1Z0-1127-25 OCI Generative AI Service Practice Test 3 — 15 Questions

Question 1

A security engineer needs to allow the DataScience group to use OCI Generative AI resources (Chat, Embedding, and Summarisation) in the compartment 'genai-compartment', but not allow them to create dedicated AI clusters. Which IAM policy statement achieves this?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Allow group DataScience to use generative-ai-family in compartment genai-compartment and allow group DataScience to inspect dedicated-ai-clusters in compartment genai-compartment. Granting 'use generative-ai-family' in genai-compartment gives the DataScience group access to the Chat, Embedding, and Summarisation capabilities, while the separate 'inspect dedicated-ai-clusters' statement only permits read-level visibility and does not grant the manage/create permissions needed to provision dedicated AI clusters. In OCI IAM, 'use' on a resource family covers the verbs required to consume the service without conferring full management, and 'inspect' is a read-only verb that cannot create clusters. Option A fails because 'use dedicated-ai-clusters' would grant the ability to work with and potentially create dedicated AI clusters, which the scenario forbids. Option B is invalid because OCI policy statements do not support a 'where request.permission != ...' condition of that form for excluding a permission. Option D is invalid because OCI IAM policies are allow-only and do not support explicit 'deny' statements.

Answer

Allow group DataScience to use generative-ai-family in compartment genai-compartment and allow group DataScience to use dedicated-ai-clusters in compartment genai-compartment

Answer

Allow group DataScience to use generative-ai-family in compartment genai-compartment where request.permission != 'MANAGE_DEDICATED_AI_CLUSTER'

Answer

Allow group DataScience to use generative-ai-family in compartment genai-compartment and deny group DataScience to manage dedicated-ai-clusters in compartment genai-compartment

Question 2

A developer is using the OCI GenAI Chat API to build a multi-turn customer support chatbot. They want the assistant to always introduce itself as 'SupportBot' and never mention being an AI. How should they configure the API call?

Accepted Answer

Set the system message (preamble) to 'You are SupportBot. You must never mention that you are an AI.'. The OCI GenAI Chat API supports a 'system message' or 'preamble' parameter that sets the assistant's behavior and identity for the entire conversation. By setting the preamble to 'You are SupportBot. You must never mention that you are an AI.', the developer enforces the desired persona and restriction across all turns, ensuring the assistant introduces itself as SupportBot and avoids any reference to being an AI.

Answer

Use the Generate API instead and include the instruction in the prompt

Answer

Set the preamble override to a JSON object with the assistant's identity

Answer

Set the first user message to 'Introduce yourself as SupportBot and never say you are an AI.'

Question 3

A data scientist is creating a fine-tuning job in OCI Generative AI. They have prepared a JSONL dataset with prompt/completion pairs. What is the correct format for each line in the JSONL file?

Accepted Answer

{"prompt": "What is OCI?", "completion": "Oracle Cloud Infrastructure is a cloud computing platform."}. The OCI Generative AI fine-tuning service expects JSONL files where each line contains exactly the keys "prompt" and "completion". This format matches the service's internal training pipeline, which maps the prompt to the input and the completion to the expected output during supervised fine-tuning.

Answer

{"input": "What is OCI?", "output": "Oracle Cloud Infrastructure is a cloud computing platform."}

Answer

{"text": "What is OCI?", "label": "Oracle Cloud Infrastructure is a cloud computing platform."}

Answer

{"question": "What is OCI?", "answer": "Oracle Cloud Infrastructure is a cloud computing platform."}

Question 4

A company wants to build a customer service chatbot that answers questions about their internal policy documents. The documents are updated monthly, and the team cannot afford to retrain a model each time. Which approach is MOST appropriate?

