1Z0-1127-25 Fundamentals of Large Language Models Practice Test 2 — 15 Questions

Question 1

During multi-turn conversation with an OCI GenAI model, the model repeats user messages from earlier turns. What is the most likely cause?

Accepted Answer

Low presence penalty. A low presence penalty reduces the model's incentive to avoid repeating previously mentioned content. In multi-turn conversations, this can cause the model to echo user messages from earlier turns because the penalty is too weak to discourage repetition of tokens that have already appeared in the context window.

Answer

Low top-p

Answer

High temperature

Answer

High frequency penalty

Question 2

Refer to the exhibit. A user in group GenAIGroup cannot see models in the Production compartment using OCI Generative AI. What is the most likely issue?

Accepted Answer

Resource type should be 'generative-ai-model'. The resource type in the policy statement must match the exact OCI resource type for Generative AI models, which is 'generative-ai-model'. Using a generic or incorrect resource type (e.g., 'generative-ai' or 'model') will cause the policy to not apply, preventing the user from seeing models even if they have the appropriate permissions.

Answer

Statement should be 'use' instead of 'read'

Answer

Missing 'inspect' permission

Answer

Policy syntax incorrect (missing quotes)

Question 3

A data scientist is using OCI Data Science to fine-tune a Cohere command model on domain-specific documents. They observe that the fine-tuned model generates repetitive text. What is the most likely cause?

Accepted Answer

The training dataset lacked diversity.. Repetitive text in fine-tuned models is a classic symptom of overfitting to a narrow or homogeneous training dataset. When the domain-specific documents lack diversity in phrasing, topics, or contexts, the model learns to latch onto the most common patterns and repeats them, rather than generalizing. This is not a hyperparameter tuning issue but a data quality issue.

Answer

The number of epochs was insufficient.

Answer

The learning rate was too high.

Answer

The batch size was too small.

Question 4

An OCI AI Language text classification request returns the output shown. Which conclusion is most accurate?

Accepted Answer

The text is classified as Positive with high confidence.. The output shows a sentiment label of 'Positive' with a confidence score of 0.98, indicating the model is highly confident in its classification. Option B correctly identifies this as a positive sentiment with high confidence, which is the most accurate conclusion based on the provided data.

Answer

The model is uncertain about the sentiment.

Answer

The API endpoint is misconfigured.

Answer

The --endpoint parameter is optional.

Question 5

During fine-tuning of a Cohere model on OCI Data Science, the loss curve shows a sharp spike after epoch 3. What is the most appropriate action?

Accepted Answer

Gradient clipping.. A sharp spike in the loss curve after epoch 3 during fine-tuning indicates a gradient explosion, where the gradients become excessively large and destabilize the model's weights. Gradient clipping is the most appropriate action because it directly caps the gradient norm (e.g., using `max_grad_norm=1.0` in Cohere's fine-tuning API) to prevent these spikes, ensuring stable training without altering the learning dynamics.

Answer

Reduce learning rate.

Answer

Add more training data.

Answer

Increase batch size.

Question 6

A team uses OCI Generative AI's summarization feature to condense legal documents. The summaries sometimes omit critical clauses. Which parameter adjustment is most likely to improve completeness?

Accepted Answer

Increase maxTokens.. Increasing maxTokens (option D) is the most direct way to improve completeness because it extends the maximum length of the generated summary, allowing the model to include more content from the source legal document. Critical clauses are often omitted when the token limit truncates the output before the model can cover all essential sections. This parameter controls the output length, not the style or randomness of the generation.

Answer

Adjust frequencyPenalty.

Answer

Increase temperature.

Answer

Decrease topP.

Question 7

Which OCI service provides pre-trained models for custom text classification without requiring fine-tuning?

Accepted Answer

OCI AI Language. B is correct because OCI AI Language provides pre-trained models that can perform custom text classification out-of-the-box without requiring fine-tuning. It offers built-in models for common NLP tasks like sentiment analysis, entity extraction, and text classification, allowing users to classify text into custom categories defined by their own labels without additional training.

