Result of Hirshfeld-I charges is independent of initial guess. Although a good guess may lead to faster convergence, it is not easy to determine an evidently better guess than Hirshfeld charge.

Hello,

There is no so-called initial guess for calculating partial charges. Partial charges only depend on the wavefunction loaded into Multiwfn.

Assume we want to fit partial charges in a way to get the best fit to ESP (as usual e.g. in ChelpG and MK methods). There should be some initial guesses to be improved in each iteration of fitting or I am missing something?
Similarly, for Hirshfeld partitioning approaches that work iteratively, (e.g. Hirhfeld-I), if we start from a good guess of initial partial charges, I suppose we can reduce the number of iterations. Is that correct?

There is no so-called initial guess for calculating partial charges. Partial charges only depend on the wavefunction loaded into Multiwfn.