Accepted Answer

Use Retrieval-Augmented Generation (RAG) with the policy documents indexed in a vector store. Retrieval-Augmented Generation (RAG) is the most appropriate approach because it allows the chatbot to answer questions by retrieving relevant chunks from the policy documents stored in a vector store at inference time, without requiring any model retraining. When documents are updated monthly, you only need to re-index the new content into the vector store, and the LLM can use the retrieved context to generate accurate answers. This decouples knowledge updates from model training, making it cost-effective and agile for frequently changing internal documents.

Answer

Fine-tune a base LLM on the policy documents monthly

Answer

Train a custom model from scratch on the policy documents each month

Answer

Use a larger foundation model with a longer context window and paste all documents into each prompt

Question 5

A regulatory compliance team needs to restrict access to the OCI Generative AI service so that only users in the 'AI_Engineers' group can create fine-tuning jobs and endpoints. Which IAM policy statement should be used?

Accepted Answer

Allow group AI_Engineers to use generative-ai-family in compartment ABC. To allow a group to manage generative AI resources, you allow them to use the 'generative-ai-family' resource type. Specific verbs like 'inspect' limit visibility but not management. 'Read' only allows viewing, not creating.

Answer

Allow group AI_Engineers to inspect generative-ai-family in compartment ABC

Answer

Allow group AI_Engineers to manage generative-ai-family in tenancy

Answer

Allow group AI_Engineers to read generative-ai-family in compartment ABC

Question 6

A developer is using the Chat API for a multi-turn conversation. They want the assistant to adopt a formal tone and always identify itself as 'OracleBot'. How should they configure the API request?

Accepted Answer

Include the persona instructions in the 'preamble_override' field and the conversation history in the 'messages' field of the Chat API. The Chat API supports a 'preamble_override' parameter to set system-level instructions, and 'messages' array for conversation history. The preamble is the correct place to set persona. Temperature and max tokens do not control persona. The Generate API does not support system prompts.

Answer

Set the 'temperature' parameter low and 'max_tokens' high

Answer

Use the Embedding API to embed the persona instructions and pass them with each request

Answer

Set the 'system_prompt' in the Generate API

Question 7

During fine-tuning a model in OCI Generative AI, the training loss does not decrease after several epochs. The dataset has 5,000 prompt-completion pairs. What is the MOST likely cause?

Accepted Answer

The dataset format is incorrect (e.g., missing 'prompt' or 'completion' fields). In OCI Generative AI fine-tuning, the dataset must be a JSON file with exactly 'prompt' and 'completion' fields. If these fields are missing or incorrectly named, the service cannot parse the training data, so the model never learns from the actual content, causing the training loss to remain flat across epochs. This is the most direct cause of a non-decreasing loss when the dataset structure is invalid.

Answer

The dataset is too large, causing underfitting

Answer

The dataset contains many duplicate prompt-completion pairs

Answer

The learning rate is too high, causing the loss to oscillate

Question 8

Which OCI Generative AI API should be used to generate a summary of a long legal document?

Accepted Answer

Generate API. The OCI Generative AI Generate API (option D) is specifically designed for text generation tasks, including summarization, by taking a prompt and returning a coherent response. For a long legal document, you would provide the document text as part of the prompt and instruct the model to generate a summary, making the Generate API the correct choice for this use case.

Answer

Embedding API

Answer

Rerank API

Answer

Chat API

Question 9

A data scientist needs to fine-tune a Llama 3 model for a legal document classification task. They have a dataset of 10,000 labeled examples. Which fine-tuning technique available in OCI Generative AI is most suitable for efficiently adapting the model with limited computational overhead?

Accepted Answer

T-Few fine-tuning. T-Few fine-tuning is a parameter-efficient technique that updates only a small fraction of parameters and is designed to adapt large language models with minimal computational overhead. For a legal document classification task with 10,000 labeled examples, T-Few provides efficient adaptation without the cost of full fine-tuning.

Answer

Full fine-tuning all model parameters

Answer

LoRA (Low-Rank Adaptation)

Answer

Prefix tuning

Question 10

A developer is building a summarization pipeline using OCI Generative AI. They want to ensure the summary includes key points from the entire document without truncation. Which parameter should they primarily adjust?

