Pith. sign in

REVIEW 2 minor 61 references

Cavity equations for path measures give exact dynamical mean-field theory on tree graphs and the correct finite-time description on locally tree-like sparse graphs.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-27 17:38 UTC pith:MSMAUUVC

load-bearing objection This paper gives a cavity derivation for continuous-time DMFT on sparse graphs that treats reciprocity at the path level and recovers dense limits.

arxiv 2606.08689 v1 pith:MSMAUUVC submitted 2026-06-07 cond-mat.dis-nn

Dynamical cavity method for continuous-time complex systems on sparse random graphs

Fernando L. Metz , Isaac P\'erez Castillo This is my paper
classification cond-mat.dis-nn
keywords dynamical mean-field theorycavity methodsparse random graphscontinuous-time dynamicspath measuresstochastic processesreciprocal interactionspopulation dynamics
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper derives continuous-time cavity equations that reduce the stochastic dynamics of systems with general pairwise interactions on sparse random graphs to a set of self-consistent equations over trajectory probabilities. These equations are obtained by removing one node and writing the effect of its neighbors as an effective field whose statistics are determined by the remaining graph. The resulting closure is exact when the graph is a tree because each incoming branch is independent of the others. On graphs that are only locally tree-like the same equations become exact in the thermodynamic limit after ensemble averaging. The derivation makes explicit that the form of the effective field depends on whether edges are directed or reciprocal, with the latter case requiring conditional kernels that incorporate the receiver's past trajectory.

Core claim

We develop a continuous-time cavity derivation of sparse-network DMFT at the level of path measures for stochastic dynamics with general pairwise interactions on sparse random graphs. The cavity equations are exact on trees and yield the finite-time thermodynamic description on locally tree-like graphs. They show explicitly how reciprocity changes dynamical closure: fully directed graphs recover the sparse directed path-probability equation, whereas reciprocal or bidirected edges require conditional path kernels driven by the imposed history of the receiving node. Ensemble averaging gives laws over path-probability messages, with barycenters and higher-message moments closing by multilineari

What carries the argument

Path-probability messages whose ensemble averages close exactly by multilinearity and statistical independence of incoming branches on locally tree-like graphs.

Load-bearing premise

The graphs are locally tree-like, so that the incoming branches to any node remain statistically independent after the cavity removal.

What would settle it

Running exact Monte Carlo simulations of the same stochastic dynamics on a graph that contains many short loops and comparing the resulting trajectory distributions to those predicted by the cavity equations.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Fully directed graphs recover the known sparse directed path-probability equation.
  • Reciprocal or bidirected edges produce conditional path kernels that depend on the imposed history of the receiving node.
  • Ensemble averaging produces closed equations for the distribution of path-probability messages.
  • A causal discrete-time version yields population-dynamics algorithms that distinguish trajectory populations for directed graphs from conditional branch-law populations for reciprocal graphs.
  • High-connectivity limits of the same equations recover standard low-dimensional DMFT when path-level information is not required.

Where Pith is reading between the lines

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

  • The distinction between directed and reciprocal closures could be used to decide when a network model must retain full trajectory conditioning rather than single-time marginals.
  • Finite-memory truncations of the path kernels offer a practical route to numerical simulation of large sparse systems without full Monte Carlo sampling.
  • The same cavity construction may extend to dynamics that include higher-order interactions provided the local tree structure still holds.
  • Testing the equations on graphs whose loop density can be tuned continuously would map the range of validity of the locally tree-like assumption.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

0 major / 2 minor

Summary. The paper develops a continuous-time cavity derivation of sparse-network DMFT at the level of path measures for stochastic dynamics with general pairwise interactions on sparse random graphs. The cavity equations are exact on trees and close to yield a finite-time thermodynamic description on locally tree-like graphs via multilinearity and branch independence. It distinguishes directed vs. reciprocal cases (recovering sparse directed path-probability equations or requiring conditional path kernels), derives population-dynamics representations, formulates finite-memory closures tested on an additive RNN model, and recovers dense DMFT limits as projections.

