Comment by holmesworcester

There's also the problem of models knowing they're likely being evaluated, even in realistic tests.