Potential Logic Bug: Rating your own content
I wanted to ask if this is really how it is intended, that you partially rate your own prompts and answers? While I can understand that it can make to answer the own prompt, I doubt that rating the own content is intended and serves the training data quality.
I can't reproduce it each time, but I've experienced this multiple times while grinding.
Please correct my assumption if this is really the intention.
This is a bug, in order to fix it this query needs to be adjusted.
There is a different use case for this query which needs to be user-independent for ranking backlog activation .. it currently isn't so this is another bug.
I've had a slightly different setting where I was asked to rate an assistant prompt reply to my message (which was an answer to another message) so basically : Someone elese first question First assistant answer My answer (2nd question) second assistant answer
And I had to rate that second assistant answer. Not sure if that influences what I would rate but if possible should be avoided ? I've seen it a couple of other times too now (even multiple times on rating an assistant's response to my question) I can 👍 or 👎 my own question so that's probably an issue (same as mentioned by the creator of the post)
I noticed it too.Russian is currently small language.So I notice it all the time.I believe i did enough damage.I believe my prompts are not bad.But it is not for me to decide.