We consider the problem of efficient inference of the Average Treatment
Effect in a sequential experiment where the policy governing the assignment of
subjects to treatment or control can change over time. We first provide a
central limit theorem for the Adaptive Augmented Inverse-Prob