Significance. If the derivation holds, the work supplies a rigorous path-measure framework for DMFT on sparse heterogeneous graphs that handles reciprocity explicitly and clarifies when dense closures apply, extending beyond Gaussian assumptions of dense DMFT. Strengths include the exact-on-trees claim, the population-dynamics representation, and the high-connectivity projection analysis, all grounded in standard locally tree-like assumptions without ad-hoc parameters.

minor comments (2)
  1. The abstract is information-dense; consider splitting the description of directed vs. reciprocal closures and the population-dynamics step into separate sentences for readability.
  2. The finite-memory closure and RNN test are mentioned but lack explicit equation numbers or parameter values in the provided abstract; ensure the main text labels these clearly (e.g., as Eq. (X) or Algorithm Y) so readers can reproduce the numerical check.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and detailed summary of our work, the recognition of its strengths (exactness on trees, population-dynamics representation, and high-connectivity projection), and the recommendation for minor revision. No specific major comments were listed in the report, so we have no point-by-point rebuttals to provide. We will incorporate any editorial or minor clarifications requested by the editor in the revised version.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained from tree cavity principles

full rationale

The paper derives continuous-time cavity equations for path measures on sparse graphs by starting from exactness on trees and closing under ensemble averaging via multilinearity plus branch independence on locally tree-like graphs. These closure steps are standard for cavity methods and are invoked precisely where the locally tree-like assumption is stated; they do not reduce to fitted inputs or self-referential definitions. No load-bearing self-citations, uniqueness theorems from the authors, or ansatzes smuggled via prior work appear in the derivation chain. The high-connectivity limit is obtained as a projection rather than a renaming of known results. The central claim therefore remains independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Review performed on abstract only; full derivation details unavailable. Ledger therefore limited to claims visible in abstract.

axioms (2)
  • standard math Cavity equations are exact on trees
    Stated directly in abstract as the foundation for the finite-time description.
  • domain assumption Graphs are locally tree-like
    Required for the ensemble-averaged closure via multilinearity and branch independence.

pith-pipeline@v0.9.1-grok · 5769 in / 1322 out tokens · 17023 ms · 2026-06-27T17:38:17.642909+00:00 · methodology

0 comments
read the original abstract

Dynamical mean-field theory (DMFT) reduces dense high-dimensional disordered dynamics to a self-consistent effective stochastic process. For sparse and heterogeneous networks, however, local fields contain finitely many strong inputs, so the Gaussian closure mechanisms of dense DMFT need not apply. We develop a continuous-time cavity derivation of sparse-network DMFT at the level of path measures for stochastic dynamics with general pairwise interactions on sparse random graphs. The cavity equations are exact on trees and yield the finite-time thermodynamic description on locally tree-like graphs. They show explicitly how reciprocity changes dynamical closure: fully directed graphs recover the sparse directed path-probability equation, whereas reciprocal or bidirected edges require conditional path kernels driven by the imposed history of the receiving node. Ensemble averaging gives laws over path-probability messages, with barycenters and higher-message moments closing by multilinearity and independence of incoming branches. A causal discrete-time derivation yields the corresponding population-dynamics representation, distinguishing trajectory populations for directed graphs from conditional branch-law or finite-depth tree populations for reciprocal graphs. We also formulate finite-memory numerical closures and test them in an additive-input recurrent neural network specialization. Finally, high-connectivity limits are obtained as projections of the sparse path-measure theory, clarifying when dense drift, noise, and response channels reduce to standard low-dimensional DMFT and when path-level descriptions remain essential.

Figures

Figures reproduced from arXiv: 2606.08689 by Fernando L. Metz, Isaac P\'erez Castillo.