Accepted Answer

Max_tokens. Max_tokens controls the maximum length of the generated output, so if summaries are being truncated before covering the whole document, increasing max_tokens allows the model to produce a longer, more complete summary. The other parameters influence style, repetition, and randomness, not output length.

Answer

Frequency_penalty

Answer

Temperature

Answer

Top_p

Question 11

A company needs to classify customer support tickets into 20 categories. They have a labeled dataset of 50,000 examples. They want to use OCI Generative AI Embedding API to generate embeddings, then train a classifier. Which input type should they use for the embedding API when processing the training examples?

Accepted Answer

classification. The OCI Generative AI Embedding API accepts an input type parameter that tells the model how the text will be used. For training a classifier on labeled examples, the correct input type is 'classification', which optimizes embeddings for classification tasks. Using the wrong input type (e.g., search_document or search_query) produces embeddings tuned for retrieval, degrading classifier accuracy.

Answer

search_query

Answer

clustering

Answer

search_document

Question 12

An organization wants to deploy a chatbot that uses a custom fine-tuned model. They have provisioned a Dedicated AI Cluster with 4 model units. During peak hours, they observe high latency and want to reduce it. What is the most cost-effective change?

Accepted Answer

Increase the number of model units in the dedicated cluster. Adding more model units increases throughput and reduces latency by distributing the load. This is the direct way to improve performance on a dedicated cluster.

Answer

Reduce the max_tokens parameter in the application

Answer

Switch to shared infrastructure to get more capacity

Answer

Enable response streaming

Question 13

A company wants to build a customer service chatbot that answers questions about their internal policy documents. The documents are updated monthly, and the team cannot afford to retrain a model each time. Which approach is MOST appropriate?

Accepted Answer

Use Retrieval-Augmented Generation (RAG) with the policy documents indexed in a vector store. RAG is the correct approach because it separates knowledge (the policy documents) from the model, allowing documents to be updated in the vector store without retraining. The chatbot retrieves relevant chunks at query time and passes them to the LLM as context, so monthly document updates only require re-indexing, not model retraining. This is both cost-effective and operationally sustainable.

Answer

Fine-tune a base LLM on the policy documents monthly

Answer

Train a custom model from scratch on the policy documents each month

Answer

Use a larger foundation model with a longer context window and paste all documents into each prompt

Question 14

A developer is using the OCI Generative AI Chat API with a system prompt to guide the assistant's behavior. They notice that after a few turns, the assistant starts ignoring the system instructions. What is the MOST likely cause?

Accepted Answer

The system prompt (preamble) is only sent once and may be truncated as the conversation grows. In OCI Generative AI chat, the system prompt (preamble) is typically sent as the first message in the conversation. As the conversation grows, the model's context window fills with user and assistant turns, and the system prompt can be truncated or pushed out of the effective attention window, causing the model to drift from its original instructions. The fix is to re-send or reinforce the system prompt on each turn.

Answer

The temperature is set too high, causing the model to ignore instructions

Answer

The model is not fine-tuned to follow system prompts

Answer

The conversation history is not being sent with subsequent requests, so the model loses context

Question 15

Which of the following is the correct format for a training dataset used in OCI Generative AI fine-tuning?

Accepted Answer

JSONL file with 'prompt' and 'completion' fields. OCI Generative AI fine-tuning expects a JSONL (JSON Lines) file where each line is a JSON object containing 'prompt' and 'completion' fields. This format is the documented standard for supervised fine-tuning datasets in OCI Generative AI. Other formats like CSV, TXT, or Parquet are not accepted for this purpose.

Answer

CSV file with columns 'input' and 'output'

Answer

TXT file with one prompt-completion pair per line separated by a tab

Answer

Parquet file with 'text' and 'label' columns