REVIEW 5 major objections 6 minor 14 references
Universality of scaling of correlations across probability distributions
T0 review · 5 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read This paper shows that the two-point correlation exponent of any translation-, rotation-, and scale-invariant lattice distribution over finitely many states per site matches the critical exponent of an equilibrium statistical-mechanical…
desk verdict Bold question, but the central assumption (Prop. 3) is the conclusion; the proof collapses once nonlocal effective actions are admitted. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is the Fourier representation of a translation-invariant distribution, combined with a path-integral identity. Each mode of the lattice variable is labelled by source fields $J$; the paper defines a mode-resolved connected correlation $C(|a-b|)_J$ and asserts it decays as $e^{-\lambda_J |a-b|}/|a-b|^{\alpha_J}$ (Proposition 3). The partition sum is then rewritten as an integral over auxiliary fields $\psi_i$, and in the continuum limit the accumulated action is asserted to be local, with the infrared behaviour controlled only by $c_1 \nabla\psi\cdot\nabla\psi+c_2\psi^2$ (Eq. 19). This locality is what lets the paper invoke standard effective-field-theory decay, and the derivative-matching argument then forces all contributing modes to have the same $\alpha$, with at least one mode massless.
What would settle it
Compute the connected two-point correlation for a deliberately non-Boltzmann lattice distribution—for example, the uniform distribution over all configurations of a 2D lattice with fixed total magnetization—and compare the power-law exponent with the equilibrium critical exponent expected from the system's symmetry. A mismatch would falsify the claim; expanding the effective action beyond lowest order could also reveal whether non-local terms such as $\int d^d x\,\psi(x)\psi(y)/|x-y|^\sigma$ survive in the infrared.
Extended reading notes
Core claim
On its own terms, the paper argues for a universality result for arbitrary lattice probability distributions. Let $f(\{S_i\})$ be a translation-, rotation-, and scale-invariant distribution over configurations whose site variables each take the same finite set of values. The paper's central claim is that at large separation the connected correlation $\langle S_a S_b\rangle-\langle S_a\rangle\langle S_b\rangle$ decays as $C/|a-b|^\alpha$, with the same $\alpha$ that a critical equilibrium model with Boltzmann weights would have in that symmetry class. The proof Fourier-expands $f$, writes the mode-resolved correlation as $e^{-\lambda_J |a-b|}/|a-b|^{\alpha_J}$ (Proposition 3), converts the partition sum into a path integral over auxiliary fields, reduces the continuum theory to the local action $c_1 \nabla\psi\cdot\nabla\psi+c_2\psi^2$, and then uses repeated differentiation to force all contributing modes to share one $\alpha$, with at least one mode at $\lambda=0$. The paper concludes that arbitrary non-Boltzmann distributions fall into the universality classes of critical Boltzmann systems.
Load-bearing premise
The load-bearing premise is that each Fourier component's connected correlation decays as a pure exponential over a power law at large distances and that the summed lattice distribution reduces to a strictly local field theory whose infrared behaviour has only two terms; if either fails, the universality claim collapses.
Editorial extensions
If this is right
- Measured correlation exponents in non-thermal scale-invariant lattice systems—neural populations, natural images, or driven granular and biological assemblies—should match the exponents of equilibrium statistical-mechanical universality classes whenever the stated symmetries and finite-state condition hold.
- The Boltzmann weight is not special: any distribution with the same symmetries produces the same long-distance decay, so universality-class exponents can be computed from whichever lattice distribution is analytically most convenient.
- Any such system that displays algebraic decay must sit at an effective critical point, in the sense that at least one Fourier mode of its distribution has $\lambda=0$.
- The result turns universality into a tool for calculation: choosing a simple non-Boltzmann distribution should yield the same critical exponent as the corresponding equilibrium model, potentially bypassing difficult Boltzmann sums.
Reading between the lines
- A natural extension not pursued in the paper is to continuous or unbounded local variables; if the local-effective-action reduction survives, the same exponent identity should hold for compact continuous target spaces.
- The derivative-matching argument also suggests that subleading corrections are universal: modes with $\lambda_J>0$ contribute only exponentially damped terms with the same power-law prefactor, so the approach to the asymptotic power law may be governed by a single length scale.
- If the locality assumption fails for some symmetric non-Boltzmann distribution, the likely observable signature would be an anomalous exponent that depends on the distribution's higher-order moments; a systematic scan over such moments would test the boundary of the claim.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript claims that for any lattice system with finite-valued site variables governed by an arbitrary translation-, rotation-, and scale-invariant probability distribution, the large-distance decay exponent of the two-point correlation function is the same as the critical exponent of an equilibrium statistical mechanical model in the same universality class. The argument proceeds by Fourier-expanding the arbitrary distribution, introducing an auxiliary field to rewrite the partition function as a path integral, asserting that the resulting effective action has the local form of Eq. (19), and then using repeated differentiation of the correlation function identity (Eq. (21)) to show that all per-mode exponents alpha_J coincide. The paper concludes that arbitrary non-Boltzmann distributions belong to statistical-mechanics universality classes.
