The paper works in an area of mathematics that connects algebra and geometry by studying how complicated symmetry objects called "gerbes" can be. A gerbe is a geometric object that generalizes the idea of a group acting on a space, and it measures a kind of hidden symmetry or twisting. One important question is: how many parameters do you need to describe a gerbe, at least approximately? This count is called the "essential dimension." The paper extends a known result (due to Karpenko and Merkurjev) that previously only worked for simple, well-behaved symmetry groups, to a much broader class of gerbes built from prime-power symmetry groups. The key finding is that the essential dimension, measured relative to a prime p, equals a more concrete quantity called the "faithful rank," which counts the minimum size of a representation needed to faithfully encode the gerbe after you simplify the number theory by ignoring primes other than p.
The paper also proves a relative version of this result. Instead of looking at a single gerbe in isolation, it looks at maps between gerbes and asks how much complexity is added when you move along the map. The answer is that the relative complexity of the map equals the supremum, meaning the worst-case value, of the essential dimensions of the individual fibers, which are the gerbes you see sitting over each point in the base. This is a clean and natural generalization of the main theorem to a more flexible setting.
The third contribution introduces a new geometric measurement called the "quotient compression dimension." This quantity asks: what is the smallest dimension of a space with quotient singularities, meaning singularities that look locally like a space modded out by a finite group, that can realize the gerbe? The paper proves that this new dimension is always either equal to or exactly one more than the essential dimension at p. This is a striking result because it means the essential dimension, which is a purely algebraic invariant, almost completely determines the geometry of the simplest singular spaces that can encode the gerbe, with an ambiguity of at most one.