NCA-GENL Experimentation Practice Question
You are running a NeMo fine-tuning experiment where validation loss decreases for the first three epochs, then rises steadily while training loss keeps falling. You want to confirm overfitting and select the most appropriate intervention. Which experiment action should you take first?
⚠ Common exam trap
The trap here is assuming more training will eventually close the gap, when a rising validation curve signals that additional epochs will only deepen the overfitting.
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
✓
Add regularization such as dropout or weight decay, then re-run the experiment and compare validation curves.
Rising validation loss alongside falling training loss indicates the model is fitting training-specific patterns that do not generalize. Adding regularization such as dropout or weight decay directly limits effective capacity and is a controlled, reversible change. Re-running with the same split and seed isolates the intervention, so the resulting validation curve tells you whether overfitting was indeed the dominant cause.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Switch the optimizer from Adam to SGD without changing any other hyperparameter.
Why it's wrong here
Changing the optimizer alone does not directly constrain model capacity, and SGD often needs retuned learning rates and schedules to behave comparably. Without controlling those variables, the experiment becomes confounded. This action does not specifically target overfitting and would likely require additional tuning before any conclusion about generalization could be drawn.
- ✗
Reduce the size of the validation set to lower evaluation noise.
Why it's wrong here
Shrinking the validation set increases variance in the validation loss estimate, making the rising trend noisier and harder to interpret. It also does not change model capacity or training dynamics, so it cannot mitigate overfitting. The observed pattern is a generalization gap, not an evaluation-set size problem, so this action misdiagnoses the root cause.
- ✗
Increase the number of training epochs and re-run to see if validation loss recovers.
Why it's wrong here
Extending training when validation loss is already rising will typically worsen the divergence between training and validation performance. More epochs let the model memorize the training set further, so validation loss will likely continue climbing. This action does not address the observed overfitting signal and wastes compute without validating the hypothesis that the model is over-regularized.
- ✓
Add regularization such as dropout or weight decay, then re-run the experiment and compare validation curves.
Why this is correct
A widening gap between falling training loss and rising validation loss is the classic overfitting signature. Introducing dropout or weight decay constrains model capacity and is the standard first intervention. Re-running with the same data split and seed lets you isolate the regularization effect, confirming whether overfitting is the cause rather than a data or learning-rate artifact.
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.