Assume the setup of the paper's fl-RDT theorem, including the quantities ψrp, ψrd, p^, q^, and c^. Define the algorithmically achievable value ψrp,a by
ψrp,a=max{ψrpn→∞limP(ψrp can be achieved in polynomial time)=1}.
Let ψrd(r)=ψrd, with ψrd as in the stated fl-RDT theorem. Parametric fl-RDT algorithmic conjecture. There is an r-sequence ψrd(r) with a decreasing sequence c^> such that
ψrp=r→∞limψrd(r)(p^,q^,c^>).
If no subinterval of [0,1] contains no elements of p^ and q^, then
The conjecture proposes a general parametric fl-RDT mechanism relating computationally achievable values to limiting finite-parameter predictions; the source provides no resolution, so it remains open.
Sources & referencesView supporting material
Primary source
Mihailo Stojnic, “Parametric RDT approach to computational gap of symmetric binary perceptron”, arXiv:2601.10628 (2026).