Remeshed vortex methods (RVM) are semi-Lagrangian approaches that solve the Navier-Stokes equations in their vorticity-velocity formulation. The advection of the vorticity field is computed by discretizing the vorticity field on numerical particles following their motion. The particles are then remeshed on an underlying Cartesian grid, allowing the use of Eulerian methods in an originally Lagrangian approach. These methods have proven to be a stable and less dissipative alternative to more classical Eulerian methods. These features make them well adapted for Large Eddy Simulation (LES) of turbulent flows. As for subgrid-scale (SGS) modeling, the Variational Multiscale (VMS) variant of the Smagorinsky model and the Spectral Vanishing Viscosity (SVV) approach seems to be well-suited to the present context as they add diffusion to only the smallest scales of vorticity. However, these SGS models contain parameters that have to be a-priori set, and it is not granted that the values calibrated for classical Navier-Stokes work well for the RVM approach. To assess the sensitivity of both SGS models to their parameters (specifically, the model coefficient and explicit filter order), an uncertainty quantification analysis is conducted, using the Polynomial Chaos Expansion (PCE) as a surrogate model. The test case considered is the Taylor-Green Vortex (TGV) at Re = 5000. The study leads to the determination of optimal coefficients, which are subsequently applied to different Reynolds numbers and grid resolutions, as well as on a new test case, the decay of Homogeneous Isotropic Turbulence (HIT), demonstrating the robustness of parameter calibration.