최신 IBM Certified watsonx Generative AI Engineer - Associate C1000-185 무료샘플문제:

1. When creating a prompt-tuned model, which of the following factors is most critical to ensure that the model generates accurate and contextually relevant responses?

A) Matching the training data prompts as closely as possible to real-world deployment scenarios.
B) Employing low learning rates to prevent the model from adapting too quickly to new prompts.
C) Using a diverse set of unrelated tasks in the tuning dataset to increase model generalization.
D) Ensuring that the prompt length does not exceed the model's input token limit.

2. You are deploying a new version of a generative AI model in IBM Watsonx, and you want to maintain the integrity of prompt versioning throughout the deployment lifecycle.
Which of the following methods is the most effective for ensuring that the correct prompt version is used with the corresponding model version in production?

A) Hard-code the prompt within the model deployment script to ensure that the correct prompt is always used
B) Use semantic versioning for both the model and the associated prompts, and track them independently in separate systems.
C) Use the latest available prompt version for every deployment, without specifying an exact version
D) Implement model registry tags that associate a specific prompt version with each model version during deployment.

3. In a RAG system, you need to select an appropriate retriever to fetch relevant documents from a large corpus before generating an answer. You are considering different types of retrievers, including embedding-based and keyword-based retrievers.
Which of the following describes a scenario where an embedding-based retriever using a vector database is the best choice?

A) When most of the queries consist of structured queries with precise Boolean operators and relational database-style searches
B) When exact keyword matching is required, and synonyms or contextual understanding are irrelevant
C) When documents are labeled with metadata, and only metadata needs to be searched
D) When retrieval must rely on semantic similarity between a query and documents, even if the exact words in the query don't appear in the document

4. A team is fine-tuning a large language model (LLM) for a healthcare application. They have decided to implement a taxonomy tree-based curation to prepare their dataset of medical records and patient interactions.
What is the primary benefit of using a taxonomy tree in the curation process for such a model?

A) It reduces the overall size of the dataset by filtering out irrelevant data.
B) It ensures that the model only focuses on specific medical specialties by eliminating other categories.
C) It eliminates the need for data cleaning and preprocessing since the taxonomy provides the structure.
D) It provides a hierarchical structure that helps the model understand the relationships between different medical concepts.

5. After prompt-tuning a language model, you notice that certain outputs are semantically correct but syntactically flawed.
Which of the following actions is most appropriate to resolve this issue and optimize the tuned model's performance?

A) Lower the learning rate during the tuning phase
B) Fine-tune the prompt template to emphasize grammar
C) Increase the model's training dataset size
D) Use a higher temperature during the generation process

질문과 대답: