IBM watsonx Generative AI Engineer - Associate C1000-185 Prüfungsfragen mit Lösungen:

1. You are tasked with optimizing a prompt-tuned large language model (LLM) using IBM Watsonx for a customer service chatbot. The chatbot needs to handle a variety of tasks, such as answering frequently asked questions (FAQs), providing detailed product descriptions, and troubleshooting user issues.
What is the most appropriate task to focus on during the initial tuning experiment?

A) Tune the model for text summarization, condensing user queries into shorter forms.
B) Focus on prompt-tuning the model for multi-turn dialogue to simulate more natural conversations.
C) Fine-tune the model to generate product descriptions using longer contextual prompts.
D) Optimize the model for extractive question-answering from a predefined knowledge base.

2. You want a generative AI model to summarize a lengthy text in one sentence. You provide the following prompt: "Summarize the following paragraph in one sentence: 'Artificial intelligence (AI) is the simulation of human intelligence in machines that are programmed to think and learn. The field of AI includes everything from speech recognition to problem-solving and robotics.'" No prior examples are given.
What type of prompting is being used, and what are the expectations?

A) Zero-shot prompting, with the expectation that the model can summarize common concepts like AI due to pre-training.
B) Zero-shot prompting, but it requires the addition of a few examples for the model to generate a summary.
C) Zero-shot prompting, but the model may fail without detailed instructions on what aspects of the text to focus on.
D) Few-shot prompting, but the model will likely struggle without a few examples of summaries provided.

3. You are working on a Retrieval-Augmented Generation (RAG) system where large-scale document retrieval is a critical component. To improve the efficiency and accuracy of retrieval, you need to store and query vector embeddings. Given that the system needs to handle billions of high-dimensional embeddings while maintaining low latency for search queries, you are evaluating the use of a vector database.
Which of the following databases would be the most appropriate choice for this purpose, and why?

A) A document-based NoSQL database like MongoDB, utilizing full-text search capabilities.
B) A vector database like Pinecone or Weaviate that supports approximate nearest neighbor (ANN) search.
C) A graph database like Neo4j, which is designed for traversing relationships between data points.
D) Relational databases with B-tree indexes.

4. You are tuning a generative AI model to reduce repetitive outputs, which often occur when generating long texts.
Which of the following parameter adjustments would most likely reduce the model's tendency to repeat words or phrases without compromising the quality of the generated text?

A) Lowering the temperature to 0.1
B) Reducing the beam size to 1
C) Increasing the repetition penalty to 1.2
D) Increasing the Top-k value to 200

5. You are working with IBM watsonx's generative AI model and wish to reduce the likelihood of generating rare, low-probability tokens while still retaining some level of creativity. You decide to use top-k sampling for this purpose.
Which of the following settings for the top-k parameter would be most effective in achieving a balance between creativity and maintaining coherent outputs?

A) Set top-k to 5
B) Set top-k to 1
C) Set top-k to 50
D) Set top-k to 1000

Fragen und Antworten: