NCA-GENL Core Machine Learning and AI Knowledge Practice Question
A data scientist is preparing a labeled dataset of 50,000 customer support tickets for supervised fine-tuning of an LLM. Each ticket must be assigned exactly one of eight department labels. Which loss function is most appropriate for training this classification head?
⚠ Common exam trap
A common mix-up: candidates confuse single-label multi-class classification, which uses categorical cross-entropy, with multi-label tagging, which uses binary cross-entropy.
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
✓
Categorical cross-entropy loss
Because every support ticket carries exactly one of eight mutually exclusive department labels, the task is single-label multi-class classification. Categorical cross-entropy, paired with a softmax output layer, directly maximizes the probability of the correct department while normalizing across all eight classes, giving the strongest and most stable training signal for this scenario.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Categorical cross-entropy loss
Why this is correct
Categorical cross-entropy compares the predicted probability distribution over the eight mutually exclusive department labels with the one-hot true label, penalizing probability mass placed on incorrect classes. It is the standard objective for single-label multi-class classification and produces well-calibrated softmax outputs, making it the right choice when each ticket belongs to exactly one department.
- ✗
Mean squared error loss
Why it's wrong here
Mean squared error measures numeric distance between predicted and target values and is designed for regression, not for selecting among discrete categories. Applying it to a softmax output over eight departments yields weak gradients when predictions are confidently wrong and does not correspond to a proper likelihood, making convergence slower and less reliable than cross-entropy for classification.
- ✗
Contrastive loss
Why it's wrong here
Contrastive loss pulls embeddings of similar pairs together and pushes dissimilar pairs apart, which is used for representation learning, retrieval, or siamese networks. It does not directly produce a probability distribution over eight fixed department labels, so it is not the correct objective for a single-label classification head on support tickets.
- ✗
Binary cross-entropy loss
Why it's wrong here
Binary cross-entropy treats each output as an independent yes/no decision applied per label, which suits multi-label problems where several tags can apply simultaneously. Here each ticket has exactly one of eight mutually exclusive departments, so independent sigmoid outputs would fail to enforce that only one label is chosen and would not properly model the competition between classes.
About these practice questions
This NCA-GENL question is part of Courseiva's 367-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. 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.