In the high-accuracy numerical solution of the generalized Rosenau-KdV-RLW equation, the mismatch of the temporal and spatial accuracy orders often forces the calculation to adopt a fine grid to balance stability and accuracy. This not only significantly increases the computational cost but also intensifies the accumulation of rounding errors. To address this issue, this paper constructs a compact finite difference scheme with matched high-order spatiotemporal accuracy. For time discretization, the Crank-Nicolson scheme is adopted; Richardson extrapolation is then applied to elevate the temporal approximation accuracy to fourth order. Spatial discretization is carried out using the fourth-order compact finite difference method, which, while maintaining high spectral resolution, significantly reduces the width of the computational stencil and the storage requirements of the bandwidth matrix. We have established the stability and convergence theory of this scheme and proved the discrete conservation property; meanwhile, we have provided a strict analysis of the existence and uniqueness of the discrete solution. Numerical experiments demonstrate that the proposed scheme not only exhibits excellent validity and reliability but also maintains high-precision evolution on relatively coarse grids–thereby substantially improving numerical stability, conservation properties, and computational efficiency in long-term simulations. Mathematics Subject Classification. 65M12, 65M06, 35Q75.