991 questions total · 14pages · All types, answers revealed
Page 14 of 14
A user repeatedly gets the same phrase output by the model. Which parameter adjustment is MOST likely to reduce such repetitive patterns?
Frequency penalty penalizes tokens that have been used, reducing repetition.
Why this answer
Frequency penalty reduces the likelihood of repeating tokens that have already appeared, directly combating repetition.
Which component of the transformer architecture allows the model to weigh the importance of different words in a sentence when processing input?
Self-attention computes pairwise relevance scores and produces context-aware representations for each token.
Why this answer
The self-attention mechanism is the core component of the transformer architecture that enables the model to dynamically assign weights to each word in a sentence relative to every other word. This allows the model to capture contextual relationships and dependencies, such as determining which words are most relevant to the current word being processed, regardless of their distance in the sequence.
Exam trap
Cisco often tests the distinction between components that provide positional information (positional encoding) versus those that compute relational importance (self-attention), leading candidates to confuse the role of positional encoding with the weighting of word significance.
How to eliminate wrong answers
Option A is wrong because layer normalization stabilizes training by normalizing activations across features, not by weighing word importance. Option B is wrong because positional encoding adds information about the order of words in a sequence, but does not perform any weighting of importance. Option D is wrong because the feed-forward neural network applies non-linear transformations to each position independently after attention has been computed, and does not weigh word importance.
What is the primary purpose of an embedding model in a RAG pipeline?
Embedding models encode text semantically into vectors.
Why this answer
The primary purpose of an embedding model in a RAG pipeline is to convert text into numerical vectors (embeddings) that capture semantic meaning. These vectors enable the retrieval component to efficiently find relevant documents by measuring similarity (e.g., cosine similarity) between the query and stored document embeddings. Without this conversion, the system cannot perform semantic search over unstructured text.
Exam trap
Cisco often tests the distinction between the embedding model's role (conversion to vectors) and the LLM's role (generation), so candidates may mistakenly attribute response generation or summarization to the embedding model.
How to eliminate wrong answers
Option B is wrong because generating human-like responses is the role of the large language model (LLM) in the generation step, not the embedding model. Option C is wrong because ranking search results is typically performed by a reranker or the retrieval algorithm (e.g., using vector similarity scores), not by the embedding model itself. Option D is wrong because summarizing long documents is a task for the LLM or a dedicated summarization model, not the embedding model, which only produces vector representations.
An AI team is fine-tuning a large language model using OCI Data Science and plans to deploy the fine-tuned model using the Generative AI service's custom model deployment. What is the required format for the model artifacts?
The custom model deployment requires a zip archive with all necessary files.
Why this answer
The OCI Generative AI service requires custom model artifacts to be packaged as a .zip archive containing the model weights, configuration files (e.g., config.json, tokenizer files), and any necessary inference code. This format ensures the service can extract and load the model correctly into its managed inference infrastructure, aligning with the standard Hugging Face model repository structure.
Exam trap
The trap here is that candidates may confuse OCI Generative AI's custom model deployment with OCI Data Science model deployment, which does support Docker images, leading them to incorrectly select Option C.
How to eliminate wrong answers
Option A is wrong because a Git repository URL is not a supported artifact format for OCI Generative AI custom model deployment; the service expects a static artifact file, not a live repository reference. Option B is wrong because a single .pth file contains only PyTorch model weights without the required configuration files (e.g., config.json, tokenizer.json) and inference code, making it incomplete for deployment. Option C is wrong because OCI Generative AI custom model deployment does not accept Docker images; it uses a serverless, managed inference environment that expects a .zip archive of model artifacts, not a containerized application.
A team fine-tuned a Cohere Command R model in OCI GenAI and validated it. They now need to deploy it for production inference with a dedicated endpoint. What is the correct sequence of steps?
The correct order: allocate cluster, deploy model, then expose via endpoint.
Why this answer
Option D is correct because in OCI Generative AI, the correct sequence for deploying a fine-tuned model to a dedicated endpoint is: first provision a Dedicated AI Cluster (which provides the isolated compute infrastructure), then deploy the model to that cluster, and finally create an endpoint that exposes the deployed model for inference. This ensures the model is hosted on dedicated resources before the endpoint is created.
Exam trap
The trap here is that candidates often confuse the order of provisioning infrastructure versus deploying the model, mistakenly thinking the endpoint can be created first and then attached to a cluster later, but OCI GenAI requires the cluster to exist and the model to be deployed before the endpoint can be created.
How to eliminate wrong answers
Option A is wrong because it attempts to create an endpoint before the Dedicated AI Cluster is provisioned, which would fail as the endpoint requires an existing cluster to attach to. Option B is wrong because it incorrectly suggests deploying to a shared endpoint first, which is not the intended path for dedicated production inference; dedicated endpoints require a Dedicated AI Cluster, not a shared one. Option C is wrong because it creates an endpoint before the Dedicated AI Cluster is provisioned and before the model is deployed, which is invalid since the endpoint must reference an already deployed model on an existing cluster.
