The LoCA Regret: A Consistent Metric to Evaluate
  Model-Based Behavior in Reinforcement Learning

       Harm van Seijen1 , Hadi Nekoei2 , Evan Racah2 , Sarath Chandar2,3,4
            1
              Microsoft Research Montréal, 2 Mila - Quebec AI Institute,
           3
             cole Polytechnique de Montréal, 4 Canada CIFAR AI Chair

                        Abstract
     Deep model-based Reinforcement Learning (RL) has the potential to substantially
     improve the sample-efficiency of ...