1Z0-1127-25 Practice Test 36 — 15 Questions

Question 1

A RAG application is hallucinating because the LLM receives irrelevant context from the retrieval step, even when topK is set to 3. Which strategy would best reduce hallucination by improving the relevance of retrieved documents?

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

Accepted Answer

Add a reranking step after retrieval to select the most relevant chunks. Adding a reranking step after retrieval directly addresses the core issue: even with a low topK, the initial retrieval may return chunks that are semantically similar but not precisely relevant to the query. Reranking uses a cross-encoder model to score each retrieved chunk against the query, reordering them so that only the most contextually relevant chunks are passed to the LLM. This reduces the chance of the LLM receiving irrelevant context, thereby minimizing hallucination.

Answer

Reduce the chunk size to one sentence per chunk

Answer

Implement a query rewriting mechanism

Answer

Increase topK to 10 to provide more context

Question 2

Which TWO factors are most likely to cause hallucinations in LLMs?

Accepted Answer

High temperature. Option A (High temperature) is correct because temperature scales the sampling distribution's logits before softmax; higher values flatten the distribution, increasing the probability of selecting low-probability tokens and thus producing fabricated or incoherent content. Option E (Inadequate training data) is correct because sparse, biased, or outdated corpora leave the model without grounded knowledge, so it fills gaps by generating plausible-sounding but false statements. Option B (Short context window) is not a primary hallucination cause; it limits how much input the model can attend to but does not inherently make outputs factually false. Option C (Excessive fine-tuning) can cause overfitting or catastrophic forgetting, but that is a distinct degradation rather than a canonical hallucination driver. Option D (Low top-p) is incorrect because restricting nucleus sampling to a smaller cumulative probability mass makes output more conservative and deterministic, reducing rather than increasing hallucination.

Answer

Short context window

Answer

Excessive fine-tuning

Answer

Low top-p

Question 3

Which TWO of the following are common applications of large language models in enterprise settings?

Accepted Answer

Summarizing lengthy legal documents.. Option A is correct because large language models excel at natural language understanding and generation tasks such as abstractive summarization, which is widely used in enterprise legal, compliance, and contract-review workflows to condense lengthy documents. Option C is correct because LLMs are commonly applied in developer tooling (e.g., GitHub Copilot, Amazon CodeWhisperer) to translate natural language prompts into boilerplate code, unit tests, and scaffolding. Option B is not a typical LLM application: real-time audio signal processing relies on DSP algorithms and streaming frameworks (e.g., FFT, WebRTC), not transformer text models. Option D is incorrect because relational databases provide ACID transactions, schemas, and indexing that LLMs cannot replace; LLMs may assist with query generation but not storage. Option E is incorrect because image super-resolution is a computer-vision task handled by convolutional or diffusion models (e.g., SRCNN, ESRGAN), not by large language models.

Answer

Performing real-time signal processing for audio streams.

Answer

Replacing relational databases for data storage.

Answer

Enhancing low-resolution images through super-resolution.

Question 4

Refer to the exhibit. The user requested a long story but the response is cut short. What is the most likely cause?

Accepted Answer

The max_tokens parameter is too low for the requested length.. The `max_tokens` parameter explicitly limits the number of tokens (words/subwords) the model can generate in a single response. If the user requests a long story but `max_tokens` is set too low, the model will stop generating once the token limit is reached, resulting in a truncated output. This is a common configuration issue when using OCI Generative AI Service or any large language model API.

Answer

The model content filter blocked part of the output.

Answer

There was a network error during inference.

Answer

The model is not capable of generating long stories.

Question 5

A company wants to deploy a fine-tuned model for real-time inference with consistent low latency. They are evaluating dedicated AI clusters. Which THREE factors should they consider when provisioning the cluster?

Accepted Answer

Number of clusters (for high availability). Option B is correct because provisioning multiple dedicated AI clusters enables high availability, so if one cluster becomes unavailable the inference endpoint can still serve real-time requests without a single point of failure. Option C is correct because the region where the cluster is provisioned determines network proximity to users and data-residency/compliance constraints, both of which directly affect consistent low latency for real-time inference. Option D is correct because the number of model units per cluster sets the compute capacity and concurrency available for serving the fine-tuned model, which must be sized to sustain low-latency throughput under expected load. Option A is not relevant because fine-tuning job timeout settings govern the training run, not the provisioning or latency behavior of a dedicated inference cluster. Option E is not relevant because the base model is already fixed by the fine-tuning process and does not determine how the dedicated inference cluster is provisioned.

Answer

Fine-tuning job timeout settings

Answer

The base model used for fine-tuning

Question 6

Which OCI Generative AI model family is optimized for generating text embeddings that capture semantic meaning for tasks like clustering and classification?

Accepted Answer

Cohere Embed. Cohere Embed models are specifically designed for embedding text into vectors. Cohere Command and Meta Llama are generative models, not embedding models.

Answer

Cohere Rerank

Answer

Meta Llama 3

Answer

Cohere Command R

Question 7

A company wants to use OCI Generative AI to build a multilingual customer support chatbot. They need to understand customer queries in multiple languages and generate responses in the same language. Which TWO actions should they take? (Choose two.)

