Secondary sulfate and organic aerosols from natural sources have been recognized as a major source of uncertainty in estimating aerosol effective radiative forcing (ERF) for over a decade, affecting climate predictions in the past, present, and future. Accurate simulations of these aerosols and their impacts on clouds and radiation depends on appropriate model representations of complex physical and chemical mechanisms driving their formation, growth, and activation. Here, we quantify how each component of aerosol ERF (namely, the cloud albedo effect, liquid water path adjustment, and cloud amount adjustment) is affected by (1) newly discovered chemical mechanisms for forming secondary sulfate and organic aerosols, (2) favored condensational growth for ultrafine particles using dynamic partitioning that accounts for particle phase diffusion limitation, and (3) enhanced activation of newly formed ultrafine aerosols using machine learning based surrogate of an accurate parcel model to account for kinetic limitation. The U.S. Department of Energy’s Energy Exascale Earth System Model is configured to run at kilometer scale using regionally refined meshes to reduce the uncertainty associated with convection parameterization. We show that the improved representations of natural secondary aerosols bring the present-day model simulations of aerosol and clouds into much better agreement with the observations in both pristine and polluted regions. The stubborn low droplet number concentration bias in climate models is largely alleviated. Total aerosol ERF is significantly affected by the revised representation of natural secondary aerosols, where the cloud albedo effect is affected most significantly. This study combines expertise in aerosol, cloud, climate modeling, and machine learning to address a critical model deficiency. We demonstrate that a much more realistic simulation of the natural secondary sulfate and organic aerosols, and consequently a better estimate of aerosol ERF, can be achieved by incorporating complex physics in a climate model at an affordable computational cost.