AWS Certified Cloud Practitioner

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