The subgraph count
is the number of distinct subgraphs of a finite simple graph
that are isomorphic to
a fixed graph
(Alon 1986). Copies need not be induced,
and different labelings of the same copy are counted only once. The definition extends
to hypergraphs by counting their subhypergraphs. For
an integer
and a fixed nonempty finite
simple
-uniform
hypergraph
without isolated vertices, let
be the maximum count over finite simple
-uniform hosts with exactly
hyperedges. The exclusion of
isolated vertices is necessary, since arbitrarily many isolated host vertices can
be added without changing
.
Kuang et al. (2026, Theorem 1.2) proved that there is a constant such that
|
(1)
|
where
is the fractional edge cover number
and
, with
and
fixed. This strengthens the earlier order-of-magnitude result
of Friedgut and Kahn (1998) to convergence of the normalized count. For graphs, it
proves the conjecture of Alon (1986). The count of labeled embeddings is
, where
is the automorphism
group of
. Its leading constant is at most
, and equals this value whenever
.
The leading constant has a variational description over finite hypergraphs and weights. For every , any host can be replaced by a complete
blow-up of a hypergraph with at most
vertices, without increasing its number of
hyperedges and while retaining at least a proportion
of the embeddings. Here a blow-up
replaces each vertex by a class of vertices and every template hyperedge
by all choices of one vertex from each of its classes. For graphs, complete finite
cores suffice in the variational formula; the weights attain a maximum for each fixed
core, but whether some finite core always attains the overall supremum remains open
(Kuang et al. 2026, Corollary 1.3 and Problem 5.2).
For the path graph on
vertices, Kuang et al.
(2026, Proposition 5.1) also obtained
|
(2)
|
|
(3)
|
for each fixed positive integer , where
and
is the falling factorial.
The maximum defining
is attained. Kuang et al. (2026) credit GPT-6 Astra with extending their graph
argument to globally rainbow embeddings and obtaining the hypergraph proof,
which the authors verified and rewrote.