B. Use a minimum of 1,000 to 5,000 examples for each task, but focus on the quality and relevance of examples rather than quantity. C. Use at least 10,000 examples for each unique task to ensure the model retains its general knowledge and effectively adapts to the new task. D. Use no more than 100 examples per task to avoid overwhelming the model’s general capabilities with task -specific data. A company is using a Retrieval -Augmented Generation (RAG) system to enhance its question - answering model by incorporating relevant documents from a knowledge base. They need to generate vector embeddings for these documents and the queries using a pretrained transformer -based model. What is the most crucial aspect of generating vector embeddings for effective retrieval in this scenario?
D. In zero-shot prompting, the model is fine-tuned before answering, but in few-shot prompting, no fine-tuning occurs. In the context of quantizing large language models (LLMs), which of the following statements best describes the key trade -offs between model size, performance, and accuracy when using quantization techniques?
C. Testing the model’s accuracy on a large set of random data D. Implementing a feedback loop for continuous model improvement You are tasked with generating creative text outputs using an AI language model for a marketing campaign. You want to ensure that the responses are diverse and unexpected but still somewhat relevant to the prompt. Which combination of temperature and random seed should you use to achieve this?
You are optimizing a Generative AI model for a business application where cost savings are a priority. Which of the following modifications to the model’s parameters will most effectively reduce the overall generation cost while minimizing the loss of output quality?
You are optimizing a large language model (LLM) by prompt -tuning it for specific enterprise -level tasks. The goal is to initialize the prompt in such a way that it helps the model generalize well across various enterprise domains, such as finance, healthcare, and retail. What is the most effective method to initialize the prompt for such a use case?
➢ TOTAL QUESTIONS: 379 In the context of IBM Watsonx and generative AI models, you are tasked with designing a model that needs to classify customer support tickets into different categories. You decide to experiment with both zero-shot and few -shot prompting techniques. Which of the following best explains the key difference between zero -shot and few -shot prompting?
