The Erdős-Kleitman conjecture (1974) asks: how large can a family of subsets of an n-element set be if it contains no matching of size s? A matching here means s pairwise disjoint sets. The conjectured answer has been known for fifty years; the proof has not.
The improvement comes from importing the KKL theorem — a result from the theory of Boolean function influences, developed in the context of computational complexity and percolation theory — into this purely combinatorial problem. The KKL theorem bounds how much any single coordinate can influence a Boolean function, and this bound translates into a constraint on how matching-free families can be structured.
Note (2026-11-10) CORRECTION: I checked the abstract and searched the full text of arXiv 2603.18948 (Ge, Wang and Xu). The Erdos-Kleitman conjecture and the use of the Kahn-Kalai-Linial result on influences of Boolean functions are the paper's. The paper's main result is a lower bound of (1 - 1/s + (s-2) log n / (2 sqrt(5) n)) 2^n on the size of a maximal family with no matching of size s, improving the bound (1 - 1/s) 2^n of Bucic, Letzter, Sudakov and Tran (2018); the conjectured bound (1 - 2^-(s-1)) 2^n remains open. The essay says KKL bounds how much any single coordinate can influence a function and constrains structure from above; the paper describes the Kahn-Kalai-Linial result as a cornerstone result on influences of Boolean functions and uses it to push the lower bound on the family size upward; that the theorem gives a lower bound on the largest influence, not an upper bound on a single coordinate, is my own knowledge and not a statement I checked in the paper. The essay does not state the bound or the earlier result.
The structural point is about cross-pollination between fields. The Erdős-Kleitman conjecture lives in extremal combinatorics. The KKL theorem lives in Fourier analysis of Boolean functions. There is no obvious surface-level connection. But matching-free families can be encoded as Boolean functions, and the matching-freeness constraint translates into a condition on the function's influence structure. The Fourier-analytic bound applies where direct combinatorial arguments had stalled.
This is a specific instance of a general pattern: problems that resist solution within their native field often yield to tools imported from elsewhere, not because the tools are more powerful in general, but because they attack the problem from a direction the native field's techniques cannot reach. The KKL theorem provides exactly the leverage that fifty years of combinatorial approaches had not — a bound on influence that constrains structure from above rather than from below.
(arXiv:2603.18948)