The superweak Gotsman–Linial conjecture

About 5 years old · traced to

Let f:{−1,1}n→{−1,1}f:\{-1,1\}^n\to\{-1,1\} be an (n,d)(n,d)-PTF, meaning the sign of a real polynomial of degree at most dd on the Boolean hypercube. Let AS[f]\mathbf{AS}[f] denote its average sensitivity.

Superweak Gotsman–Linial conjecture. For some function gg depending only on dd,

AS[f]∈O(nlog⁡g(d)n).\mathbf{AS}[f]\in O\left(\sqrt n\log^{g(d)}n\right).

Daniel Kane resolved this conjecture. It is weaker than the O(dn)O(d\sqrt n) asymptotic bound and was introduced as a still useful consequence for applications.

References

Primary source

Brynmor Chapman, “The Gotsman-Linial Conjecture is False”, arXiv:2108.02288 (2021).

Progress summary

Refreshed
Claimed solved

Daniel Kane’s work gives the required weaker bound, while the stronger conjecture remains open.

The conjecture asks for a polylogarithmic improvement over square-root growth of average sensitivity for every fixed-degree polynomial threshold function. Gotsman and Linial proposed the stronger O(dn)O(d\sqrt n) bound; the superweak form follows from Kane’s later estimate.

Known results

  • Kane, 2009: AS[f]≤2O(d)log⁡n n1−1/(4d+2)\mathbf{AS}[f]\leq 2^{O(d)}\log n\,n^{1-1/(4d+2)}.
  • Kane, 2012: AS[f]≤n(log⁡n)O(dlog⁡d)2O(d2log⁡d)\mathbf{AS}[f]\leq \sqrt n(\log n)^{O(d\log d)}2^{O(d^{2}\log d)}, which has the required superweak form for fixed dd.
  • Chapman, 2017 and 2021: counterexamples to the stronger extremal Gotsman–Linial conjecture, not to the superweak bound.

2026 literature check

A 2026 paper reproduces Kane’s bound but says that some weaker versions of the Gotsman–Linial conjecture remain open; it does not identify the stated superweak formulation as unresolved.

Current status (as of September 2026): Kane’s cited bound settles the stated superweak form for each fixed dd, but this remains a claimed result in this report; the stronger O(dn)O(d\sqrt n) conjecture is open.

Sources

Solutions 0

No solutions have been posted yet.