Never? Edit: To defend my response, the model definitely knows when it hasn't yet found a correct response, but this is categorically different from knowing th

Never? Edit: To defend my response, the model definitely knows when it hasn't yet found a correct response, but this is categorically different from knowing that it does not know (and of course monkeys and typewriters etc., can always find a proof eventually if one exists).