Vu’s conjecture on spectral determination for symmetric ±1 matrices

Establish that, for almost all n × n symmetric matrices with entries in {−1, +1}, the multiset of eigenvalues uniquely determines the matrix among symmetric ±1 matrices; equivalently, show that the probability a random symmetric ±1 matrix is determined by its spectrum tends to 1 as n → ∞.

Background

Following Haemers’ conjecture for graphs, Vu proposed an analogous statement in the setting of random symmetric ±1 matrices, seeking a high-probability spectral characterization in a non-graph, matrix ensemble context.

The present paper’s findings for digraphs suggest the importance of the real symmetric condition in such spectral determination phenomena, underscoring why this conjecture remains significant.

References

Vu conjectured a parallel version for symmetric $\pm 1$ matrices.

One can almost never hear the shape of a digraph  (2604.02165 - Zhao, 2 Apr 2026) in Section 1 (Introduction)