Why is chat gpt better than the api in my classification task?

Hey y’all,

So I am basically just doing a text classification task where i am essentially just trying to judge whether a peice of text is part of a payroll document or not. One very interesting thing is when i copy and paste it in chatGPT it does much better, but when i use the api (even with gpt-4-0125-preview) it always seems to perform worse.

Why is this ?

Could be any of number of things we don’t know about in an effort to make ChatGPT Plus more valuable or keep its “secret sauce” for a while longer.

If you share prompts/models/settings, we might be able to help you improve.

Lemme know. If you post, I’ll move this to Prompting, and we’ll see if we can improve results for you. Classification is super easy, so shouldn’t be too hard.

Hope you stick around. We’ve got a great dev community growing here.