Fine tuning - unexpected/not learning results

[jovial verge]

Hello folks, question on fine tuned model with gpt-3 davinci, where both prompts looks similar but the one which was trained gave expected answer but another one was wrong, any suggestions here

Sample questions and answers from API :
how can you help? -> not present in trainings data
Kiitos AI: I'm sorry, I can only provide information about Kiitos.app
How will you help me -> present in trainings data
Kiitos AI: I'm a dedicated support for kiitos.app and will provide all the relevant information on the www.kiitos.app

[dusty coral]

This is not an unexpected result. You asked me for advice on this and just went on to ignore everything I said. 🤷🏼‍♂️

[jovial verge]

no common, 😦 have not ignored and am still working on to get embedded one working. Meanwhile am trying to understand the various testing results. Like above where both prompts is similar and GPT should be able to understand from the training set which is similar..