nerdexam
Microsoft

DP-100 · Question #565

Note: This question is part of a series of questions that present the same scenario. Each question in the series contains a unique solution that might meet the stated goals. Some question sets might…

The correct answer is B. No. Increasing Top-p (nucleus sampling probability) does NOT directly produce more creative responses in the way the question intends. Top-p controls the cumulative probability mass of tokens considered during sampling - a higher value broadens the candidate token pool slightly…

Optimize language models for AI applications

Question

Note: This question is part of a series of questions that present the same scenario. Each question in the series contains a unique solution that might meet the stated goals. Some question sets might have more than one correct solution, while others might not have a correct solution. After you answer a question in this section, you will NOT be able to return to it. As a result, these questions will not appear on the review screen. You manage an Azure AI Foundry project in your subscription. You deploy a gpt-4o model. You must test the model before you use it in an existing front-end application. You need to adjust the parameters to get more creative responses. Solution: Increase Top probabilities. Does the solution meet the goal?

Options

  • AYes
  • BNo

How the community answered

(33 responses)
  • A
    27% (9)
  • B
    73% (24)

Explanation

Increasing Top-p (nucleus sampling probability) does NOT directly produce more creative responses in the way the question intends. Top-p controls the cumulative probability mass of tokens considered during sampling - a higher value broadens the candidate token pool slightly, but it is Temperature that is the primary parameter for controlling creativity and randomness. Increasing Temperature causes the model to produce more varied, imaginative, and less predictable outputs. To meet the goal of 'more creative responses,' the correct action is to increase Temperature, not Top-p. Therefore, this solution does not meet the goal.

Topics

#LLM parameters#Generative AI#Azure AI Foundry#Model testing

Community Discussion

No community discussion yet for this question.

Full DP-100 Practice