Significance. If established, the claim would represent a major extension of the concept of universality beyond Gibbs-Boltzmann statistics, with potential implications for criticality in biological, neural, and other data-driven systems. The paper is clearly structured and attempts to present the argument as a series of propositions. However, the central steps are not derivations but assumptions: Proposition 3's asymptotic form and the local effective action in Eq. (19) are posited rather than proven; Proposition 5's derivation is circular; and the proof assumes a power-law correlation function rather than deriving it from scale invariance. The manuscript also contains no numerical or exactly solvable consistency check. At present, the core claim is unsupported.
major comments (5)
- [Proposition 1 and Eq. (8)] Proposition 1 is not established by the argument given. From Eq. (6), the constancy of <S_a> in a implies only that the integral of the corresponding integrand is independent of a; it does not imply that a({J_i}) is invariant under J_i -> J_{i+1}. The same objection applies to the higher correlation functions. Therefore the Fourier representation Eq. (8), which underpins all subsequent steps, is not justified for arbitrary probability distributions.
- [Proposition 3, Eq. (14), and Eq. (19)] Proposition 3 is an assumption, not a consequence. The only support is the claim that after summing over the S_i the effective action reduces to the local form Eq. (19). But integrating out the S_i from Eq. (18) yields an action S(psi) that is generically nonlocal: the J couplings in Eq. (8) have arbitrary ranges, so kernels such as integral dx dy psi(x) K(x-y) psi(y) with non-polynomial K survive. A nonlocal kernel with Fourier transform ~ |k|^sigma changes the scaling dimension of psi and can alter the correlation exponent; it cannot be dismissed as an irrelevant higher-derivative term by the cited EFT power-counting argument, since that argument assumes a local derivative expansion to begin with. Thus Eq. (14) effectively assumes the conclusion that the arbitrary distribution defines a local statistical field theory.
- [Proposition 5, Eqs. (22)-(27)] The step from Eq. (22) to Eq. (23) drops the lambda_J terms by asserting that they 'dominate' and therefore must vanish separately for consistency. This is not a valid inference: the equality must hold for the entire integral, not for each integrand, and the relative contributions of the lambda_J and alpha_J terms depend on the unknown distribution of (lambda_J, alpha_J) over the support of a(J) Z(J). Moreover, the only way the argument can work is if a mode with lambda_J = 0 exists, which is exactly Proposition 6; using that to drop the lambda_J terms makes the derivation circular. Equations (24)-(27) inherit this problem, so the conclusion alpha_J = alpha is not established.
- [Proposition 6 and Eq. (21)] The paper assumes from the outset that the left-hand side of Eq. (21) is C/|a-b|^alpha, i.e., a pure power law. This is a critical-scaling form; it is not derived from the assumed scale invariance of the probability distribution, and no argument is given that the original lattice distribution is at its critical point. Proposition 6 then 'proves' that at least one lambda_J = 0, which is equivalent to the existence of a massless mode, i.e., criticality. Since the power-law ansatz is already an assumption of criticality, the proof does not extend universality to arbitrary non-critical distributions; it only re-derives a known statement about critical field theories under the locality assumptions of Proposition 3.
- [From one to higher dimensions, around Eq. (19)] The extension of Proposition 2 to two and three dimensions is asserted rather than proved. The text states that the terms in Eq. (8) must be 'isotropic enough' to be reducible to a rotational and translationally invariant field theory, but no condition on the J couplings is given. A lattice model with only discrete rotational symmetry (e.g., a square lattice) can have a continuum limit with continuous rotation invariance for local terms, but nonlocal terms need not respect the enlarged symmetry. The derivation of Eq. (19) in d dimensions is therefore incomplete.
minor comments (6)
- [Eq. (15)] The notation 'psi in [-infinity, infinity]' is imprecise; it should read 'psi_i in (-infinity, infinity)' for each lattice site i.
- [Notation throughout] The notation 'J1, J 2, J3...' and the statement that 'J2 represents all J_m^2' is confusing; a consistent multi-index notation should be introduced before Eq. (8).
- [Proposition 2] The claim that the listed translation-invariant sums cannot be written as products of each other is imprecise; for example, powers of sum_i S_i generate product terms, and the argument is not needed for the main result.
- [After Eq. (19)] The sentence 'We didn't consider rotational invariance until now as we were considering one dimensional systems for simplicity' appears after Eq. (19), which is already written in d dimensions; the exposition is inconsistent.
