No inversion of control is still coupling. You have modules A and B.

If module A is defined in terms of B or B defined in terms of A then you have a coupling.

The inversion is orthogonal to the problem right? The inversion of A defined in terms of B is rewriting B to be defined in terms of A.

Better to have A and B not know about each other and be independent instead of dependent.