Is self-interaction corrected density functional theory straying from the path toward the exact functional?

As recently pointed out [1] higher rung functionals (e.g. MGGA or hybrids) may deliver a better description of the total energy but not necessarily a better description of the density for the calculated systems. We present an additional perspective to this discussion and present the performance of self-interaction corrected densities for LDA, GGA and MGGA rungs of density functionals for an extended benchmark set considering atoms, molecules and stretched bond situations.

[1] M. G. Medvedev et al., Science, 355, 6320, 49-52, 2017