Figure 1
Figure 1. Figure 1: Finite-depth causal tree for the bidirected population-dynamics construc [PITH_FULL_IMAGE:figures/full_fig_p037_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparison between direct finite-graph simulations and finite-depth pop [PITH_FULL_IMAGE:figures/full_fig_p037_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between direct finite-graph simulations and finite-depth popu [PITH_FULL_IMAGE:figures/full_fig_p038_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Graph-level view of the moving-window approximation. In the exact finite [PITH_FULL_IMAGE:figures/full_fig_p039_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison between direct finite-graph simulations and three finite [PITH_FULL_IMAGE:figures/full_fig_p042_5.png] view at source ↗

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

61 extracted references · 4 canonical work pages · 3 internal anchors

  1. [1]

    and Opper, M

    Eissfeller, H. and Opper, M. , title =. Physical Review Letters , volume =. 1992 , doi =

    1992

  2. [2]

    , title =

    De Dominicis, C. , title =. Physical Review B , volume =. 1978 , doi =

    1978

  3. [3]

    Newman, M. E. J. and Strogatz, S. H. and Watts, D. J. , title =. Physical Review E , volume =. 2001 , doi =. cond-mat/0007235 , archivePrefix =

    work page internal anchor Pith review Pith/arXiv arXiv 2001

  4. [4]

    Configuring Random Graph Models with Fixed Degree Sequences

    Fosdick, Bailey K. and Larremore, Daniel B. and Nishimura, Joel and Ugander, Johan , title =. SIAM Review , volume =. 2018 , doi =. 1608.00607 , archivePrefix =

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  5. [5]

    Random Structures & Algorithms , volume =

    Bordenave, Charles and Lelarge, Marc , title =. Random Structures & Algorithms , volume =. 2010 , doi =

    2010

  6. [6]

    Karrer, Brian and Newman, M. E. J. , title =. Physical Review E , volume =. 2010 , doi =

    2010

  7. [7]

    Lokhov, Andrey Y. and M. Inferring the origin of an epidemic with a dynamic message-passing algorithm , journal =. 2014 , doi =

    2014

  8. [8]

    Cavity master equation for the continuous time dynamics of discrete-spin models , journal =

    Aurell, Erik and Del Ferraro, Gino and Dom. Cavity master equation for the continuous time dynamics of discrete-spin models , journal =. 2017 , doi =

    2017

  9. [9]

    Physical Review E , volume =

    Barthel, Thomas and De Bacco, Caterina and Franz, Silvio , title =. Physical Review E , volume =. 2018 , doi =

    2018

  10. [10]

    Journal of Statistical Mechanics: Theory and Experiment , volume =

    Barthel, Thomas , title =. Journal of Statistical Mechanics: Theory and Experiment , volume =. 2020 , doi =

    2020

  11. [11]

    PRX Life , volume =

    Tonolo, Tommaso and Angelini, Maria Chiara and Azaele, Sandro and Maritan, Amos and Gradenigo, Giacomo , title =. PRX Life , volume =. 2026 , doi =

    2026

  12. [12]

    2025 , eprint =

    Tarabolo, Mattia and Dall'Asta, Luca and Mulet, Roberto , title =. 2025 , eprint =

    2025

  13. [13]

    and Crisanti, A

    Sompolinsky, H. and Crisanti, A. and Sommers, H. J. , title =. Physical Review Letters , volume =. 1988 , doi =

    1988

  14. [14]

    Physical Review E , volume =

    Bunin, Guy , title =. Physical Review E , volume =. 2017 , doi =

    2017

  15. [15]

    Reviews of Modern Physics , volume =

    Pastor-Satorras, Romualdo and Castellano, Claudio and Van Mieghem, Piet and Vespignani, Alessandro , title =. Reviews of Modern Physics , volume =. 2015 , doi =

    2015

  16. [16]

    and Peron, Thomas K

    Rodrigues, Francisco A. and Peron, Thomas K. D. M. and Ji, Peng and Kurths, J. The. Physics Reports , volume =. 2016 , doi =

    2016

  17. [17]

    , title =

    Metz, Fernando L. , title =. Physical Review Letters , volume =. 2025 , doi =

    2025

  18. [18]

    Newman, Mark E. J. , title =. 2010 , isbn =

    2010

  19. [19]

    Newman, M. E. J. and Strogatz, S. H. and Watts, D. J. , title =. Physical Review E , volume =. 2001 , doi =

