>>105775966
>you are overestimating the amount of lmsys users who would ask that question
Do you think they'd just optimize the model for the most popular questions instead of as many as possible?