Free 1Z0-1158-26 – Oracle Cloud Infrastructure Enterprise AI Professional Practice Test Question

Select an option, then click Submit answer.

• ○

  A multimodal model, because it specializes in describing images of source code screenshots

• ○

  An embedding model, because converting code is a similarity-search problem

• ○

  A code model, because it is trained on programming languages to generate, complete, explain, and test source code

• ○

  A reranking model, because it can order legacy modules by complexity

Select an option, then click Submit answer.

• ○

  Act, because selecting a tool is the same as executing it

• ○

  Perceive, because the agent is taking in the search results for the first time

• ○

  Reason, because the agent is evaluating its current state against the goal and deciding the next action

• ○

  Observe, because the agent is generating the final answer for the user

Select an option, then click Submit answer.

• ○

  Keep the preferences in the active context window for the duration of each session only

• ○

  Hard-code each customer's preferences into the system prompt template

• ○

  Retrain the foundation model monthly with the preference data

• ○

  Store the preferences in long-term memory, persisted across sessions and retrieved when a returning customer starts a new conversation

Select an option, then click Submit answer.

• ○

  Share the tenancy administrator's console credentials with the application team

• ○

  Expose an unauthenticated public endpoint so the application can call the models directly

• ○

  Give each developer an individual API key that bypasses IAM policy evaluation

• ○

  Create an IAM policy that grants the developers' group permission to use the Generative AI resources in the relevant compartment, and authenticate API calls with OCI API signing keys or a principal-based method

Select an option, then click Submit answer.

• ○

  Immediately begin fine-tuning so the prototype uses a custom model

• ○

  Start with prompt engineering, using carefully designed instructions and a few example outputs (few-shot prompting) on a pretrained model

• ○

  Provision a dedicated AI cluster first, because customization is impossible without one

• ○

  Build a full RAG pipeline over the company document archive before showing anything

Select an option, then click Submit answer.

• ○

  Raise the model's temperature during peak hours so responses are generated faster

• ○

  Reduce the context window of the model so each request uses fewer tokens

• ○

  Increase the number of units allocated to the hosting dedicated AI cluster so more inference capacity (replicas) is available to absorb peak load

• ○

  Move the endpoint to the on-demand mode permanently, because dedicated clusters cannot scale

Select an option, then click Submit answer.

• ○

  It removes the need to specify a model in the request

• ○

  It guarantees the model will never produce an incorrect answer

• ○

  It provides a unified execution model where reasoning, built-in tools, function calling, and multi-turn state are handled as one response lifecycle rather than bolted together by the client

• ○

  It allows prompts of unlimited length with no token accounting

Select an option, then click Submit answer.

• ○

  Function tools are deprecated; every capability must be an MCP server

• ○

  MCP is only for tools provided by Oracle itself

• ○

  Remote MCP Calling can only be used with public internet servers

• ○

  A Function tool is appropriate for a capability owned and invoked within the application, while exposing it as an MCP server makes it a standardized, reusable tool that multiple agents — including remote ones — can share

Select an option, then click Submit answer.

• ○

  Vector stores fine-tune the model on all 40,000 pages so retrieval is unnecessary

• ○

  Embeddings-based retrieval scales to large, changing corpora by fetching only the most relevant passages per query, keeping context use and update cost low

• ○

  The Files API compresses all 40,000 pages into a single prompt

• ○

  Vector stores guarantee the model will never hallucinate

Select an option, then click Submit answer.

• ○

  Create an IAM policy that grants the agent access to OCI services

• ○

  Delete any previous versions of the agent from the tenancy

• ○

  Open the hosted agent endpoint to the public internet

• ○

  Package the agent and its dependencies as a container image

Select an option, then click Submit answer.

• ○

  Embed an OCI user's API signing key in a Kubernetes secret mounted into the container

• ○

  Create a dedicated service user account and rotate its password every 90 days

• ○

  Store the credentials in the container image and restrict who can pull the image from the registry

• ○

  Run the agent with an instance principal / workload identity backed by a dynamic group, and grant access with IAM policies scoped to the model endpoint and bucket

Select an option, then click Submit answer.

• ○

  Keep the cluster unchanged, because dedicated AI clusters cannot be resized after creation

• ○

  Add more replicas of the agent to consume the unused units and justify the spend

• ○

  Right-size the dedicated AI cluster to match steady-state demand, and consider hosted on-demand endpoints for spiky or low-volume workloads instead of always-on dedicated capacity

• ○

  Move the workload to a larger region so unit prices drop enough to offset the idle capacity