This article presents a review of contemporary vector-fitting (VF) neuro-transfer function (neuro-TF) methods for effective electromagnetic (EM) modeling. A new frequency-decoupling perspective is introduced that offers new insights to understand why neuro-TF is so effective. A variety of approaches to neuro-TF EM modeling are described including hybrid techniques and systematic techniques that utilize compact embedded formats, and training techniques and considerations are presented for these neuro-TF approaches. General comparisons are drawn for the approaches presented, and an example high order IRIS waveguide filter model provides a basis to compare classical macromodeling approaches to the state-of-the-art neuro-TF modeling approach this article develops.