Selection often favours different phenotypes in males and females, driving the evolu- tion of sex-specific genetic architectures that facilitate sexual dimorphism. Previous theory based on single-locus models has shown how such architecture can evolve through modifiers of gene expres- sion and allelic dominance in males and females. Here I consider the evolution of gene expression in polygenic traits experiencing sex-specific selection. In such traits, I find that sexual dimorphism evolves more readily through sex-specific gene amplification, whereby genes differ in their absolute expression levels between the sexes, than through the evolution of sex-specific dominance across loci, which requires strong sexual antagonism. Furthermore, I show that the type of genetic architecture that evolves through sex-specific amplification, namely the number and nature of sex-biased genes, is highly sensitive to the distance between optimal male and female trait values relative to the number of loci contributing to a trait, i.e, to the level of genetic redundancy for sexual dimor- phism. Together these results indicate that genetic architectures resulting from sex-specific selection are highly dependent on the genetic basis of the trait of interest, and that this creates challenges when interpreting current metrics of sexual antagonism.