NVIDIA GenAI LLM Associate Practice Test

Answer choices

1. A. A model with no access to the articles, within the rag and knowledge integration context.
2. B. A prompt that asks users to provide their own citations, within the proposed design.
3. C. Retrieval-augmented generation that retrieves relevant articles before answering, under the described rag and knowledge integration criteria.
4. D. A random response generator for variety, for the described technical objective and its associated operational control requirements, for this scenario.

Wrong-answer review

• A. A model with no access to the articles, within the rag and knowledge integration context.: A model with no access to the articles is not appropriate here because without article access, the model must rely on memory or guesses.
• B. A prompt that asks users to provide their own citations, within the proposed design.: A prompt that asks users to provide their own citations is not appropriate here because users should not have to supply the evidence the assistant is meant to retrieve.
• D. A random response generator for variety, for the described technical objective and its associated operational control requirements, for this scenario.: A random response generator for variety is not appropriate here because random generation undermines consistency and grounding.

Answer choices

1. A. Throughput, latency, resource use, and quality for the same workload
2. B. Only the price of the keyboard, for the specified implementation requirement.
3. C. The number of icons on the dashboard, for the stated experimentation and evaluation requirement.
4. D. A different prompt set for each configuration, within the stated policy framework.

Answer choices

1. A. NIM is only a handwritten prompt template, as the organization’s selected response.
2. B. NIM is intended to package optimized inference for portable deployment targets, for this task.
3. C. NIM removes the need for hardware planning, under the stated technical, operational, and governance constraints.
4. D. NIM exists only to edit source documents, as the recommended implementation across the complete governed service lifecycle.

Answer choices

1. A. That the system is always perfect, within the safety, governance, and responsible ai context.
2. B. Nothing about AI involvement, for the described technical objective and its associated operational control requirements.
3. C. The system’s AI nature, limitations, and appropriate use boundaries, for review.
4. D. That citations are forbidden, for the described technical objective.

Answer choices

1. A. Removing every table before indexing, within organization-wide risk-and-accountability boundaries.
2. B. Increasing answer creativity, for the described technical objective.
3. C. Document parsing that preserves table structure and relationships, within the rag and knowledge integration context.
4. D. Sorting chunks only by upload time, within the documented scope, ownership, and validation boundaries.

Answer choices

1. A. Triton Inference Server, for the required outcome.
2. B. A markdown-only model card, for the stated deployment and inference requirement.
3. C. A local text editor theme, for the affected environment.
4. D. An employee travel calendar, in practice.

Answer choices

1. A. Whether every document file name is short, for the required outcome.
2. B. Whether the UI uses square buttons, within the defined security and accountability boundaries.
3. C. Whether the embedding model matches the domain and query style, within the described operational context.
4. D. Whether the generator has no prompt, for the described technical objective and its associated operational control requirements.

Answer choices

1. A. A guarantee that longer prompts always improve accuracy
2. B. A GPU driver requirement for shorter filenames, for the required business outcome.
3. C. A sign that retrieval is never useful, within the defined security and accountability boundaries.
4. D. The lost-in-the-middle effect in long-context use, for this requirement.

Answer choices

1. A. Only a speed test of the login page, as the recommended implementation across the complete governed service lifecycle.
2. B. A RAG evaluation workflow covering retrieval relevance and answer faithfulness, as the proposed design for the complete governed operational workflow.
3. C. A count of how many documents exist, under organization-wide implementation-governance requirements.
4. D. A rule that citations are never reviewed, as the primary implementation for the described business requirement.

Answer choices

1. A. Checks should only run once per year, within the stated policy framework.
2. B. Safety checks should run before unsafe output reaches the user
3. C. The model should never be evaluated, under the documented operational and governance requirements.
4. D. Users should fix unsafe answers manually, as the primary implementation for the described business requirement.