Drop a random walker on a 2D lattice. Let it walk until it returns to its starting point, forming a loop. Collect many such loops. The random walk loop soup is the ensemble of all these loops, weighted by their probability and a parameter controlling the intensity (how many loops per unit area).
Lawler & Peltola (arXiv:2603.13161) prove universality for the 2D random walk loop soup: in the scaling limit (lattice spacing → 0), the collection of macroscopic loops converges to the same continuum object regardless of the underlying lattice. Square lattice, triangular lattice, hexagonal lattice — the macroscopic loops are the same. The limit is the Brownian loop soup, a fundamental object in conformal field theory.
Universality for individual random walks has been known since Donsker's theorem (1951): a single walk converges to Brownian motion. But the loop soup is a collection of infinitely many overlapping loops, and convergence of the collection is much harder than convergence of individual trajectories. The loops interact through their mutual avoidance constraints and their collective covering of the plane.
The proof requires controlling the loop soup at all scales simultaneously. Short loops are numerous and local; long loops are rare but stretch across the entire domain. The universality applies to both: the short-loop statistics determine the local intensity, while the long-loop statistics determine the conformal structure. Both must converge, and they must converge jointly.
For conformal field theory, this matters because the Brownian loop soup is conjectured to describe the scaling limits of critical statistical mechanics models — the Ising model, percolation, the Potts model at specific temperatures. Proving universality for the loop soup is a step toward proving universality for these physical systems: the critical behavior doesn't depend on the lattice, only on the dimension and symmetry.