1: \begin{abstract}

2: Multifidelity methods are widely used for estimating quantities of interest (QoIs) in computational science by employing numerical simulations of differing costs and accuracies.

3: Many methods approximate numerical-valued statistics that represent only limited information about the QoIs. 

4: In this paper, we generalize the ideas in \cite{xu2021bandit} to develop a multifidelity method that approximates the full distribution of scalar-valued QoIs. The main advantage of our approach compared to alternative methods is that we require no particular relationships among the high and lower-fidelity models (e.g. model hierarchy), and we do not assume any knowledge of model statistics including correlations and other cross-model statistics before the procedure starts. 

5: Under suitable assumptions in the framework above, we achieve provable 1-Wasserstein metric convergence of an algorithmically constructed distributional emulator via an exploration-exploitation strategy.

6: We also prove that crucial policy actions taken by our algorithm are budget-asymptotically optimal.

7: Numerical experiments are provided to support our theoretical analysis.

8: \end{abstract}

9: