Free NCA-GENL: NVIDIA Certified Associate — Generative AI LLMs Exam Questions and Answers

Which spaCy pipeline component labels spans of text such as people, organizations and locations?

• AA. Dependency parser
• BB. Named entity recognizer
• CC. Part-of-speech tagger
• DD. Lemmatizer

What does horizontal scaling of an LLM inference service involve?

• AA. Running additional copies of the service behind a load balancer
• BB. Replacing the deployed model with a smaller model that has fewer parameters
• CC. Moving the service to a single larger GPU with more memory and bandwidth
• DD. Lowering the numeric precision of the model weights to speed up each request

Which NVIDIA open-source toolkit lets developers add programmable rules that control the topics and safety behavior of an LLM-based application?

• AA. NeMo Guardrails
• BB. TensorRT-LLM
• CC. NVIDIA Nsight Systems
• DD. NVIDIA DALI

Which kind of model is trained so that sentences with similar meaning map to nearby dense vectors?

• AA. A sequence-to-sequence model trained for translation
• BB. A bi-encoder transformer trained with a contrastive objective
• CC. A convolutional classifier trained on character n-grams
• DD. An autoregressive decoder with a language modeling head

A company needs a question-answering service over 2 million internal PDF pages. Policy requires that no document text leave the company's own data center. Which set of components meets this requirement?

• AA. An embedding model, a vector database, and a locally hosted LLM
• BB. A tokenizer, a relational database, and a text classifier
• CC. An embedding model, a vector database, and a cloud-hosted LLM API
• DD. A reward model, a keyword index, and a locally hosted LLM

A developer must pull company names, people and dates out of incoming support emails using spaCy. Which capability should the developer use?

• AA. Sentence segmentation through doc.sents
• BB. Lemmatization through token.lemma_
• CC. Named entity recognition through doc.ents
• DD. Dependency parsing through token.dep_

In Keras, which method trains an already compiled model on training data for a set number of epochs?

• AA. model.evaluate()
• BB. model.fit()
• CC. model.summary()
• DD. model.predict()

A retailer wants to discover natural customer segments in purchase records that have no predefined labels. Which data mining technique fits?

• AA. Logistic regression
• BB. k-means clustering
• CC. Decision tree classification
• DD. Linear regression

A team has fine-tuned a new version of the model behind an internal assistant that runs on GPU servers. The new version passed all offline benchmarks. The team wants to limit the risk to users during release. Which approach is most appropriate?

• AA. Send about 5% of live traffic to the new version, compare its error rate and latency with the current one, then widen the rollout
• BB. Replace the current version for all users once it passes the offline benchmarks, and keep the old container image archived
• CC. Deploy the new version to every replica at once during off-peak hours and watch the dashboards for the first hour
• DD. Run the new version on an idle replica for 24 hours without live traffic, then switch all users to it

A developer is building a summarizer for 300-page reports. Each report is far longer than the model's context window. Which approach is most appropriate?

• AA. Truncate each report to the first part that fits in the context window and summarize that
• BB. Split the report into chunks, summarize each chunk, then summarize the combined chunk summaries
• CC. Embed the report and return the sentence whose vector is closest to the report's average
• DD. Raise the temperature so the model produces longer and more varied summaries