1Z0-1127-25 Prompt Engineering Practice Test 4 — 15 Questions

Question 1

In OCI Generative AI, when using the Cohere Command model, which parameter is used to discourage the model from repeating the same phrases?

Accepted Answer

frequency_penalty. In OCI Generative AI, the frequency_penalty parameter is used to discourage the model from repeating the same phrases by penalizing tokens based on how often they have already appeared in the generated text. A higher frequency_penalty reduces repetition.

Answer

presence_penalty

Answer

temperature

Answer

top-k

Question 2

A developer wants the model to generate a bullet list of steps to troubleshoot a network issue. Which output format specification should be included in the prompt?

Accepted Answer

Request 'bullet list' in the output format. To get a bullet list of troubleshooting steps, the prompt should explicitly request a 'bullet list' in the output format specification. Explicit format instructions are the most direct and reliable way to control the model's output structure, and they work regardless of temperature or examples.

Answer

Use a few-shot example with numbered steps

Answer

Provide a table format

Answer

Set temperature to 0

Question 3

An organization maintains a library of prompt templates for various use cases. Which three practices are essential for effective prompt management? (Choose three.)

Accepted Answer

Store prompt templates in a shared, accessible repository. Option B is correct because storing prompt templates in a shared, accessible repository centralizes them so teams can reuse, review, and update prompts consistently instead of duplicating them in silos. Option C is correct because defining evaluation criteria gives objective, repeatable metrics (such as accuracy, relevance, or task success rate) to judge whether a prompt actually performs well. Option E is correct because versioning each prompt template and tracking changes provides an audit trail, enables rollback to a known-good prompt, and supports reproducibility across model or data updates. Option A is not essential to prompt management itself, since temperature and top-p are inference-time generation parameters that may be tuned per use case but are not required for managing the template library. Option D is also not essential, because A/B testing every single prompt change is impractical and unnecessary; testing should be risk-based rather than mandatory for all edits.

Answer

Adjust temperature and top-p for each prompt variant

Answer

Conduct A/B testing on every prompt change

Question 4

A prompt engineer is troubleshooting a chatbot that consistently fails to follow instructions when the user includes adversarial input. Which two strategies can mitigate prompt injection attacks? (Choose two.)

Accepted Answer

Use instruction shielding: clearly separate system instructions from user input. Option B is correct because instruction shielding (e.g., delimiting system prompts with tags like ... or using role separation) helps the model distinguish trusted developer instructions from untrusted user input, making it harder for injected text to override the original directives. Option E is correct because input validation and sanitization can detect and strip or neutralize known adversarial patterns (e.g., phrases like "ignore previous instructions" or encoded payloads) before they reach the model, reducing the attack surface. Option A is incorrect because raising temperature increases randomness and does not prevent injection; it can actually make behavior less reliable. Option C is incorrect because using a smaller model does not inherently mitigate prompt injection and may reduce instruction-following ability. Option D is incorrect because adding few-shot examples with safe outputs does not reliably block adversarial inputs and may even be exploited if examples are not carefully isolated.

Answer

Increase temperature to make model less predictable

Answer

Use a smaller model to reduce capability

Answer

Add more few-shot examples with safe outputs

Question 5

What is the primary purpose of setting the 'top-k' parameter in an LLM?

Accepted Answer

To limit the next token selection to the k most probable tokens. The top-k parameter restricts the model's next-token selection to only the k most probable tokens in the vocabulary distribution. All other tokens are assigned zero probability, and the model samples from this truncated set. This prevents the model from selecting extremely low-probability tokens that could produce incoherent output.

Answer

To truncate the output after k tokens

Answer

To control the creativity of the output by scaling logits

Answer

To penalize tokens that have already appeared in the generated text

Question 6

A developer is using a Cohere Command model via OCI Generative AI. They want the model to generate responses strictly in JSON format for a specific task, but the model sometimes outputs additional explanatory text. Which prompt engineering technique is MOST effective?

Accepted Answer

Add a preamble: 'You are a JSON generator. Output only valid JSON. Do not include any other text.'. A system-level preamble that explicitly instructs the model to act as a JSON generator and to output only valid JSON is the most effective technique because it sets a persistent behavioral constraint at the start of the prompt. Cohere Command models, like most instruction-tuned LLMs, weight system/role instructions heavily and will suppress extraneous explanatory text when told to do so. This directly addresses the root cause (the model's tendency to add commentary) rather than trying to constrain it after the fact.

Answer

Include a single example of a JSON output in the user message

Answer

Set the temperature to 0.0 to make the model deterministic

Answer

Use a stop sequence '}' to force the model to stop after the JSON

Question 7

A data scientist wants to generate a concise summary of a long legal document. The model should output a bullet list of key points. Which prompt component is LEAST important for this task?

Accepted Answer

Few-shot examples of summaries. Few-shot examples are the least important component here because the task — summarizing a legal document into bullet points — is a straightforward, well-understood instruction that modern LLMs can execute reliably from a clear task description alone. The context (the document itself) is essential input, the output format specification ensures the bullet-list structure, and the task instruction defines the objective. Few-shot examples add token overhead and complexity without materially improving output quality for such a simple, explicit task.

