The paper on arXiv proposes an extension of multifidelity Monte Carlo (MFMC) for high-dimensional quantities of interest. MFMC is a technique for estimating statistics of quantities of interest computed via expensive high-fidelity models by using cheaper low-fidelity models as control variates.
The abstract states the extension but is cut off before details. It does not report specific numerical results, error bounds, or comparisons. Therefore the available text only establishes the problem setting and intended contribution.
Since there is only one source, there are no differing accounts to reconcile. The significance is conditional: if the extension holds, it could make MFMC applicable to problems where quantities of interest are high-dimensional, but the abstract alone does not demonstrate performance.