The classical (p,q)-theorem in combinatorics says something like this: if you have a large collection of convex sets where among any p of them, some q overlap, then there is a small finite set of points that collectively stabs all the sets. A "definable" version of this asks whether, in certain logical settings, the stabbing set can be chosen in a definable and structured way rather than just existing abstractly. This question had been open for so-called NIP formulas, which are formulas appearing in mathematical logic that capture a wide class of well-behaved mathematical structures, including ordered fields, valued fields, and many geometric settings.
The authors resolve this open conjecture by proving that the definable (p,q)-theorem holds for NIP formulas in any first-order logical theory, not just in specific familiar structures. The key technical concept here is "NIP," which roughly means the formula cannot encode arbitrarily complex combinatorial patterns, a property that makes the underlying structure tameable and well-behaved. The proof shows that this tameness is enough to guarantee the existence of definable, structured hitting sets whenever the combinatorial overlap condition is satisfied.
Beyond proving the main conjecture, the authors establish a sharper, more refined version of the result and use it to draw conclusions about "dense pairs," which are structures consisting of a mathematical universe together with a distinguished dense subset, like the real numbers paired with the rationals. Dense pairs are notoriously tricky to analyze because the extra subset introduces new complexity, so extending these combinatorial-geometric results to that setting is a meaningful application showing the reach of the new theorem.