NCP-GENL Fine-Tuning Practice Question
A research team is fine-tuning a model with NVIDIA NeMo and wants to reduce the risk of catastrophic forgetting of general capabilities while still adapting to a specialized domain. They have a small domain dataset and limited compute. Which fine-tuning approach best balances domain adaptation with retention of pretrained knowledge?
⚠ Common exam trap
The trap here is equating strong domain adaptation with full fine-tuning, when a small dataset and limited compute make parameter-efficient methods both safer and more practical.
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
✓
Parameter-efficient fine-tuning with LoRA using a modest rank and a low learning rate.
LoRA with a modest rank and low learning rate freezes the base model and trains small adapters, which constrains updates and preserves pretrained knowledge while still adapting to the domain. Full fine-tuning with high learning rates, training from scratch, or freezing all layers and training only a head either cause forgetting, require excessive resources, or fail to adapt the generative model.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Full-parameter fine-tuning with a high learning rate for many epochs on the domain dataset.
Why it's wrong here
Full-parameter fine-tuning with a high learning rate and many epochs on a small dataset is a classic recipe for catastrophic forgetting. The model can overfit to the narrow domain and lose general capabilities. While it may adapt strongly to the domain, it directly conflicts with the goal of retaining pretrained knowledge under limited compute.
- ✗
Freezing all layers and training only the final classification head on the domain dataset.
Why it's wrong here
Training only a classification head does not adapt the generative model's internal representations to the domain. It also assumes a classification task, while the scenario describes general domain adaptation of an LLM. This method limits learning capacity and is not appropriate for generative fine-tuning with NeMo.
- ✓
Parameter-efficient fine-tuning with LoRA using a modest rank and a low learning rate.
Why this is correct
LoRA freezes the pretrained weights and trains small low-rank adapters, which limits drift from the original model and reduces forgetting. A modest rank and low learning rate further constrain updates, making it well suited for small datasets and limited compute. This approach balances domain adaptation with retention of general capabilities.
- ✗
Training from scratch on the domain dataset using the same architecture and tokenizer.
Why it's wrong here
Training from scratch discards all pretrained knowledge and requires far more data and compute than the team has. It would not retain general capabilities and would likely underperform a fine-tuned model on the domain task. This approach is the opposite of what is needed for a small dataset with limited compute.
About these practice questions
This NCP-GENL question is part of Courseiva's 352-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 NCP-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 NCP-GENL exam.