You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As discussed, 4 of the 20 'withVector' batch questions failed to find the expert eval - and every one of the questions is an identical question from the trial. I've added the source chatId into a new column on the attached spreadsheet.
Will take some sleuthing on your part to figure out why they didn't match. Maybe not a bug? But if not a bug, it's confusing.
If it's because the evals are 'too old' we'll really have to work on that - sounds like @anniecrombie will do some thinking on that. As we said, some policies tend to change often/quickly like EI, some rarely or never have.