Hello everyone, I created a bot using Chat completions. I gave close to 50 user assistance examples and all of them in English. I asked our company’s CTO. The question I asked was in Turkish, but the first answer he gave was one letter, and when I asked the second, he gave them all.
This doesn’t always happen. It happens every now and then.
A solution suggestion from the gut if you use the base API: You could add the following upstream string to the prompts via the backend:
“Translate the following in English, phrase your reply in Turkish: ()”
Why does this help? The translation forces the model to process the entire text. In the next step, the answer is generated in Turkish. You have a slight performance degradation due to the double translation, but it should improve the stability of the responses.
Oh this is quite interesting, because to my best knowledge and what I have seen here on other forums finetuneing for gpt 3.5 turbo is not (yet) available. Interesting that you got it to work - great work!
So to get back to your question: I think in your case it is because we are at the intersection of two not yet perfected functions (Turkish language processing and gpt-3.5-turbo-training) which causes these dropouts. I think the suggestion I just made to always translate the messages to English and then back to the user’s original language will make it easier for your model to match the training data and thus increase reliability.