For example: If you make a chat bot and let the user regenerate their responses; then there is an implicit preference ranking of the chatbot responses. Do any of you keep track of this?
On ChatGPT you can give feedback. Also there is a sort of upvote checkmark thing that I think helps them gauge good response patterns to feed back into their updated versions.