https://platform.openai.com/playground/p/9UgzQTVFRKLMbFxCGFrW0X0E?model=gpt-3.5-turbo
At first, I set up Temperature 1 and Top P 1, and changed my prompt to see how the answer would be changed.
Prompt 1 : You are a friendly travel advisor.
The base of my prompt is travel advisor. For Prompt 1, I added “friendly” because I wanted something a little more exciting rather than simply being told about the destination when asking which city to recommend for Japan travel.
It gave me a concise explanation of Tokyo, but it sounded a lack of “friendly” part.
Prompt 2 : You are a friendly travel advisor specializing in Asia.
I added “specializing in Asia” in hopes that it would alleviate the lack of “friendly” part.
In terms of content, there are no major differences between prompts 1 and 2.
Prompt 3 : You are a friendly travel advisor specializing in Japan.
I changed from specializing in Asia to in Japan because I expected that a travel advisor specializing in Japan would speak about Japan more interestingly, compared to the one specializing in Asia.
It gave me the charms of Tokyo in more detail. Also, it sounded like “friendly advisor”.
Next, I changed Temperature and Top P under Prompt 3 (You are a friendly travel advisor specializing in Japan).
Temperature 2, Top P 1
Some sentences came out that I couldn’t understand, like garbled characters.
Temperature 0.5, Top P 1
The content is not much different from Temperature 1, but it seemed like it couldn’t fit into the number of characters, so it ended halfway.
Temperature 1, Top P 0.5
The content was a slightly improved version of the answer in Prompt 1. Compared to other answers in Prompt 3, the content sounded less friendly.
With all that in mind, my best is Prompt 3 (You are a friendly travel advisor specializing in Japan) with Temperature 1 and Top P 1.
