A cloud-resolving model (CRM) is used to simulate the multiple-layer mixed-phase stratiform (MPS) clouds that occurred during a three-and-a-half day subperiod of the Department of Energy-Atmospheric Radiation Measurement Program's Mixed-Phase Arctic Cloud Experiment (M-PACE) and to examine physical processes responsible for multilayer production and evolution. The CRM with a two-moment cloud microphysics is initialized with concurrent meteorological, aerosol, and ice nucleus measurements and is driven by time-varying large-scale advective tendencies of temperature and moisture and surface sensible and latent heat fluxes. The CRM reproduces the dominant occurrences of the single- and double-layer MPS clouds as revealed by the M-PACE observations although the simulated first cloud layer is lower and the second cloud layer is thicker compared to observations. The aircraft measurements suggest that the CRM qualitatively captures the major characteristics in the vertical distribution and interperiod variation of liquid water content (LWC), droplet number concentration, total ice water content (IWC), and ice crystal number concentration (nis). However, the magnitude of LWC is overestimated and those of IWC and nis are underestimated. In particular, the simulated nis is one order of magnitude smaller than the observed. Sensitivity experiments suggest that both the surface fluxes and large-scale advection control the formation of the lower cloud layer while the large-scale advection initiates the formation of the upper cloud layer but the maintenance of multilayer structures relies on the longwave (LW) radiative effect. The LW cooling near cloud top produces a more saturated environment and a stronger dynamical circulation while cloud base radiative warming of the upper layer creates the stability gap between the two cloud layers. Both cloud layers are sensitive to ice-forming nuclei number concentration since ice-phase microphysics provides a strong sink of cloud liquid water mass.