pith. sign in

arxiv: 2607.01967 · v1 · pith:VD3K4FGGnew · submitted 2026-07-02 · ⚛️ physics.soc-ph · cond-mat.dis-nn· math.CO

The ring wants to be broken

Alexei Vazquez This is my paper

Pith reviewed 2026-07-03 03:18 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cond-mat.dis-nnmath.CO
keywords Ramsey community numbercirculant ring latticestochastic block modelBayesian evidencecommunity emergencelocal connectivity
0
0 comments X

The pith

A plain cycle lattice is never better described by communities, but adding next-nearest links creates a finite size where partitions become preferred.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper computes the size at which a homogeneous ring lattice prefers a two-community partition over the unpartitioned description, using exact Bayesian evidence under a fixed detection model. For nearest-neighbor rings the evidence for any partition decays as a power law in network size, so the unpartitioned description remains favored at all scales. For next-nearest-neighbor rings the evidence instead grows linearly once the network exceeds roughly 35 nodes. This isolates the minimal local wiring change that allows community structure to appear without any node heterogeneity.

Core claim

In the circulant ring C_n(1,...,c) the closed-form evidence comparison shows that c=1 yields a two-community posterior decaying as n^{-(2α+3)} so r_κ=∞, while c=2 produces a finite r_κ≃35 after which the log-evidence for the balanced partition grows as (ln 2)n.

What carries the argument

Exact Bayesian evidence ratio between the unpartitioned model and the balanced two-community partition, obtained in closed form for the Bernoulli stochastic block model with symmetric Beta priors applied to the circulant ring.

If this is right

  • The plain cycle (c=1) never acquires a preferred partition at any finite size.
  • The next-nearest ring (c=2) switches to preferring the partition above approximately 35 nodes.
  • Above the transition the log-evidence for the partition grows linearly with network size.
  • Community preference can arise from local connectivity rules alone in a fully homogeneous graph.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same exact-evidence method could be applied to rings with larger interaction range c to map how r_κ decreases with c.
  • The linear growth for c=2 supplies a concrete baseline against which to test whether other regular graphs without labels exhibit similar transitions.
  • Because the transition point is model-dependent, repeating the calculation with different priors would show how sensitive community emergence is to the detection rule.

Load-bearing premise

The community detection rule must be the Bernoulli stochastic block model with symmetric Beta priors; the location of the transition depends on this specific choice.

What would settle it

Direct numerical evaluation of the evidence ratio for a concrete ring of size n=40 with c=2, checking whether the partition evidence exceeds the unpartitioned evidence by the predicted (ln 2)n amount.

Figures

Figures reproduced from arXiv: 2607.01967 by Alexei Vazquez.

Figure 1
Figure 1. Figure 1: FIG. 1. (a) Posterior weight of the balanced two-community partition, Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
read the original abstract

The Ramsey community number $r_\kappa$ is the minimum network size at which a graph's connectivity is better described by a partition into communities than by no partition, under a prescribed community-detection rule. It was introduced through numerical simulations of networks grown by local rules, which suggested that community structure can emerge without any node heterogeneity. Here I compute $r_\kappa$ analytically for the simplest homogeneous, locally wired graph: the circulant ring lattice $C_n(1,\dots,c)$. Using a Bernoulli stochastic block model with symmetric $\mathrm{Beta}$ priors as the detection rule, the Bayesian evidence for a balanced two-community partition and for the unpartitioned network are both obtained in closed form, so the transition between them can be located exactly. The result is a sharp dependence on the interaction range: the plain cycle ($c=1$) is never partitioned, its two-community posterior decaying as $n^{-(2\alpha+3)}$, so $r_\kappa=\infty$; but the next-nearest-neighbour ring ($c=2$) acquires a finite $r_\kappa\simeq 35$ nodes, above which the partition is preferred with a log-evidence growing as $(\ln 2)\,n$. This provides an exactly solvable instance of community emergence in a network with no built-in communities, and shows that a minimal amount of local connectivity is enough to break the ring.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript defines the Ramsey community number r_κ as the smallest n at which a balanced bipartition is preferred over the single-community null under a fixed Bernoulli SBM detection rule with symmetric Beta(α,α) priors. For the circulant ring C_n(1,…,c) it derives closed-form marginal likelihoods, showing that the two-community posterior decays as n^{-(2α+3)} (hence r_κ=∞) when c=1, while for c=2 the partitioned model overtakes at a finite n≃35 with log-evidence growing linearly as (ln 2)n.

