AI-102 Question 474
Select 3You are designing an Azure OpenAI-based chatbot to assist users by generating responses to customer queries. To ensure the chatbot provides concise and relevant answers while minimizing randomness in the responses, which parameters should you configure?
- A
Temperature
- B
Max Tokens
- C
Top-p (Nucleus Sampling)
- D
Frequency Penalty
- E
Stop Sequences
Show answer and explanation
Correct answers: A, B, C
Explanation
To control the generative behavior of an Azure OpenAI model, parameters such as Temperature, Max Tokens, and Top-p are critical. Temperature reduces randomness, Max Tokens limits response length, and Top-p narrows the range of token choices, ensuring the chatbot generates concise and relevant answers. While other parameters like Frequency Penalty and Stop Sequences are useful in specific scenarios, they do not directly impact the desired level of conciseness and relevance in this case.
- A. Correct.
Temperature controls the randomness of the generated responses. Lowering the value makes the chatbot more deterministic, ensuring concise and reliable answers.
- B. Correct.
Max Tokens limits the length of the chatbot's response, ensuring the answers remain concise and do not exceed a specific length.
- C. Correct.
Top-p (Nucleus Sampling) determines the range of tokens considered at each generation step, helping to balance randomness and relevance.
- D. Incorrect.
Frequency Penalty adjusts the likelihood of the model repeating the same phrases. While useful in some contexts, it does not directly control response conciseness or randomness.
- E. Incorrect.
Stop Sequences define where the generation should stop, but they do not directly affect the chatbot's randomness or overall conciseness.