Answer

OCI Generative AI

Answer

OCI Data Science

Answer

OCI Vision

Question 8

A user has a prompt that exceeds the model's token limit. What is the best practice to handle this?

Accepted Answer

Summarize the earlier parts of the prompt and include the summary.. When a prompt exceeds the model's token limit, the best practice is to summarize the earlier parts of the prompt and include the summary. This preserves the essential context without exceeding the token limit, as the model's context window is fixed (e.g., 4,096 tokens for GPT-3.5 or 8,192 for GPT-4). Summarization reduces token count while retaining key information, enabling the model to process the entire input within its constraints.

Answer

Increase the max tokens parameter in the API call.

Answer

Truncate the prompt and hope the model understands.

Answer

Split the input into multiple calls and merge results.

Question 9

An application using OCI GenAI experiences high response times. Which change will most directly reduce latency?

Accepted Answer

Reduce the number of output tokens requested.. Reducing the number of output tokens requested directly decreases the amount of text the model must generate per request. Since latency in OCI GenAI is proportional to the total number of tokens processed (input + output), fewer output tokens mean fewer autoregressive decoding steps, which directly reduces response time.

Answer

Increase the max tokens parameter.

Answer

Switch to a fine-tuned version of the same model.

Answer

Enable batched inference.

Question 10

Which three factors most significantly affect the quality of an LLM's output? (Select THREE)

Accepted Answer

Clarity of the prompt. Option B (Clarity of the prompt) is correct because the prompt is the primary instruction the model conditions on; ambiguous, underspecified, or poorly structured prompts lead to vague or off-target completions, while clear prompts with explicit task, format, and constraints yield more accurate and relevant output. Option D (Temperature setting) is correct because temperature directly controls the sampling distribution's randomness: low values (e.g., 0.0-0.3) make output more deterministic and factual, while high values (e.g., 0.8-1.2) increase diversity and creativity, so the setting materially shapes output quality for a given task. Option E (Quality of training data) is correct because the model's capabilities, factual grounding, and biases are learned from its pretraining and fine-tuning corpora; noisy, outdated, or skewed data degrades accuracy and coherence regardless of how the model is prompted. Option A (Model's context window size) is not among the marked answers: while it limits how much input the model can attend to, it is a capacity constraint rather than a direct determinant of output quality for typical prompts. Option C (Number of GPUs used during inference) is not marked because inference hardware affects latency, throughput, and cost, not the intrinsic quality of the generated text.

Answer

Model's context window size

Answer

Number of GPUs used during inference

Question 11

Which three characteristics of LLMs can lead to hallucinations? (Select THREE)

Accepted Answer

Ability to generate plausible-sounding text. Option B is correct because LLMs are optimized to produce fluent, statistically plausible continuations, so they can generate convincing text that is factually wrong—this plausibility is exactly what makes hallucinations hard to detect. Option C is correct because LLMs have no grounded connection to the real world (no live sensors, databases, or verified knowledge sources); they operate purely on learned token patterns, so outputs are not anchored to verified facts. Option D is correct because when the training corpus lacks coverage of a topic or contains sparse/outdated information, the model still generates an answer by interpolating patterns, producing confident but unsupported content. Option A is not among the marked correct answers: overconfidence describes a model's tone/calibration rather than a root cause of hallucination, and it is not one of the three selected characteristics. Option E is not correct because a large vocabulary size affects tokenization and output diversity, not the tendency to fabricate facts.

Answer

Overconfidence in predictions

Answer

Large vocabulary size

Question 12

A machine learning team is fine-tuning a 7B parameter Llama 2 model on a custom dataset of 10,000 documents using OCI Data Science and GPU instances. They encounter out-of-memory (OOM) errors during the fine-tuning process. They are using a batch size of 8 and a sequence length of 2048. They cannot increase the GPU memory. Which change should they prioritize to resolve the OOM?

