NCA-GENL Experimentation Practice Question
During an ablation study on a retrieval-augmented LLM in NeMo, an engineer removes the reranking stage and observes that answer accuracy drops by 12 points, but latency improves by 40 percent. A stakeholder asks whether reranking should be kept. Which experimental next step best supports a defensible recommendation?
⚠ Common exam trap
The trap here is treating a two-point ablation as sufficient evidence for a binary keep-or-remove decision, when the useful answer is often an intermediate configuration found by sweeping cost and quality.
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
✓
Measure the accuracy-latency trade-off across several reranker sizes and retrieval depths, then compare against the application's latency budget and quality target.
A single ablation establishes that reranking trades latency for accuracy but does not show whether a middle ground exists. Sweeping reranker sizes and retrieval depths produces a trade-off curve, and evaluating those points against the application's latency budget and quality target converts measurements into a recommendation. Asserting a priority, removing retrieval, or compensating with more documents does not answer the stakeholder's question.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Keep reranking because accuracy is always more important than latency in production systems.
Why it's wrong here
Prioritizing accuracy unconditionally ignores the application's service-level objectives and the observed latency trade-off. Some deployments, such as interactive assistants, may not tolerate the extra latency even at a quality cost. A defensible recommendation requires measuring the trade-off against requirements rather than asserting a universal priority, so this choice substitutes opinion for evidence.
- ✓
Measure the accuracy-latency trade-off across several reranker sizes and retrieval depths, then compare against the application's latency budget and quality target.
Why this is correct
The single ablation shows a trade-off but not the shape of the curve or whether a cheaper configuration can retain most of the accuracy gain. Sweeping reranker sizes and retrieval depths reveals intermediate operating points, and mapping them to the latency budget and quality target turns the data into a concrete recommendation. This is the experiment that supports a defensible decision rather than a binary choice.
- ✗
Increase the number of retrieved documents while keeping the reranker disabled to recover the lost accuracy.
Why it's wrong here
Adding more candidates without reranking may partially recover recall, but it also lengthens the context and increases inference cost, so latency could worsen. It addresses reranking only indirectly and does not map the trade-off between reranker cost and answer quality. The scenario needs a systematic comparison of reranking configurations, not a workaround that changes a different part of the pipeline.
- ✗
Remove the retrieval stage entirely to see whether the model's parametric knowledge is sufficient.
Why it's wrong here
Removing retrieval tests a different hypothesis about knowledge sourcing and does not clarify the reranker trade-off that was observed. The result might be interesting for architecture decisions but leaves the accuracy-versus-latency question unanswered. The stakeholder asked specifically about reranking, so the next experiment should vary reranking cost and quality, not eliminate retrieval.
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.