The paper works in a branch of mathematics that blends algebraic geometry with a modern technique called "quadratic enrichment." Classical algebraic geometry counts geometric objects (curves, intersection points, etc.) using ordinary integers. Over the past decade, mathematicians have developed ways to replace those integer counts with richer algebraic objects that carry extra information about the geometry, particularly information sensitive to the field of numbers you are working over. These enriched counts live in something called the Grothendieck-Witt ring, which packages together both a usual integer count and additional symmetry data. The paper extends this enrichment program to a broader and more technically demanding setting: Deligne-Mumford stacks, which are geometric spaces that have built-in symmetries and arise naturally when you want to parametrize objects like curves with automorphisms.
The central construction is a "quadratically enriched pushforward," a way of moving geometric data from one space to another along a map, while keeping track of the extra quadratic structure throughout. The authors build this using a classical tool called Grothendieck-Serre duality, which relates geometric data on a space to data on its dual in a precise way. From this pushforward they derive enriched versions of two important classical constructions: Euler classes, which measure how a vector bundle twists over a space, and Euler numbers, which are the global counts you get by collapsing that data to a single algebraic element.
The paper then applies these tools toward an ambitious goal in enumerative geometry. Quantum K-invariants are numbers (or more precisely, elements of a ring) that count curves in a geometric space in a sophisticated way, incorporating ideas from mathematical physics and string theory. Defining them requires integrating over moduli spaces of stable maps, which parametrize all the curves at once. The authors propose a quadratically enriched version of these invariants, using their new enriched fundamental classes and pushforwards, valid whenever the relevant moduli space is smooth and satisfies certain tameness and orientation conditions. This opens the door to extracting finer arithmetic information from curve-counting problems than classical methods allow.