#Which prompt style is "better" for GPT-3/GPT-4 LLM?

1 messages · Page 1 of 1 (latest)

coral salmon
#

I want to provide the user's (structured) personal information to the LLM using the system message.

{"role": "system", "content": "The user's information..."}

I am able to provide these in these two different format:

User's information:
name: John
occupation: software engineer, barista
age: 30
pronoun: he
nationality: Czech Republic
hobbies: programming, swimming
The user's name is John. He is a software engineer and a barista. He is 30 years old. He is from the Czech Republic. His hobbies are programming and swimming.

From a quick test, the token size is roughly the same and the accuracy are both pretty accurate.

Which style is more "better" and why?