Oracle Cloud Infrastructure 2025 Generative AI Professional Practice Exam 2025: Latest Questions

A team is building an internal assistant and wants to reduce hallucinations by ensuring answers are grounded in the company’s policies stored in a document repository. Which approach is MOST appropriate?

In the OCI Generative AI service, you want to quickly validate a prompt for summarization before writing any application code. Which OCI capability is best suited for this task?

You are integrating OCI Generative AI into an application that runs on OCI Compute instances. What is the RECOMMENDED way for the application to authenticate to OCI services without storing long-lived credentials on disk?

A user reports that the assistant’s answers are sometimes overly verbose. Which prompt adjustment is MOST likely to help while keeping factuality intact?

A customer support team wants a chatbot that answers using only content from an approved knowledge base, and must cite which article sections were used. Which design best meets this requirement on OCI?

A developer calls a generative model endpoint from a private subnet and the request times out. The subnet has no internet gateway. Which OCI network configuration is MOST likely required to reach OCI public service endpoints while keeping the workload private?

Your organization requires that only a specific team can invoke Generative AI endpoints, and all requests must be auditable. Which combination of OCI capabilities best satisfies this?

You observe that responses degrade when the prompt includes large, unstructured context pasted from multiple documents. What is the BEST improvement to both quality and token efficiency?

A bank must prevent its generative AI application from returning sensitive personal data (for example, account numbers) in responses. The assistant uses RAG over internal documents that may contain PII. Which strategy is MOST effective as a defense-in-depth control?

A team wants to add an LLM-powered feature to an existing microservices application. Requirements: low latency, stable behavior across deployments, and the ability to gradually roll out prompt changes with the option to rollback. Which approach best meets these requirements?

A team wants to reduce hallucinations in an OCI Generative AI chat assistant without changing the underlying foundation model. Which approach is MOST appropriate?

You need to call OCI Generative AI from an application running on an OCI compute instance. Which OCI security best practice should you use to avoid storing long-lived credentials on the instance?

An architect is choosing between using an LLM for free-form text generation versus using embeddings. Which use case is BEST suited for embeddings?

A developer reports that calls to OCI Generative AI succeed from a laptop using the OCI CLI, but fail from an OCI Function with a 401/NotAuthenticated error. The function is deployed in a private subnet. What is the MOST likely fix?

Your organization must ensure that only approved compartments and projects can use OCI Generative AI, and that usage is traceable for audits. Which combination is MOST appropriate?

A team wants a chat assistant that answers questions only using content from an internal PDF library stored in Object Storage. They also need the assistant to cite sources. Which architecture pattern is MOST appropriate?

A company integrates OCI Generative AI into a customer-support portal. During load testing, responses are intermittently slow and sometimes time out. Which mitigation is MOST appropriate at the application layer?

A security review flags that prompts may contain personally identifiable information (PII). The organization wants to minimize PII exposure to the LLM while preserving utility. Which approach is BEST?

A regulated enterprise requires that the model's response must be traceable to the exact inputs used, including retrieved documents, prompt templates, and model parameters. Which design is MOST appropriate?

You deploy a RAG solution and notice correct documents are retrieved, but the LLM frequently ignores the provided context and answers from general knowledge. Which change is MOST likely to improve context adherence?

A retail team is prototyping a chatbot using an OCI Generative AI chat model. They notice the model sometimes invents product return-policy details. They want a quick improvement without changing the base model or building a complex pipeline. What is the MOST appropriate first step?

A developer calls an OCI Generative AI endpoint from a function but receives an authorization error. The function is deployed in OCI and runs successfully, but the Generative AI call fails. Which fix is MOST likely to resolve the issue following OCI best practices?

A team is building an internal Q&A assistant that must answer based on thousands of company documents stored in Object Storage. Users complain the assistant often ignores relevant documents even though retrieval is enabled. Which change is MOST likely to improve retrieval quality without changing the LLM?

A healthcare customer wants to ensure that no sensitive patient identifiers are sent to a Generative AI model, but they still want summaries of clinical notes. Which approach BEST addresses this requirement?

An enterprise is designing a multi-team GenAI platform on OCI. Each team must use shared foundation models but cannot access other teams' prompts, outputs, or fine-tuned artifacts. The security team requires centralized governance with least privilege. Which architecture BEST meets these goals?