Abstract: In the Monte Carlo simulation of a deep-penetration problem, only a small number of particles can penetrate the shielding layer and reach the target region, resulting in a very low computational efficiency. In order to solve the deep-penetration problem, the hybrid Monte-Carlo-Deterministic method is studied based on the consistent adjoint driven importance sampling method (CADIS). The hybrid method can automatically generate the input parameters required for deterministic SN calculation from the constructive solid geometry in the Monte-Carlo modeling. The adjoint SN calculation is used to generate consistent source biasing and weight-window parameters for forward calculation of the Monte-Carlo method. On the other hand, the mesh-based weight window applied in the hybrid method will encounter memory bottleneck in large-scale problems. A new structure of nested mesh is developed for mesh coarsening to save the memory of the weight-window parameters. The coarse mesh does not affect the variance reduction effect of the importance sampling. Based on NECP-MCX code system, the hybrid method and mesh coarsening method are implemented. The numerical results of HBR-2 benchmark show that the figure of merit (FOM) of the hybrid method is up to two orders of magnitude higher than that of MCNP. The weight window mesh can be reduced by 226 times without affecting the accuracy and efficiency of the final results.