Significance. If the derivations hold, the work supplies an exactly solvable, parameter-controlled example of community emergence driven solely by local wiring in an otherwise homogeneous graph. The closed-form Beta integrals and the explicit n-dependence of the evidence ratio constitute a clear technical contribution that can be used to benchmark more complex community-detection settings.

major comments (1)
  1. [abstract and results for c=2] The numerical threshold r_κ≃35 is stated without an explicit value (or functional dependence) for the prior parameter α; because the crossing point is the root of an α-dependent ratio of Beta integrals, the concrete number cannot be reproduced or generalized from the given information.
minor comments (1)
  1. [derivation of marginal likelihoods] The abstract asserts closed-form expressions but the main text should include the explicit integral evaluations or intermediate edge-count expressions that yield the n^{-(2α+3)} and (ln 2)n scalings.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment and for identifying the missing detail on the prior parameter. We address the single major comment below and will make the corresponding minor revision.

read point-by-point responses

  1. Referee: [abstract and results for c=2] The numerical threshold r_κ≃35 is stated without an explicit value (or functional dependence) for the prior parameter α; because the crossing point is the root of an α-dependent ratio of Beta integrals, the concrete number cannot be reproduced or generalized from the given information.

    Authors: The referee is correct. The manuscript uses the uniform prior α=1 for the numerical illustration of the c=2 case (as stated in the methods section), which yields the reported crossing near n=35; the general α-dependence is implicit in the closed-form Beta integrals but was not restated in the abstract or results. We will add the explicit value α=1 to the abstract and results, and include a brief remark on how the threshold scales with α. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is direct computation from prescribed model

full rationale

The paper defines r_κ explicitly as the size at which the chosen detection rule (Bernoulli SBM + symmetric Beta priors) prefers a balanced bipartition over the null on the exact circulant adjacency matrix of C_n(1..c). It then derives the marginal likelihoods in closed form by evaluating the Beta integrals on the intra- and inter-block edge counts of that specific graph, yielding the exact n^{-(2α+3)} decay for c=1 and the (ln 2)n growth for c=2. These expressions follow immediately from the model definition and the ring's deterministic edge structure; α remains an explicit prior parameter whose effect is reported rather than fitted or hidden. No self-citation, ansatz smuggling, or renaming of external results occurs. The central claim is therefore the direct, parameter-free consequence of applying the stated rule to the stated graph.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the specific Bayesian detection rule and the regularity of the circulant graph that permits closed-form marginals; no new entities are postulated.

free parameters (1)
  • alpha
    Shape parameter of the symmetric Beta prior on block probabilities; controls the power-law exponent for c=1 and the numerical value of r_κ for c=2.
axioms (1)
  • domain assumption The community detection rule is defined to be the comparison of Bayesian evidence under a Bernoulli SBM with symmetric Beta priors.
    This rule is prescribed by the definition of r_κ and determines when the partition is preferred.

pith-pipeline@v0.9.1-grok · 5774 in / 1418 out tokens · 52948 ms · 2026-07-03T03:18:10.015612+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

  1. [1]

    certainty

    the value is more prior depen- dent, ranging fromr κ = 14 (α= 3) to 30 (α= 0.3): a weaker prior demands a larger network before commu- nities emerge, while a stronger prior lowers the bar. As the certainty requirement is tightened toward the almost- certain regime of [12], this prior dependence washes out andr κ concentrates near∼35 nodes. Two features co...

    work page

  2. [2]

    Fortunato, Phys

    S. Fortunato, Phys. Rep.486, 75 (2010)

    work page 2010

  3. [3]

    Fortunato and D

    S. Fortunato and D. Hric, Phys. Rep.659, 1 (2016)

    work page 2016

  4. [4]

    Girvan and M

    M. Girvan and M. E. J. Newman, Proc. Natl. Acad. Sci. U.S.A.99, 7821 (2002)

    work page 2002

  5. [5]

    M. E. J. Newman and M. Girvan, Phys. Rev. E69, 026113 (2004)

    work page 2004

  6. [6]

    M. E. J. Newman, Proc. Natl. Acad. Sci. U.S.A.103, 8577 (2006)

    work page 2006

  7. [7]

    Rosvall and C

    M. Rosvall and C. T. Bergstrom, Proc. Natl. Acad. Sci. U.S.A.105, 1118 (2008)

    work page 2008

  8. [8]

    P. W. Holland, K. B. Laskey, and S. Leinhardt, Soc. Net- works5, 109 (1983)

    work page 1983

  9. [9]

    Karrer and M

    B. Karrer and M. E. J. Newman, Phys. Rev. E83, 016107 (2011)

    work page 2011

  10. [10]

    Decelle, F

    A. Decelle, F. Krzakala, C. Moore, and L. Zdeborov´ a, Phys. Rev. E84, 066106 (2011)

    work page 2011

  11. [11]

    T. P. Peixoto, Phys. Rev. X4, 011047 (2014)

    work page 2014

  12. [12]

    E. Abbe, J. Mach. Learn. Res.18, 1 (2018)

    work page 2018

  13. [13]

    Vazquez, Phys

    A. Vazquez, Phys. Rev. E111, 064314 (2025)

    work page 2025

  14. [14]

    Vazquez, Phys

    A. Vazquez, Phys. Rev. E67, 056104 (2003)

    work page 2003

  15. [15]

    F. P. Ramsey, Proc. London Math. Soc.s2-30, 264 (1930)

    work page 1930

  16. [16]

    D. J. Watts and S. H. Strogatz, Nature393, 440 (1998)

    work page 1998