method that shapes the model's output distribution toward the team's requirements.
- ✗
Reducing the model's parameter count
Why it's wrong here
Model size affects capacity and computational cost but does not inherently improve alignment or safety. A smaller model may even degrade performance on complex tasks and still exhibit biases present in its training data. Parameter count is not a lever for ensuring factual accuracy or harmlessness in a customer support context.
- ✗
Increasing the model's temperature during inference
Why it's wrong here
Higher temperature increases randomness in token selection, which can lead to more diverse but also more incoherent, biased, or factually incorrect outputs. It does not enforce safety or accuracy; instead, it often amplifies undesirable behavior. For a customer support chatbot, this would undermine reliability and is not an alignment technique.
- ✓
Reinforcement learning from human feedback (RLHF)
Why this is correct
RLHF fine-tunes the model using human preferences to optimize for helpfulness, harmlessness, and honesty. It directly aligns outputs with desired behaviors by training a reward model on human comparisons and then optimizing the policy via reinforcement learning. This makes it highly effective for reducing toxicity and improving factual consistency in customer support scenarios where nuanced judgment is required.
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.