A developer implements a RAG chatbot using OCI Generative AI with streaming enabled. The chatbot fails to remember earlier conversation turns during a session. What is the most likely cause?
Session memory requires the client to send the conversation history in the messages list.
Why this answer
Option C is correct because a RAG chatbot with streaming enabled still requires the application to manage conversation state by including previous messages in each request. The OCI Generative AI streaming endpoint processes each request independently; without explicitly passing the conversation history, the model has no context of prior turns, causing it to fail to remember earlier interactions.
Exam trap
Cisco often tests the misconception that streaming endpoints inherently preserve conversation history, when in fact they are stateless and require explicit inclusion of prior messages in each request.
How to eliminate wrong answers
Option A is wrong because max_tokens controls the maximum length of the generated response, not the retention of conversation history; a low max_tokens might truncate output but does not cause loss of memory across turns. Option B is wrong because the OCI Generative AI streaming endpoint does support conversation history when the application includes previous messages in the request; streaming itself does not disable context. Option D is wrong because temperature affects the randomness of token selection, not the model's ability to recall prior context; a high temperature might produce more varied responses but does not erase conversation memory.
A retail company uses OCI Generative AI Service to build a RAG chatbot for product recommendations. The chatbot should consider both the user's query and the retrieved product descriptions. Which component of the RAG pipeline is responsible for combining these inputs before sending to the LLM?
Merges user query and context into a single prompt.
Why this answer
The prompt template is the component in a RAG pipeline that structures the final input to the LLM by combining the user's query with the retrieved product descriptions. It defines the format and instructions (e.g., 'Based on these product descriptions, recommend...') that the LLM uses to generate a coherent response. Without a prompt template, the raw query and documents would be sent without context, leading to poor or irrelevant outputs.
Exam trap
Oracle often tests the misconception that the embedding model or retriever handles input combination, when in fact those components only deal with vector representation and retrieval, not prompt assembly.
How to eliminate wrong answers
Option A is wrong because a reranker reorders retrieved documents based on relevance scores after initial retrieval, but it does not combine inputs with the user query for the LLM. Option B is wrong because the document retriever fetches relevant documents from the vector store using similarity search, but it does not merge them with the query into a single prompt. Option C is wrong because the embedding model converts text into vector representations for search, but it plays no role in assembling the final input to the LLM.
Which LangChain document loader would be most appropriate to load content from a public website for inclusion in a knowledge base?
WebBaseLoader fetches content from a given URL and loads it as a Document.
Why this answer
WebBaseLoader is specifically designed to load documents from web URLs, fetching the HTML content and converting it to LangChain Document objects. PDFLoader, CSVLoader, and TextLoader are for local files of specific formats.
An organization wants to use OCI Generative AI for summarizing long legal documents. Which OCI Generative AI service component is specifically designed for this task?
Summarisation is a dedicated component for summarizing documents.
Why this answer
Option D is correct because the Summarisation API in OCI Generative AI is a dedicated endpoint optimized for condensing long texts into concise summaries. It uses specialized model configurations and prompt engineering to handle the context window and extraction requirements of legal documents, unlike general-purpose generation endpoints.
Exam trap
Cisco often tests the distinction between a general-purpose generation API (Generate API) and a task-specific API (Summarisation), leading candidates to incorrectly choose the Generate API because they assume any text generation endpoint can handle summarization equally well.
How to eliminate wrong answers
Option A is wrong because the Chat API is designed for multi-turn conversational interactions, not for single-document summarization tasks, and lacks the specific prompt templates and length controls needed for legal document summarization. Option B is wrong because the Embedding API converts text into vector representations for semantic search or clustering, not for generating summaries. Option C is wrong because the Generate API is a general-purpose text generation endpoint that can produce summaries but is not specifically optimized or designed for summarization tasks, lacking the built-in summarization-specific parameters and model tuning that the Summarisation API provides.
A company is building a chatbot using OCI Generative AI service. They want to ensure that the model responses are grounded in their internal knowledge base. Which approach should they use?
RAG retrieves relevant documents from a knowledge base and uses them to generate grounded responses.
Why this answer
Retrieval-Augmented Generation (RAG) is the correct approach because it retrieves relevant documents from the company's internal knowledge base at inference time and provides them as context to the LLM, ensuring the model's responses are grounded in verifiable, up-to-date information without modifying the model itself. This directly addresses the requirement to ground responses in an internal knowledge base while avoiding the cost and complexity of retraining.
Exam trap
The trap here is that candidates often confuse fine-tuning (Option B) as the only way to incorporate proprietary data, overlooking that RAG provides a more flexible, cost-effective, and updatable method for grounding responses in a dynamic knowledge base without altering model weights.