    2001

  20. [20]

    and Larremore, Daniel B

    Fosdick, Bailey K. and Larremore, Daniel B. and Nishimura, Joel and Ugander, Johan , title =. SIAM Review , volume =. 2018 , doi =

    2018

  21. [21]

    The Structure of Ecological Networks Across Levels of Organization , journal =

    Guimar. The Structure of Ecological Networks Across Levels of Organization , journal =. 2020 , doi =

    2020

  22. [22]

    Science Advances , volume =

    Asllani, Malbor and Lambiotte, Renaud and Carletti, Timoteo , title =. Science Advances , volume =. 2018 , doi =

    2018

  23. [23]

    Hatchett, J. P. L. and Wemmenhove, B. and P. Parallel dynamics of disordered. Journal of Physics A: Mathematical and General , volume =. 2004 , doi =

    2004

  24. [24]

    Hatchett, J. P. L. and P. Dynamical Replica Analysis of Disordered. Physical Review Letters , volume =. 2005 , doi =. cond-mat/0504313 , archivePrefix =

    work page internal anchor Pith review Pith/arXiv arXiv 2005

  25. [25]

    and Boll

    Neri, I. and Boll. The cavity approach to parallel dynamics of. Journal of Statistical Mechanics: Theory and Experiment , year =

  26. [26]

    SciPost Physics , volume =

    Tarabolo, Mattia and Dall'Asta, Luca , title =. SciPost Physics , volume =. 2025 , doi =

    2025

  27. [27]

    Martin, P. C. and Siggia, E. D. and Rose, H. A. , title =. Physical Review A , volume =. 1973 , doi =

    1973

  28. [28]

    Janssen, H. K. , title =. Zeitschrift f. 1976 , doi =

    1976

  29. [29]

    and Roudi, Yasser and Sollich, Peter , title =

    Hertz, John A. and Roudi, Yasser and Sollich, Peter , title =. Journal of Physics A: Mathematical and Theoretical , volume =. 2017 , doi =

    2017

  30. [30]

    Information, Physics, and Computation , publisher =

    M. Information, Physics, and Computation , publisher =. 2009 , isbn =

    2009

  31. [31]

    and Siggia, E

    Martin, Paul C. and Siggia, E. D. and Rose, H. A. , title =. Physical Review A , volume =. 1973 , doi =

    1973

  32. [32]

    Zeitschrift f

    Janssen, Hans-Karl , title =. Zeitschrift f. 1976 , doi =

    1976

  33. [33]

    , title =

    De Dominicis, C. , title =. Journal de Physique Colloques , volume =. 1976 , doi =

    1976

  34. [34]

    and Sompolinsky, H

    Crisanti, A. and Sompolinsky, H. , title =. Physical Review E , volume =. 2018 , doi =

    2018

  35. [35]

    Physical Review X , volume =

    Kadmon, Jonathan and Sompolinsky, Haim , title =. Physical Review X , volume =. 2015 , doi =

    2015

  36. [36]

    Journal of Physics A: Mathematical and Theoretical , volume =

    Roy, Felix and Biroli, Giulio and Bunin, Guy and Cammarota, Chiara , title =. Journal of Physics A: Mathematical and Theoretical , volume =. 2019 , doi =

    2019

  37. [37]

    Physical Review Letters , volume =

    Altieri, Ada and Roy, Felix and Cammarota, Chiara and Biroli, Giulio , title =. Physical Review Letters , volume =. 2021 , doi =

    2021

  38. [38]

    Hatchett, J. P. L. and Wemmenhove, B. and P. Parallel dynamics of disordered Ising spin systems on finitely connected random graphs , journal =. 2004 , doi =

    2004

  39. [39]

    and Boll

    Neri, I. and Boll. The cavity approach to parallel dynamics of Ising spins on a graph , journal =. 2009 , doi =

    2009

  40. [40]

    Mimura, Kazushi and Coolen, A. C. C. , title =. Journal of Physics A: Mathematical and Theoretical , volume =. 2009 , doi =

    2009

  41. [41]

