search

Lichang-Chen / InstructZero

Official Implementation of InstructZero; the first framework to optimize bad prompts of ChatGPT(API LLMs) and finally obtain good prompts!
https://lichang-chen.github.io/InstructZero/
178 stars 14 forks source link

more details about the use of APE #11

Open MozerWang opened 1 year ago

MozerWang commented 1 year ago

Nice work!

I would appreciate it if you could provide me with more about the details when you use APE as the baseline.

I think you use chatgpt to generate the prompts. But what is your score function when you use APE? What are your criteria for choosing prompts in the first part of APE. (I have seen the FQA, you don't use log because of API.)

And what prompts (system_content) you use to instruct Chatgpt?

thanks

MozerWang commented 1 year ago

Besides, I also want to post a question about your answer to [Why is the performance of APE quite poor on ChatGPT? ]

For the main result, I think the APE also uses ACC as the score function in their code.

So what is the more precise answer?