Answers are grounded in this paper and cite their sources.
We study the number of internal vertices needed to realize a complete unsigned matching signature on four labelled boundary vertices with d colours. The signatures are even: their entries for odd sets of boundary vertices are zero, as parity forces for any graph with a nonzero entry for the empty set. Edge weights are arbitrary complex numbers, boundary edges and one nonzero overall scalar are allowed, and equality is required for every set of boundary vertices and every assignment of colours to it. Every signature whose coefficient for the empty boundary set is nonzero has a realization with at most 4d^2 + 2 internal vertices, while a generic one requires at least sqrt(2) d^2 - 4d + O(1). If the internal graph is restricted to disjoint edges, generic and worst-case complexity have cubic order: the number of pairs lies between ceil(d^4/(8d - 4)) generically and d^3 universally. We also give an explicit family, with one signature for each even m >= 4 and d = m colours, whose unrestricted minimum is exactly m and whose disjoint-edge minimum has order m^2. The orders of growth follow from standard tools: fourth cumulants, parameter counts, tensor-rank bounds, flattening ranks and known matchgates. New are the explicit constant of the universal construction, the criterion that identifies the disjoint-edge cost with a decomposition rank, and the explicit family. The upper construction uses O(d^4) weighted edges; the result concerns vertices, not coefficient-storage size or the cost of counting matchings. Files. paper.pdf is the manuscript, revision 4 (17 pages); paper.tex is its LaTeX source and paper.md the same paper in Markdown. checking-programs.zip holds the five Python checking programs named in Section 8 of the paper, each with its saved output; their finite checks corroborate the proofs and do not establish the universal statements. Prior work. The closest earlier results are credited in Section 1 and in the references of the paper. Disclosure. The author directed the project and used OpenAI Codex to produce the results, proofs, checking programs and text, and Anthropic Claude, whose reviews supplied the question studied and several of the tools, to review the manuscripts and draft the changes in this revision; the paper's statement on AI use gives the details. The author has not verified the mathematics; no human has checked the proofs, and no human peer review or formal proof-assistant verification is claimed. No AI system is an author; the author publishes under a pseudonym. All checks named in the paper were made by AI systems and their agents; they are not peer review. Corrections are welcome at ignoramus110926@gmail.com.
No takes yet. Share an insight, caveat, or question.
Ignoramus (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: