1Z0-1127-25 Fundamentals of Large Language Models Practice Test 6 — 15 Questions

Question 1

A data scientist wants to improve the accuracy of a summarization model on medical texts. Which OCI service feature is most suitable?

Accepted Answer

OCI Generative AI fine-tuning. C is correct because OCI Generative AI fine-tuning allows a data scientist to adapt a pre-trained large language model (LLM) specifically for medical text summarization by training it on domain-specific data. This improves accuracy by aligning the model's outputs with the terminology, context, and nuances of medical literature, which generic models may not capture well.

Answer

OCI Data Flow

Answer

OCI Language service

Answer

OCI Anomaly Detection

Question 2

An architect needs to ensure that an LLM deployed in OCI does not reveal sensitive information in its outputs. Which technique should be used?

Accepted Answer

Output filtering via custom inference wrapper. Output filtering via a custom inference wrapper allows the architect to inspect and sanitize the model's generated text before it reaches the user, preventing the leakage of sensitive information such as PII, credentials, or internal data. This technique operates at the application layer, intercepting the LLM's response and applying rules or regex patterns to redact or block prohibited content, which is essential for compliance and data security in production deployments.

Answer

Limiting max tokens

Answer

OCI Data Safe masking

Answer

Input sanitization

Question 3

A developer notices that an LLM's responses are too verbose. Which parameter adjustment would most effectively reduce verbosity?

Accepted Answer

Decrease max_tokens. Decreasing max_tokens directly limits the maximum length of the LLM's response, which is the most straightforward way to reduce verbosity. This parameter caps the number of tokens the model can generate, forcing it to produce shorter completions. Other parameters like frequency_penalty, top_p, and temperature influence the style, diversity, or randomness of the output but do not directly control response length.

Answer

Increase frequency_penalty

Answer

Increase top_p

Answer

Decrease temperature

Question 4

A team wants to deploy an LLM for real-time inference with low latency. Which OCI deployment option is best?

Accepted Answer

OCI Data Science Model Deployment with GPU shapes. OCI Data Science Model Deployment with GPU shapes is the best option because it provides managed, scalable, low-latency inference endpoints for LLMs. GPU shapes (e.g., VM.GPU.A10) are essential for the parallel matrix computations required by transformer-based models, and the deployment service supports auto-scaling and load balancing to maintain real-time response times.

Answer

OCI Functions with CPU

Answer

OCI Events

Answer

OCI Streaming

Question 5

An AI specialist is troubleshooting why a fine-tuned model produces inconsistent results across different inference calls. What is the most likely cause?

Accepted Answer

The temperature is set too high. Temperature controls the randomness of token sampling during inference. A high temperature (e.g., >1.0) increases the probability of selecting less likely tokens, causing the model to produce varied outputs for the same input across different calls. This is the most direct cause of inconsistent results when the base model and fine-tuning are otherwise sound.

Answer

The base model is not suitable

Answer

The model is overfitted

Answer

The fine-tuning dataset is too small

Question 6

A company uses RAG (Retrieval-Augmented Generation) with OCI OpenSearch and OCI Generative AI. The system retrieves irrelevant documents. What is the first step to debug?

Accepted Answer

Check the embeddings quality. When RAG retrieves irrelevant documents, the most common root cause is poor embedding quality—vectors that fail to capture semantic similarity between the query and the documents. Checking embeddings (e.g., cosine similarity scores, dimensionality, or model used) is the logical first step before adjusting retrieval parameters or the LLM itself.

Answer

Use a different LLM

Answer

Increase the number of retrieved documents

Answer

Lower the temperature

Question 7

A developer is using the OCI Generative AI Python SDK. They receive a 400 error 'InvalidParameter'. What is the most likely reason?

Accepted Answer

Invalid model name. The 400 'InvalidParameter' error in the OCI Generative AI Python SDK specifically indicates that the request contained a parameter that is not recognized or is malformed. The most common cause is providing an incorrect or unsupported model name, as the SDK validates model names against a predefined list of available models in the OCI Generative AI service. Other parameter errors, such as invalid temperature or top_p values, would also produce this error, but model name mismatches are the most frequent trigger.

Answer

Exceeded token limit

Answer

Network timeout

Answer

Missing API key

Question 8

Which TWO of the following are benefits of using OCI Generative AI service compared to self-hosting an LLM?

Accepted Answer

Built-in content safety filters. Option C is correct because OCI Generative AI provides built-in content moderation and safety filters (guardrails) that screen prompts and responses for harmful or inappropriate content, which a self-hosted LLM would require you to implement and maintain yourself. Option D is correct because the managed service automatically scales inference capacity to handle varying request volumes without you provisioning or managing GPU infrastructure, whereas self-hosting requires manual capacity planning and scaling. Option A is not necessarily true: latency depends on network path, region, and model, and self-hosting can sometimes yield lower latency when the model runs close to the application. Option B is incorrect because data egress costs depend on where the service is accessed from and are not inherently eliminated by using the managed service. Option E is incorrect because OCI Generative AI offers pre-trained and fine-tunable models but does not give you full control over the underlying model weights, which self-hosting would provide.

Answer

Lower latency always

Answer

No data egress costs

Answer

Full control over model weights

Question 9

Which TWO techniques can help reduce bias in LLM outputs?

Accepted Answer

Using diverse training data. Option C (Using diverse training data) is correct because bias in LLM outputs often stems from skewed or unrepresentative corpora; training on data that spans different demographics, cultures, languages, and viewpoints exposes the model to a broader distribution and reduces the tendency to overgeneralize from a narrow slice of the population. Option E (Applying adversarial debiasing) is correct because it is a recognized fairness technique in which an adversary tries to predict a protected attribute (e.g., gender or race) from the model's representations while the main model is trained to prevent that prediction, thereby minimizing encoded bias in the learned features. Option A (Setting temperature to 0) is not correct because temperature only controls sampling randomness and makes outputs more deterministic; it does not remove the underlying biases baked into the model's weights. Option B (Using only English data) is not correct because restricting training to a single language reduces linguistic and cultural diversity, which typically amplifies rather than reduces bias. Option D (Increasing model size) is not correct because larger models can capture more nuanced patterns but do not inherently eliminate bias and may even amplify stereotypes present in the training data.

Answer

Setting temperature to 0

Answer

Using only English data

Answer

Increasing model size

Question 10

Refer to the exhibit. What is the primary reason the response is incomplete?

Accepted Answer

The max-tokens limit is too low.. The response is incomplete because the max-tokens limit is too low, causing the model to truncate its output before completing the full answer. When the token budget is exhausted, the generation stops mid-sentence or mid-thought, leaving the response unfinished regardless of prompt length or other parameters.

Answer

The temperature is not set.

Answer

The model-id is incorrect.

Answer

The prompt is too short.

Question 11

Refer to the exhibit. What is the solution?

Accepted Answer

Use a different base model that supports fine-tuning.. The exhibit indicates that the base model does not support fine-tuning, which is a prerequisite for adapting a large language model to a specific task or domain. Using a different base model that supports fine-tuning allows the model to be customized through supervised learning on task-specific data, enabling it to learn new patterns and improve performance. This is the correct solution because without fine-tuning capability, the model cannot be effectively adapted regardless of other hyperparameter adjustments.

Answer

Change the learning rate.

Answer

Increase the training epochs.

Answer

Use a different compartment.

Question 12

Refer to the exhibit. A user in GenAI-Users group tries to run a text generation inference but gets permission denied. What is the most likely issue?

Accepted Answer

The operation condition is too restrictive.. The policy attached to the GenAI-Users group includes a condition that restricts the operation to a specific compartment or resource, but the user is attempting to run inference in a different compartment or without meeting the condition. Since the condition is too restrictive, the IAM policy denies the action even though the user is in the correct group and the resource type is valid.

Answer

The policy resource type is wrong.

Answer

The group name mismatch.

Answer

The user is not in the compartment.

Question 13

A company wants to use OCI Generative AI to summarize customer reviews. Which model parameter should be adjusted to control the creativity of the summary?

Accepted Answer

Temperature. Temperature controls the randomness of token selection in the model's output distribution. A higher temperature (e.g., 0.9) makes the summary more creative and diverse, while a lower temperature (e.g., 0.1) makes it more deterministic and focused. For summarizing customer reviews, adjusting temperature directly influences how novel or conservative the generated text will be.

Answer

Frequency penalty

Answer

Top-k

Answer

Presence penalty

Question 14

An engineer sets beam search width to 1 during inference on OCI Generative AI. What is the most likely effect on output?

Accepted Answer

Faster inference. Beam search width of 1 is equivalent to greedy decoding, where only the single most probable token is selected at each step. This eliminates the need to maintain and compare multiple candidate sequences, significantly reducing computational overhead and memory access, which directly speeds up inference.

Answer

More memory usage

Answer

More diverse outputs

Answer

Better quality

Question 15

Which technique allows an LLM to be adapted to a new task with only a few examples?

Accepted Answer

Few-shot learning. Few-shot learning (A) is the correct technique because it enables an LLM to adapt to a new task by providing a small number of input-output examples directly in the prompt, without updating the model's weights. This leverages the model's in-context learning ability, where it generalizes from the provided examples to perform the task. It is distinct from fine-tuning, which requires retraining on a labeled dataset.

Answer

Fine-tuning

Answer

Pre-training

Answer

Prompt engineering