- [Epsilon prescription, Eq. (10)] The operator Gamma is defined for x>0 and x<0, but not at x=0; this should be specified.
- [References] Reference [13] is cited for the effective-field-theory argument, but no particular section or theorem is identified, and the cited argument concerns local effective actions, which is exactly the point at issue.
Circularity Check
The claimed universality is loaded in through Proposition 3: arbitrary distributions are assumed to reduce to the local action Eq. (19), and Proposition 5 then discards the λ terms that spoil the power law. The conclusion is an input, not a derivation.
-
ansatz smuggled in via citation
[Proposition 3, Eqs. (15)-(20), especially Eq. (19) and the paragraph after it]
"The RHS in the second equation after summing over all Si’s can be written as e^{−iS(..ψi,ψi+1,...)}, where S(..ψi,ψi+1,... ) is a function of all values of ψi’s. Because every term in every exponent in the integral above is translation invariant S(..ψi,ψi+1,... ) is translation invariant, in the continuum limit we have S = ∫ ddx[c1▽ψ·▽ψ +c2ψ2 + ∑ m,n.p cmnpψm(▽ψ·▽ψ)n▽2pψ], (19) ... It is well known that for a system described by a field theory, the correlation function at long distances r scales as ∼ e−λJ1,J2,J3..r / rαJ1,J2,J3.. . ... This furnishes proof of C(|a−b|)J1,J2,J3.."
Proposition 3 is the load-bearing input for every later proposition, but its proof is not derived from the arbitrary distribution f. It asserts that after integrating out the S_i variables the auxiliary-field action reduces to the local continuum action Eq. (19), a c1(∇ψ)^2 + c2ψ^2 theory whose critical correlations are precisely the Boltzmannian statistical-mechanical universality-class exponents the paper claims to derive. For a genuinely arbitrary translation-invariant f the effective action need not be local; nonlocal kernels can change the scaling dimension of ψ and hence the correlation exponent. By assuming Eq.
-
self definitional
[Proposition 5, Eqs. (21)-(27); especially the consistency step after Eq. (22)]
"Consistency for all possible values of |a−b| in the last equation (which are still large enough), requires ∫ ΠjdJj a(...Ji,Ji+1...) Z(J1,J2...) λ_{J1,J2,J3..} e^{−λ_{J1,J2,J3..}|a−b|}/|a−b|^{α_{J1,J2,J3..}} + (∗) = 0, as it dominates over ∫ ΠjdJj a(J1,J2...) Z(J1,J2...) (−α_{J1,J2,J3..}) e^{−λ_{J1,J2,J3..}|a−b|}/|a−b|^{α_{J1,J2,J3..}+1}."
To conclude α_J = α, the paper removes the λ_J-dependent terms from the moment equations, asserting that 'consistency' with Eq. (21) requires the λ integral to vanish. But Eq. (21) already assumes the full correlation is the pure power law C/|a−b|^α; for a generic superposition of sectors with λ_J > 0, the long-distance behavior is exponential, not power law. Demanding that the λ integral vanish is equivalent to assuming that a contributing sector has λ_J = 0 and that the other sectors do not affect the leading power law—exactly the content of Proposition 6, which is proved only afterward. Proposition 5 therefore presupposes Proposition 6, and the derivation of a single universality-class exponent presupposes the critical, massless behavior it is meant to establish.
full rationale
The paper contains no fitted parameters and does not rely on a self-citation chain; references [12] and [13] are standard textbooks. The circularity is structural rather than bibliographic. The central theorem is made to follow from Proposition 3, but Proposition 3 is obtained by postulating that the auxiliary-field action of an arbitrary translation-invariant distribution collapses to the local continuum action Eq. (19), a c1(∇ψ)^2 + c2ψ^2 theory whose critical correlations are the very Boltzmannian universality-class exponents the paper claims to derive. Nothing in the Fourier expansion Eq. (8) or the Hubbard-Stratonovich identity Eqs. (15)-(18) shows that the general f has no nonlocal kernels or relevant anisotropic terms; for genuinely arbitrary distributions the effective action need not be local, so the claimed universal exponent is unsupported. The later steps compound the problem: Proposition 5 discards the λ_J terms by declaring that consistency with the already-assumed pure power law requires them to vanish, which is equivalent to assuming a massless mode, i.e., Proposition 6, a proposition proved only subsequently. The proof therefore reduces the claim 'arbitrary distributions match statistical-mechanical exponents' to the assumption that arbitrary distributions are already local statistical-mechanical field theories. This is partial circularity: the Fourier/Hubbard-Stratonovich representation is independent scaffolding, but the load-bearing reduction Eq. (19) and the λ-term elimination make the main prediction an input rather than a derived result.
