We model the evolution of the mean galaxy occupation of dark-matter halos over the range $0.1<z<1.3$, using the data from the VIMOS-VLT Deep Survey (VVDS). The galaxy projected correlation function $w_p(r_p)$ was computed for a set of luminosity-limited subsamples and fits to its shape were obtained using two variants of Halo Occupation Distribution models. These provide us with a set of best-fitting parameters, from which we obtain the average mass of a halo and average number of galaxies per halo. We find that after accounting for the evolution in luminosity and assuming that we are largely following the same population, the underlying dark matter halo shows a growth in mass with decreasing redshift as expected in a hierarchical structure formation scenario. Using two different HOD models, we see that the halo mass grows by 90% over the redshift interval z=[0.5,1.0]. This is the first time the evolution in halo mass at high redshifts has been obtained from a single data survey and it follows the simple form seen in N-body simulations with $M(z) = M_0 e^{-beta z}$, and $beta = 1.3 pm 0.30$. This provides evidence for a rapid accretion phase of massive halos having a present-day mass $M_0 sim 10^{13.5} h^{-1} M_odot$, with a $m > 0.1 M_0$ merger event occuring between redshifts of 0.5 and 1.0. Futhermore, we find that more luminous galaxies are found to occupy more massive halos irrespectively of the redshift. Finally, the average number of galaxies per halo shows little increase from redshift z$sim$ 1.0 to z$sim$ 0.5, with a sharp increase by a factor $sim$3 from z$sim$ 0.5 to z$sim$ 0.1, likely due to the dynamical friction of subhalos within their host halos.