The cohomology ring of a Grassmannian is a classical object in mathematics that encodes geometric information about spaces of linear subspaces. Researchers had conjectured specific formulas, called Hilbert series formulas, that describe how these rings behave when you only allow generators up to a certain degree. These conjectures came from two groups: Reiner and Tudose, and a collaborative Polymath Jr. group studying both ordinary Grassmannians (spaces of k-dimensional subspaces of n-dimensional space) and Lagrangian Grassmannians (a symplectic geometry variant). The paper proves both sets of Hilbert series conjectures, settling these open problems completely.
To prove the formulas, the authors analyze the algebraic structure in careful detail. They identify specific maps between algebraic objects, called coefficient and cancellation maps, and work out exactly what their kernels look like. They also build explicit bases for the relevant algebraic structures and show that certain pieces of the filtration, called cancellation layers, can be identified with well-understood algebraic objects called complete intersection algebras. A key technical tool involves deforming the algebraic structures slightly, which lets them recover how the filtration builds up step by step.
However, the paper also disproves two related conjectures from the Polymath group about explicit bases for certain two-row cases. The conjectured families of basis elements turn out to be almost right but not quite: they fall short by exactly one element in a specific degree range. Concretely, for every parameter r at least 2, the proposed spanning set is one dimension too small in a particular weight. The authors show that swapping out a single index fixes this deficiency, and more broadly, that the proposed families are off by at most one element in any given weight. The overall picture is therefore one of near-complete success for the conjectures, with a precise and correctable error in the basis proposals.