1: \begin{abstract}

2: We introduce the concept of self-adaptive tensor network states (SATNS) based on multi-site correlators. The SATNS ansatz 

3: gradually extends its variational space incorporating the most important next-order correlators into the ansatz for 

4: the wave function. The selection of these correlators is guided by entanglement-entropy measures from quantum information 

5: theory. By sequentially introducing variational parameters and adjusting them to the system under study, the SATNS ansatz achieves 

6: to keep their number significantly smaller than the total number of full-configuration interaction parameters. The 

7: SATNS ansatz is studied for manganocene in its lowest-energy sextet and doublet states, the latter of which is known to be

8: difficult to describe. It is shown that the SATNS parametrization solves the convergence issues found for previous

9: correlator-based tensor network states.

10: \end{abstract}

11: