NCA-GENL Core Machine Learning and AI Knowledge Practice Question
A researcher is fine-tuning a large language model on a downstream task with a small dataset. They notice that the model achieves high training accuracy but poor validation accuracy. Which regularization technique is most appropriate to address this issue?
⚠ Common exam trap
A common mix-up: candidates confuse overfitting with underfitting; high training accuracy and low validation accuracy clearly indicate overfitting, so techniques that increase model capacity or training time are counterproductive.
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
✓
Apply dropout to the transformer layers during fine-tuning.
Overfitting on a small dataset is best addressed by regularization methods like dropout, which introduce noise during training and force the model to learn more generalizable patterns. Dropout is particularly effective in transformers and is easy to apply during fine-tuning. Increasing learning rate, reducing data, or removing early stopping would all worsen the problem.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Increase the learning rate to speed up convergence.
Why it's wrong here
Increasing the learning rate would likely worsen overfitting by causing larger weight updates that fit noise in the small dataset. The problem is poor generalization, not slow convergence. A higher learning rate could also destabilize training and lead to divergence, making validation performance even worse.
- ✗
Remove early stopping and train for more epochs.
Why it's wrong here
Training for more epochs without early stopping would allow the model to continue fitting the training data, exacerbating overfitting and further widening the gap between training and validation accuracy. Early stopping is a regularization technique that should be used, not removed, to prevent overfitting on small datasets.
- ✓
Apply dropout to the transformer layers during fine-tuning.
Why this is correct
Dropout randomly deactivates neurons during training, preventing the model from relying too heavily on specific features and reducing overfitting. In transformer fine-tuning, applying dropout to attention and feed-forward layers is a standard regularization method. It improves generalization on small datasets by encouraging more robust representations.
- ✗
Reduce the size of the training dataset further.
Why it's wrong here
Reducing the training dataset would decrease the amount of information available for learning, likely worsening both training and validation performance. Overfitting is already indicated by high training accuracy; less data would not help the model generalize better. This is counterproductive for the stated issue.
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.