arXiv:2610.41146v1 Announce Type: cross Abstract: The electronic structure of molecules is foundational for scientific advancement, however determining it remains challenging due to the growth of the molecular Hilbert space with system size. Recent approaches address this by using quantum circuits to sample candidate configurations and sample-based quantum diagonalization (SQD) to compute the ground state. We propose ZenKO, a novel gradient-based, iterative SQD method that employs a hybrid quantum-classical approach. ZenKO decomposes configuration selection into discovery, validation, and refinement, enabling targeted subspace expansion and guided exploration. Our evaluation demonstrates that, across $H_2O$, $H_{10}$, and $N_2$ molecules, ZenKO improves ground-state accuracy compared to prior iterative methods, while significantly reducing classical cost.