A New Challenge in Policy Evaluation

This paper proposes a new challenge in policy evaluation: to improve the online data efficiency of Monte Carlo methods via information extracted from offline data while maintaining the unbiasedness of Monte Carlo methods.