How to eliminate wrong answers
Option A is wrong because prompt engineering with few-shot examples only provides a handful of static examples in the prompt, which cannot dynamically retrieve or incorporate the full breadth of an internal knowledge base, leading to hallucinations on unseen or specific internal data. Option B is wrong because fine-tuning the model on the internal knowledge base would embed that data into the model's weights, making it expensive to update, prone to catastrophic forgetting, and unable to guarantee factual grounding for new or changing documents without retraining. Option C is wrong because model distillation compresses a larger model into a smaller one for efficiency, but it does not introduce external knowledge retrieval; it merely replicates the behavior of the teacher model, which still lacks access to the internal knowledge base.
In a LangChain RAG pipeline using OCI Generative AI, which THREE components are essential for ingesting documents into a vector store?
Text splitter divides documents into manageable chunks for embedding and indexing.
Why this answer
Option B is correct because text splitters like RecursiveCharacterTextSplitter are essential for breaking large documents into smaller, manageable chunks that fit within the context window limits of embedding models and LLMs. Without chunking, the vector store cannot effectively index and retrieve relevant passages, making it a core component of the ingestion pipeline.
Exam trap
Cisco often tests the distinction between the ingestion pipeline (loader, splitter, embeddings) and the retrieval/generation pipeline (retriever, LLM), leading candidates to incorrectly include the retriever or LLM as essential for ingestion.
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?
Correct: Gradient accumulation reduces memory per step without changing effective batch size.
Why this answer
Option B is correct because enabling gradient accumulation allows the effective batch size to be maintained while reducing per-step memory usage. Option A changes the model entirely, Option C may not fix the memory issue, and Option D helps but may still OOM if the batch size is too high; gradient accumulation is more directly targeted.
A developer is using OCI GenAI to generate structured data. They often get responses that include additional commentary or markdown. Which prompt engineering technique should they use to ensure only JSON output?
Correct: Direct instruction enforces format.
Why this answer
Option C is correct because explicitly instructing the model to 'Return only JSON' directly constrains the output format, reducing the likelihood of extraneous commentary or markdown. This technique leverages prompt engineering to guide the model's behavior without altering inference parameters like temperature or top_p, which control randomness rather than output structure.
Exam trap
Oracle often tests the misconception that adjusting sampling parameters (like temperature or top_p) can enforce output format, when in fact these parameters control randomness and diversity, not structural constraints—leading candidates to overlook the direct prompt engineering solution.
How to eliminate wrong answers
Option A is wrong because setting top_p to 0.1 reduces the nucleus sampling threshold, making the model more deterministic but not preventing it from generating additional text or markdown; it controls token selection diversity, not output format. Option B is wrong because a larger context window allows the model to process more input tokens but does not enforce a specific output structure; it addresses memory limitations, not format constraints. Option D is wrong because increasing temperature to 1.5 raises randomness, which can actually increase the likelihood of unpredictable or verbose responses, including unwanted commentary, rather than ensuring strict JSON output.
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?
RAG retrieves relevant document chunks at query time, ensuring the chatbot always answers from the latest uploaded documents without any model retraining.
Why this answer
RAG (Retrieval-Augmented Generation) allows the LLM to retrieve relevant document sections at inference time, so knowledge stays current without retraining. The other options either require expensive retraining for each update or lack document grounding.
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?
Frequency penalty penalizes tokens that have already appeared, discouraging the model from repeating phrases.
Why this answer
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.
A team has deployed a generative AI model and needs to monitor inference performance and set up alerts for increased error rates. Which OCI service should they integrate with?
Correct: Monitoring provides metrics and alerting for inference endpoints.
Why this answer
OCI Monitoring is the correct service because it provides metrics and alarms for tracking inference performance (e.g., latency, throughput) and error rates from deployed generative AI models. It allows you to set up threshold-based alerts on custom or predefined metrics, enabling proactive incident response. This directly addresses the requirement to monitor inference performance and alert on increased error rates.
Exam trap
Oracle often tests the distinction between monitoring (metrics/alarms) and logging (raw events) — candidates mistakenly choose OCI Logging because they think 'error rates' require log analysis, but OCI Monitoring is designed for metric-based alerting with thresholds.
How to eliminate wrong answers
Option B is wrong because OCI Cloud Guard is a security posture management service that detects misconfigurations and security threats, not a real-time performance monitoring or alerting tool for inference metrics. Option C is wrong because OCI Events is a notification service that reacts to state changes in OCI resources (e.g., object creation, instance termination) but does not natively track or alert on time-series performance metrics like error rates. Option D is wrong because OCI Logging collects and stores log data for audit and troubleshooting, but it lacks built-in metric-based alerting capabilities for monitoring inference performance trends or setting threshold alarms.
Page 14 of 14
Practice 1Z0-1127 by domain
Target a specific domain to shore up weak areas.