Accepted Answer

Select Cohere Command R as the base model for chat. Option C is correct because Cohere Command R is a multilingual generative chat model available in OCI Generative AI that can understand queries and generate responses across many languages, directly supporting the chatbot's conversational needs. Option D is correct because embed-multilingual-v3.0 produces vector embeddings that capture semantic meaning across multiple languages, enabling multilingual semantic search and retrieval (e.g., RAG) so queries in any supported language can match relevant knowledge. Option A is wrong because embed-english-v3.0 is English-only and would fail to represent non-English customer queries accurately. Option B is not required and is not one of the marked answers, since fine-tuning Llama 3 on multilingual data is unnecessary when a natively multilingual model like Command R is available. Option E is wrong because the Summarisation API condenses text and does not provide multilingual query understanding or same-language response generation for a chatbot.

Answer

Use the embed-english-v3.0 model for embedding queries

Answer

Fine-tune Meta Llama 3 on multilingual data

Answer

Use the Summarisation API for each language

Question 8

An OCI user observes that their Mistral model produces very repetitive text when temperature is set to 0.9 and top-p to 1.0. Which adjustment is most likely to reduce repetition?

Accepted Answer

Enable frequency penalty. Enabling frequency penalty directly reduces the likelihood of the model repeating the same tokens by subtracting a penalty proportional to the token's existing frequency in the generated text. This is the most targeted way to combat repetitive output without altering the core sampling parameters (temperature and top-p) that control randomness.

Answer

Decrease top-p to 0.5

Answer

Increase temperature to 1.5

Answer

Set top-k to 1

Question 9

An application uses an LLM to summarize legal documents. The summaries sometimes include hallucinations (details not in the original text). Which prompt engineering technique is MOST effective at reducing hallucinations?

Accepted Answer

Include the full document text in the prompt and instruct the model to base its summary only on that text. Including the full source document in the prompt and instructing the model to base its summary only on that text is the most effective way to reduce hallucinations. This technique, often called grounding or retrieval-augmented generation when combined with retrieval, constrains the model to the provided context. It directly addresses the root cause: the model inventing details not present in the input.

Answer

Increase the temperature to 0.9 to make the model more cautious

Answer

Use a few-shot prompt with examples of correct summaries

Answer

Set the presence penalty to a high value

Question 10

A company requires a generative AI service to automatically summarize customer support transcripts. Which OCI Generative AI model is most suitable for this task?

Accepted Answer

Cohere Command. Cohere Command is a large language model specifically designed for text generation tasks such as summarization, making it the most suitable choice for automatically summarizing customer support transcripts. Unlike embedding models or base Llama variants, Command is optimized for instruction-following and generating coherent, concise summaries from conversational data.

Answer

Llama 3 70B

Answer

Cohere Embed

Answer

Fine-tuned Llama 2

Question 11

Refer to the exhibit. Why did the embedding creation fail?

Accepted Answer

The API call was not properly authenticated. The 401 status code in the exhibit indicates an authentication failure. The API call requires a valid OAuth 2.0 token or API key in the Authorization header; without proper authentication, the embedding creation request is rejected regardless of input length, model availability, or region enablement.

Answer

The input text is too short

Answer

The model ID is not available in the us-ashburn-1 region

Answer

The region is not enabled for the Generative AI service

Question 12

An administrator runs the above CLI command to check the status of a dedicated AI cluster. The cluster is ACTIVE with capacity 10. However, a user reports that inference requests to this cluster are failing with a '429 Too Many Requests' error. What is the most likely cause?

Accepted Answer

The cluster is hitting the maximum inference requests per minute limit. The '429 Too Many Requests' error is an HTTP status code indicating rate limiting has been exceeded. In OCI Generative AI, dedicated AI clusters have a configurable 'maximum inference requests per minute' limit. Even if the cluster is ACTIVE and has capacity (e.g., 10 nodes), hitting this per-minute request cap will cause the API gateway to reject further requests with a 429 error. The administrator must increase the rate limit or implement client-side throttling to resolve this.

Answer

The cluster does not have enough nodes to handle the load

Answer

The user is not in the same compartment as the cluster

Answer

The cluster is not in ACTIVE state

Question 13

A team is prompting a model to translate English to French. They notice the model occasionally mixes French with other Romance languages. Which prompt refinement strategy would MOST likely improve language consistency?

Accepted Answer

Add a constraint in the system prompt: 'Translate only to French. Do not use any other language.'. The core issue is that the model is drifting into other Romance languages, which is a language-consistency problem best solved by an explicit instruction constraint. Adding a system-prompt directive such as 'Translate only to French. Do not use any other language.' directly narrows the model's output distribution toward French tokens and is the most reliable, low-cost refinement. System prompts are weighted more heavily than user prompts in most LLM APIs, so this constraint is respected across turns.

Answer

Use a stop sequence to halt after the first sentence

Answer

Lower top-k to 1 for deterministic output

Answer

Increase temperature to encourage more focused output

Question 14

A prompt engineer wants to generate a JSON object from an LLM response. Which prompt component is MOST critical to ensure the output is valid JSON?

Accepted Answer

Output format specification. The output format specification is the component that explicitly tells the model to return valid JSON — for example, 'Respond only with a JSON object with keys X, Y, Z.' Without an explicit format directive, the model may wrap JSON in prose, use markdown fences, or produce malformed syntax. While examples can demonstrate format, the explicit format specification is the most critical and direct control for guaranteeing valid JSON.

Answer

Context and background

Answer

Task instruction

Answer

Examples (few-shot)

Question 15

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