One of the most effective ways to test stellar evolutionary models is to measure dynamical masses for binary systems at a range of temperatures. In this paper, we present orbits of three young K+M binary systems in Taurus (Hubble 4, FF Tau, and HP Tau/G3) with VLBI parallaxes. We obtained precision astrometry with Keck-II/NIRC2, optical photometry with HST/WFC3, and low-resolution optical spectra with WIFeS on the ANU 2.3 m telescope. We fit orbital solutions and dynamical masses with uncertainties of 1-5% for the three binary systems. The spectrum, photometry, and mass for Hubble 4 are inconsistent with a binary system, suggesting that it may be a triple system where the primary component consists of two stars. For HP Tau/G3 and FF Tau, model masses derived from SED determined component temperatures and luminosities agree with the dynamical masses, with a small offset towards larger model masses. We find model ages for the primary components of these systems of ∼3 Myr, but find that the secondaries appear younger by a factor of two. These estimates also disagree with the age of the physically associated G-type star HP Tau/G2, which is older (∼5 Myr) according to the same models. This discrepancy is equivalent to a luminosity under-prediction of 0.1-0.2 dex, or a temperature over-prediction of 100-300 K, for K/M-type stars at a given model age. We interpret this as further evidence for a systematic error in pre-main sequence evolutionary tracks for convective stars. Our results reinforce that the ages of young populations determined from the locus of M-type members on the HR-diagram may require upward revision.