Assumptions & free parameters
assumptions (5)
- domain assumption The site variables S_i take the same finite range of values at every lattice site, and the probability distribution f({S_i}) is translation invariant under S_i -> S_{i+1}.
- ad hoc to paper The continuum limit of the lattice theory is a local, translation- and rotation-invariant scalar field theory with action S = integral d^d x [c1 grad psi dot grad psi + c2 psi^2 + relevant and marginal terms].
- ad hoc to paper For each Fourier mode J, the connected correlation C_J(|a-b|) has the asymptotic form e^{-lambda_J |a-b|} / |a-b|^{alpha_J} with lambda_J >= 0.
- domain assumption At long distances, only relevant and marginal operators in the effective action matter.
- ad hoc to paper The epsilon prescription (1 + i epsilon Gamma) regularizes otherwise ill-defined sums and does not alter physical results as epsilon -> 0+.
Cite this review
Pith. "Pith review of Universality of scaling of correlations across probability distributions." pith.science (2026). https://pith.science/paper/IJCNDANJ
@misc{pith2026190806025,
author = {Pith},
title = {Pith review of: Universality of scaling of correlations across probability distributions},
year = {2026},
howpublished = {\url{https://pith.science/paper/IJCNDANJ}},
note = {Machine review of arXiv:1908.06025}
}
read the original abstract
Scale invariance and the resulting power law behaviours are seen in diverse systems. In this work we consider translation, rotational and scale invariant systems defined on a lattice, such that the variables defining the state at every lattice site take on the same range of finite values, with these values collectively picked up from probability distribution that can be arbitrary. We show that the exponent that describes the scaling of the two point correlation function in these systems will match the scaling exponent of a equilibrium statistical mechanical model described by a Boltzmannian distribution at criticality. This work therefore extends the concept of universality in statistical mechanics to probability distributions that do not have a Boltzmannian form.
Reference graph
Works this paper leans on
-
[1]
Khaluf, Yara, et al. ”Scale invariance in natural and artificial collective systems: a review.” Journal of the royal society interface 14.136 (2017): 20170662
work page 2017
-
[2]
”Discrete scale invariance in supercritical percolation.” New Journal of Physics 18.1 (2016): 013042
Schr¨ oder, Malte, Wei Chen, and Jan Nagler. ”Discrete scale invariance in supercritical percolation.” New Journal of Physics 18.1 (2016): 013042
work page 2016
-
[3]
Khatami, Fatemeh, et al. ”Origins of scale invariance in vocalization sequences and speech.” PLoS computational biology 14.4 (2018): e1005996
work page 2018
-
[4]
Lin, Hai-Nan, and Yu Sang. ”Scale-invariance in the repeating fast radio burst 121102.” Monthly Notices of the Royal Astronomical Society 491.2 (2020): 2156-2161
work page 2020
-
[5]
Djenidi, L., R. A. Antonia, and S. L. Tang. ”Scale invariance in finite Reynolds number homogeneous isotropic turbulence.” Journal of Fluid Mechanics 864 (2019): 244-272
work page 2019
-
[6]
”Scaling Laws in River Runoff.” EGUGA (2018): 7707
Coutandin, Thomas, and Carlo Albert. ”Scaling Laws in River Runoff.” EGUGA (2018): 7707
work page 2018
-
[7]
Jhun, Jennifer, Patricia Palacios, and James Owen Weatherall. ”Market crashes as critical phenomena? Explanation, idealization, and universality in econophysics.” Synthese 195.10 (2018): 4477-4505
work page 2018
-
[8]
Mora, T., & Bialek, W. (2011). Are biological systems poised at criticality?. Journal of Statistical Physics, 144(2), 268-302
work page 2011
Show all 14 references
-
[9]
Lubensky, and Thomas A
Chaikin, Paul M., Tom C. Lubensky, and Thomas A. Witten. Principles of condensed matter physics. Vol. 10. Cambridge: Cambridge university press, 1995
1995
-
[10]
Cardy, J. L. (1988). Conformal invariance and statistical mechanics. Les Houches
1988
-
[11]
”The conformal bootstrap.” Nature Physics 12.6 (2016): 535-539
Poland, David, and David Simmons-Duffin. ”The conformal bootstrap.” Nature Physics 12.6 (2016): 535-539
2016
-
[12]
Altland, A., Simons, B. D. (2010). Condensed matter field theory. Cambridge university press pg 196-199
2010
-
[13]
”An introduction to effective field theory.” Annu
Burgess, Cliff P. ”An introduction to effective field theory.” Annu. Rev. Nucl. Part. Sci. 57 (2007): 329-362
2007
-
[14]
Stephens G J, Mora T, Tkacik G and Bialek W, 2008 arXiv:0806.2694 [q-bio.NC]
2008 arXiv
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.