    Journal of Statistical Mechanics: Theory and Experiment , volume =

    Roudi, Yasser and Hertz, John , title =. Journal of Statistical Mechanics: Theory and Experiment , volume =. 2011 , doi =

    2011

  42. [42]

    Physical Review E , volume =

    Aurell, Erik and Mahmoudi, Hamed , title =. Physical Review E , volume =. 2012 , doi =

    2012

  43. [43]

    Physical Review E , volume =

    Del Ferraro, Gino and Aurell, Erik , title =. Physical Review E , volume =. 2015 , doi =

    2015

  44. [44]

    Backtracking Dynamical Cavity Method , journal =

    Behrens, Freya and Hudcov. Backtracking Dynamical Cavity Method , journal =. 2023 , doi =

    2023

  45. [45]

    Europhysics Letters , volume =

    Derrida, Bernard and Gardner, Elizabeth and Zippelius, Annette , title =. Europhysics Letters , volume =. 1987 , doi =

    1987

  46. [46]

    The European Physical Journal B , volume =

    The Bethe lattice spin glass revisited , author =. The European Physical Journal B , volume =

  47. [47]

    Journal of Statistical Physics , volume =

    The cavity method at zero temperature , author =. Journal of Statistical Physics , volume =

  48. [48]

    SIAM Review , volume =

    Configuring random graph models with fixed degree sequences , author =. SIAM Review , volume =. 2018 , doi =

    2018

  49. [49]

    Physical Review , volume=

    The spherical model of a ferromagnet , author=. Physical Review , volume=. 1952 , publisher=

    1952

  50. [50]

    Zeitschrift f

    The sphericalp-spin interaction spin glass model: the statics , author=. Zeitschrift f. 1992 , publisher=

    1992

  51. [51]

    Zeitschrift f

    The spherical p-spin interaction spin-glass model , author=. Zeitschrift f. 1993 , publisher=

    1993

  52. [52]

    Physical Review B , volume=

    p-spin-interaction spin-glass models: Connections with the structural glass problem , author=. Physical Review B , volume=. 1987 , publisher=

    1987

  53. [53]

    Journal of Physics A: Mathematical and General , volume=

    Violation of the fluctuation--dissipation theorem in glassy systems: basic notions and the numerical evidence , author=. Journal of Physics A: Mathematical and General , volume=. 2003 , publisher=

    2003

  54. [54]

    Journal of Physics C: Solid State Physics , volume=

    Phase diagrams for dilute spin glasses , author=. Journal of Physics C: Solid State Physics , volume=. 1985 , publisher=

    1985

  55. [55]

    Physical Review B , volume=

    Dynamics as a substitute for replicas in systems with quenched random impurities , author=. Physical Review B , volume=. 1978 , publisher=

    1978

  56. [56]

    Slow Relaxations and nonequilibrium dynamics in condensed matter , pages=

    Course 7: Dynamics of glassy systems , author=. Slow Relaxations and nonequilibrium dynamics in condensed matter , pages=. 2003 , publisher=

    2003

  57. [57]

    Physical Review Research , volume=

    Spectral density of dense random networks and the breakdown of the Wigner semicircle law , author=. Physical Review Research , volume=. 2020 , publisher=

    2020

  58. [58]

    Journal of Physics: Complexity , volume=

    Mean-field theory of vector spin models on networks with arbitrary degree distributions , author=. Journal of Physics: Complexity , volume=. 2022 , publisher=

    2022

  59. [59]

    Journal of Physics: Complexity , volume=

    Analytic solution of the resolvent equations for heterogeneous random graphs: spectral and localization properties , author=. Journal of Physics: Complexity , volume=. 2022 , publisher=

    2022

  60. [60]

    arXiv preprint arXiv:2212.09424 , year=

    Nonequilibrium dynamics of the Ising model on heterogeneous networks with an arbitrary distribution of threshold noise , author=. arXiv preprint arXiv:2212.09424 , year=

    work page arXiv

  61. [61]

    2021 , publisher=

    Quantum field theory and critical phenomena , author=. 2021 , publisher=

    2021