Answer

Context/background (the legal document text)

Answer

Output format specification ('Output as a bullet list')

Answer

Task instruction ('Summarize the following legal document in bullet points')

Question 8

Which prompting technique involves generating multiple independent reasoning paths and then selecting the most common answer?

Accepted Answer

Self-consistency prompting. Self-consistency prompting generates multiple independent reasoning paths (typically via chain-of-thought sampling at higher temperature) and then selects the most common final answer through majority voting. This ensemble-like approach improves accuracy on reasoning tasks by reducing the impact of any single flawed reasoning chain. It was introduced by Wang et al. (2022) as an enhancement to chain-of-thought prompting.

Answer

Chain-of-thought prompting

Answer

Few-shot prompting

Answer

Zero-shot prompting

Question 9

An OCI Generative AI user notices that a model generates repetitive phrases when summarizing technical articles. Which parameter adjustment is MOST likely to reduce this repetition?

Accepted Answer

Increase the frequency penalty. Frequency penalty reduces the likelihood of repeating tokens that have already appeared, directly targeting repetition. Presence penalty also helps but frequency penalty is stronger for repeated phrases.

Answer

Decrease max tokens

Answer

Set top-p to 0.95

Answer

Increase temperature to 0.9

Question 10

What is the main advantage of using chain-of-thought (CoT) prompting over standard few-shot prompting for complex reasoning tasks?

Accepted Answer

CoT improves performance on multi-step reasoning tasks by generating step-by-step explanations. Chain-of-thought prompting improves performance on complex reasoning tasks by eliciting intermediate reasoning steps from the model. By generating step-by-step explanations, the model can decompose multi-step problems into manageable sub-problems, reducing errors that occur when jumping directly to a final answer. This is especially effective for math, logic, and multi-hop reasoning.

Answer

CoT reduces the number of tokens needed in the prompt

Answer

CoT allows the model to use any tools available in the environment

Answer

CoT eliminates the need for any examples in the prompt

Question 11

A prompt engineer is testing two versions of a prompt for a content generation task. They want to measure which version produces more factual and concise outputs. Which evaluation approach is BEST?

Accepted Answer

Run A/B tests on a diverse set of inputs and score outputs based on predefined criteria. Running A/B tests on a diverse set of inputs and scoring outputs based on predefined criteria (e.g., factuality, conciseness) provides a systematic and objective comparison. This approach reduces bias and ensures that the evaluation is representative of the task. It is the best practice for prompt evaluation.

Answer

Use only the first output from each prompt and manually compare

Answer

Ask the model to self-evaluate its outputs

Answer

Increase the temperature to see which prompt handles randomness better

Question 12

Which of the following is a common prompt injection vulnerability?

Accepted Answer

User input that contains 'Ignore previous instructions' followed by malicious commands. Prompt injection occurs when untrusted user input is interpreted by the model as instructions rather than data, allowing an attacker to override the system prompt. The classic payload 'Ignore previous instructions' followed by malicious commands directly hijacks the model's instruction-following behavior, which is the defining characteristic of a prompt injection vulnerability. This is a well-documented OWASP LLM Top 10 risk (LLM01).

Answer

Including too many few-shot examples

Answer

Setting temperature too high

Answer

Using a system prompt that is too long

Question 13

A developer uses a Cohere Command model with a preamble: 'You are a helpful assistant that answers questions based on the provided document.' They also include a user message with a query and a relevant document excerpt. However, the model often ignores the document and uses its own knowledge. Which improvement is MOST effective?

Accepted Answer

Use the document-grounded generation format with and tags around the document. Using the document-grounded generation format with and tags around the document is the most effective improvement because it explicitly signals to the model that the enclosed text is the source document to ground its response on. This format is recommended by Cohere for retrieval-augmented generation (RAG) to ensure the model prioritizes the provided document over its parametric knowledge.

Answer

Increase the temperature to 0.7 to reduce reliance on parametric knowledge

Answer

Set frequency penalty to 0.5

Answer

Remove the preamble and rely solely on the user message

Question 14

In prompt engineering, what is the role of a system prompt?

Accepted Answer

To define the model's persona, tone, and behavioral constraints. The system prompt in prompt engineering defines the model's persona, tone, and behavioral constraints, setting the overarching context for all interactions. This is distinct from user prompts, which carry the current query. By establishing these parameters, the system prompt guides the model's responses consistently across a session, ensuring alignment with desired output characteristics.

Answer

To handle error messages from the API

Answer

To specify the user's current query

Answer

To provide examples of desired outputs

Question 15

A team wants to use a prompt pattern that combines reasoning with acting, where the model alternates between thinking and taking actions (e.g., tool calls). Which pattern fits this requirement?

Accepted Answer

ReAct pattern. The ReAct pattern (Reason + Act) prompts the model to output reasoning steps and then perform actions (like API calls), enabling interactive problem-solving.

Answer

Template pattern

Answer

Role prompting

Answer

Recipe pattern