Accepted Answer

Enable gradient accumulation with steps of 4 or more.. The correct option is A: enabling gradient accumulation with steps of 4 or more. Gradient accumulation lets the team keep the effective batch size while processing smaller micro-batches sequentially, so the peak activation memory per step drops and the OOM is avoided without adding GPU memory. Mixed precision (B) can reduce memory somewhat, but it is not the most direct fix for a batch-size-driven OOM and may still fail at batch size 8 with sequence length 2048. Reducing to a 3B model (C) changes the model and may hurt quality, while decreasing epochs (D) only shortens training and does not reduce per-step memory.

Answer

Use mixed precision training (FP16).

Answer

Reduce the model size by using a 3B parameter version.

Answer

Decrease the number of training epochs.

Question 13

An e-commerce company fine-tuned a Cohere Command model on their product catalog to generate product descriptions. During inference, they notice the model outputs are too repetitive: it often repeats similar phrases across different products, and the descriptions lack diversity. The team wants to increase the variety of the generated text without sacrificing relevance. They are currently using temperature=0.8, top_p=0.9, frequency_penalty=0, and presence_penalty=0. Which parameter adjustment should they make to most effectively increase diversity?

Accepted Answer

Increase top_p from 0.9 to 0.95.. Increasing top_p from 0.9 to 0.95 expands the nucleus of tokens considered during sampling, allowing the model to select from a wider set of plausible next tokens. This directly increases output diversity while still maintaining relevance, as tokens outside the top 90% probability mass are now included. The current settings already have moderate temperature and no penalties, so broadening top_p is the most effective single adjustment to reduce repetitiveness.

Answer

Decrease temperature from 0.8 to 0.5.

Answer

Set frequency_penalty to a negative value (e.g., -0.5).

Answer

Increase max_tokens from 200 to 500.

Question 14

A company is using OCI GenAI with a Dedicated AI Cluster to serve a large language model for real-time chat applications. They notice high inference latency (average 2 seconds per response) and want to reduce it to under 500 milliseconds without significantly degrading the quality of responses. The cluster is configured with NVIDIA A100 GPUs. The model is the base Cohere Command model (52B parameters). They have explored increasing batch size, but that increases latency for interactive use cases. Which action should they take?

Accepted Answer

Deploy the model with inference optimization frameworks like vLLM, TensorRT, or ONNX Runtime.. Inference optimization frameworks like vLLM, TensorRT, and ONNX Runtime are specifically designed to reduce latency for large language models on NVIDIA A100 GPUs. These frameworks use techniques such as PagedAttention (vLLM), kernel fusion, and graph optimization to significantly lower per-request latency without degrading output quality, making them ideal for real-time chat applications where sub-500ms responses are required.

Answer

Increase batch size to process multiple queries at once.

Answer

Swap the model to a smaller variant, such as Cohere Command Light (6B).

Answer

Enable model quantization (e.g., int8) to reduce memory and computation.

Question 15

A developer is testing the OCI Generative AI API by sending a request to generate text using the Cohere Command R model. The request returns the following error: 'The model 'cohere.command-r-08-2024' is not available in this region. Please check the model availability in your region.' The developer is using the us-ashburn-1 region. What is the most likely cause of this error?

Accepted Answer

The model is not deployed in the us-ashburn-1 region.. The error message explicitly states that the model 'cohere.command-r-08-2024' is not available in the region. OCI Generative AI models are deployed regionally, and the Cohere Command R model is not available in the us-ashburn-1 (Ashburn) region. The developer must select a supported region, such as us-chicago-1, where this model is deployed.

Answer

The request body format is incorrect.

Answer

The model name is misspelled (e.g., 'cohere.command-r-08-2024' vs 'cohere.command-r-08-2024').

Answer

The API key used in the request is invalid.