Daniel Tan comments on Do models know when they are being evaluated?