The paper studies a mathematical object called an orthogonal projection matrix, which is a way of representing a geometric subspace (a flat, multi-dimensional slice through the origin) in terms of coordinates. Each such matrix has special numbers associated with it called principal minors, which are determinants of square sub-blocks sitting along the diagonal of the matrix. The central question is: what information about the subspace can you recover just from knowing all these principal minors?
The main result shows that the principal minors completely determine what the authors call the "coordinatewise absolute value" of the subspace. You can think of this as follows: take every vector lying in the subspace, apply the absolute value to each of its components separately, and collect all the resulting vectors. The set you get this way is the same for any two subspaces that share the same principal minors. In other words, even though the minors do not pin down the subspace exactly, they do pin down this particular geometric shadow of it.
The proof works by breaking the problem into cases based on a concept from combinatorics called matroid connectivity. A matroid is a structure that captures which sets of coordinates are "independent" in a combinatorial sense, and it can be read off from the projection matrix. The authors organize all subspaces into layers according to how connected their associated matroid is, prove a structural theorem describing what subspaces in each layer look like, and then handle each layer separately to establish the main result. This gives a cleaner and more organized understanding of the space of all subspaces, known as the Grassmannian, as well as a concrete link between an algebraic invariant and a geometric one.