Answer: Increase top-p

## Explanation **Correct Answer: B (Increase top-p)** **Why this is correct:** 1. **Top-p (Nucleus Sampling)**: This parameter controls the cumulative probability threshold for token selection. When you increase top-p, you allow the model to consider a broader set of possible tokens that collectively reach the specified probability threshold. This results in more diverse vocabulary while still maintaining coherence and meaning. 2. **The problem context**: The company wants "a slightly broader vocabulary without losing meaning." This is exactly what increasing top-p achieves - it expands the pool of considered tokens while still filtering out low-probability, nonsensical options. **Why other options are incorrect:** - **A. Decrease temperature**: Temperature controls randomness - decreasing it makes output more deterministic and focused on high-probability tokens, which would make the text even MORE rigid, not less. - **C. Reduce top-k**: Top-k limits the number of tokens considered to the k most probable ones. Reducing top-k would make the output MORE constrained, not broader. - **D. Increase max-tokens**: This parameter controls the maximum length of generated text, not vocabulary diversity. Increasing it would simply allow longer responses, not affect vocabulary breadth. **Key concepts for AWS Certified Cloud Practitioner:** - **Amazon Bedrock** is AWS's fully managed service for foundation models - **Model parameters** like temperature, top-p, and top-k control text generation behavior - **Top-p (nucleus sampling)** is particularly useful for balancing creativity with coherence - Understanding these parameters helps optimize model outputs for specific use cases like financial reporting where accuracy and appropriate tone are crucial

Author: Ritesh Yadav