At the submicron scale, the governing mechanism of plastic deformation changes from mutual interactions of dislocations to the operation of sources such as truncated sources and surface nucleation. For this reason, creating a realistic nucleation scheme in computational modeling is imperative to correctly craft mechanical theory on metal plasticity at this size scale. In our previous dislocation dynamics (DD) simulations, we have adopted the nucleation rate from atomistic models to accommodate dislocation sources associated with surface nucleation within the DD framework. Following an algorithm commonly used in kinetic Monte Carlo (kMC) method, our model can capture the statistical character of surface nucleation. In this study, we propose a surface nucleation model, which mimics the result of atomistic modeling where twist boundary forms from the dislocation network. With a the proposed model, we investigate the effect of the stability of the resulting dislocation network on the mechanical behavior of single crystalline metal micropillars. We follow both the evolution of the dislocation structure and the corresponding stress-strain relation. Our simulation results show the evident size effect and clear Bauschinger effect, which appear to be good agreement with latest experimental results.