Thanks for the response. I confirmed the behavior in the following test using the Test function in QnA Maker - it is easy to replicate (not just the Generate Answer API with the answers value).
I quickly created in QnA Maker with following Q-A Pairs
- a question with single turn answer
- a question with an answer with related questions
You can see the image attached the score went down after asking the same question right after a reply with related questions. Is there a way to keep the score through some parameters in API call or configuration?