NCA-GENL Data Analysis and Visualization Practice Question
A data scientist is analyzing token-level loss values produced by an LLM evaluation run on a summarization dataset. Losses are stored as a list of floats, and most values cluster around 2.1, but a few exceed 9.0. The team wants a visualization that shows the shape of the loss distribution, including those extreme values, without hiding them through bin aggregation. Which visualization is most appropriate?
⚠ Common exam trap
The trap here is assuming any distribution plot preserves extreme values, when binning or kernel smoothing can hide sparse outliers.
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
✓
A strip plot (jittered scatter) of token-level loss
A strip plot plots every token-level loss as an individual mark, so extreme values remain visible instead of being merged into a bin or smoothed away. It reveals the distribution shape and the location of outliers simultaneously. Histograms, box plots, and violin plots each aggregate or summarize the data, which would obscure the few very high losses the team needs to inspect.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
A histogram with 50 equal-width bins
Why it's wrong here
A histogram with many equal-width bins still aggregates counts into ranges. Extreme loss values above 9.0 would fall into the last bin and lose their individual magnitude. If the tail is sparse, the histogram may show a tiny bar that hides how extreme the outliers are. This fails the requirement to preserve the extreme values rather than bin them.
- ✗
A violin plot of token-level loss
Why it's wrong here
A violin plot uses kernel density estimation to show distribution shape, but it smooths the data and can misrepresent sparse extreme values. A few tokens with loss above 9.0 would be spread into a thin tail that may be visually truncated or smoothed away. It also does not show individual extreme points, so it fails the goal of preserving those values explicitly.
- ✓
A strip plot (jittered scatter) of token-level loss
Why this is correct
A strip plot places each token loss as an individual point along one axis, optionally with vertical jitter to reduce overlap. Every extreme value above 9.0 remains visible as a distinct mark, and the overall shape of the distribution, including skew and multimodality, is preserved without binning. This directly satisfies the requirement to show shape while keeping extreme values.
- ✗
A box plot of token-level loss
Why it's wrong here
A box plot summarizes median, quartiles, and potential outliers, but it collapses the actual distribution shape into a compact five-number summary. For token-level loss with a long tail, the box plot would flag extreme points but would not reveal whether the bulk is unimodal, skewed, or bimodal. The scenario explicitly asks to show the shape while preserving extreme values, which a box plot does not accomplish.
About these practice questions
One of 367 original NCA-GENL practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. 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.