Non-existence probabilities and lower tails in the critical regime via Belief Propagation
Pith reviewed 2026-05-10 19:43 UTC · model grok-4.3
The pith
The probability that a random hypergraph subset induces far fewer edges than expected can be approximated by the Bethe free energy at the unique fixed point of belief propagation.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Under simple structural conditions on the hypergraph and an upper bound on p determined by a phase transition in the hard-core model on the infinite k-uniform, Delta-regular, linear hypertree, the lower-tail probability that a p-random subset induces significantly fewer hyperedges than expected is accurately approximated by the Bethe free energy evaluated at the unique fixed point of a Belief Propagation operator on the hypergraph.
What carries the argument
The unique fixed point of the Belief Propagation operator on the hypergraph, where the Bethe free energy is evaluated to obtain the logarithmic asymptotics.
If this is right
- Logarithmic asymptotics for non-existence probabilities of subgraphs in random graphs become computable via belief propagation.
- The same asymptotics apply to the number of k-term arithmetic progressions in random subsets of the integers.
- The method fills the gap between hypergraph container techniques and Janson's inequality for a wide range of parameters.
- Efficient numerical evaluation of the fixed point replaces direct enumeration or moment calculations for these tail events.
Where Pith is reading between the lines
- The same fixed-point analysis could supply tail estimates for non-uniform or weighted hypergraph models.
- Variational characterizations of the Bethe free energy may link these combinatorial tails to phase-transition phenomena studied in statistical physics.
- Numerical solution of the belief-propagation equations on finite instances could be used to test the sharpness of the derived asymptotics.
Load-bearing premise
The hypergraph satisfies simple structural conditions and the density parameter p remains below the phase-transition threshold of the hard-core model on the infinite regular hypertree.
What would settle it
For any concrete hypergraph meeting the stated conditions, compute the true lower-tail probability numerically and check whether its logarithm equals the Bethe free energy value at the unique belief-propagation fixed point.
Figures
read the original abstract
We compute the logarithmic asymptotics of the non-existence probability (and more generally the lower-tail probability) for a wide variety of combinatorial problems for a range of parameters in the `critical regime' between the regime amenable to hypergraph container methods and that amenable to Janson's inequality. Examples include lower tails and non-existence probabilities for subgraphs of random graphs and for $k$-term arithmetic progressions in random sets of integers. Our methods apply in the general framework of estimating the probability that a $p$-random subset of vertices in a $k$-uniform hypergraph induces significantly fewer hyperedges than expected. We show that under some simple structural conditions on the hypergraph and an upper bound on $p$ determined by a phase transition in the hard-core model on the infinite $k$-uniform, $\Delta$-regular, linear hypertree, this probability can be accurately approximated by the Bethe free energy evaluated at the unique fixed point of a Belief Propagation operator on the hypergraph.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims to compute the logarithmic asymptotics of non-existence probabilities (and more generally lower-tail probabilities) for combinatorial problems in the critical regime between hypergraph container methods and Janson's inequality. In the general setting of a p-random subset of vertices in a k-uniform hypergraph inducing significantly fewer hyperedges than expected, the log-asymptotics equal the Bethe free energy evaluated at the unique fixed point of a Belief Propagation operator, provided the hypergraph satisfies simple structural conditions and p is bounded above by the phase transition threshold of the hard-core model on the infinite k-uniform, Δ-regular, linear hypertree. The result is illustrated on lower tails for subgraphs of random graphs and for k-term arithmetic progressions in random sets of integers.
Significance. If the central approximation holds, the work supplies a new analytic tool for tail probabilities in a regime where existing combinatorial methods are known to be insufficient. The explicit link to the hard-core model phase transition on the infinite hypertree and the use of the Bethe free energy at the BP fixed point provide a concrete, computable expression for the rate function. This framework has the potential to unify and extend results across several areas of probabilistic combinatorics, including random graphs and additive combinatorics, and the absence of free parameters or ad-hoc fitting in the stated claim is a methodological strength.
minor comments (2)
- [Abstract] The abstract and introduction would benefit from a brief, explicit statement of the error term (or the sense in which the approximation is 'accurate') rather than leaving it implicit in the phrase 'accurately approximated'.
- [Introduction] Notation for the BP message-update rule and the precise definition of the Bethe free energy functional should be introduced with a self-contained display equation early in the technical sections to assist readers who are not already familiar with the statistical-physics literature.
Simulated Author's Rebuttal
We thank the referee for the positive summary, the assessment of significance, and the recommendation of minor revision. We are pleased that the potential of the Bethe free energy approximation at the BP fixed point to unify results across probabilistic combinatorics is recognized. Since the report lists no specific major comments, we have no points requiring rebuttal or revision at this stage.
Circularity Check
No significant circularity
full rationale
The paper derives the logarithmic asymptotics of the non-existence probability by showing it equals the Bethe free energy at the unique fixed point of a Belief Propagation operator on the hypergraph, under structural conditions on the hypergraph and an upper bound on p taken from the hard-core model phase transition on the infinite k-uniform Δ-regular linear hypertree. This reference is external and independent; the target probability is not defined in terms of the Bethe free energy or BP fixed point, nor is any parameter fitted to the result itself and then relabeled as a prediction. No load-bearing step reduces by construction to a self-citation, ansatz smuggled via citation, or renaming of a known result. The central claim is self-contained and externally falsifiable on the listed combinatorial examples.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption simple structural conditions on the hypergraph
- domain assumption upper bound on p from hard-core phase transition on infinite hypertree
Reference graph
Works this paper leans on
- [1]
- [2]
- [3]
- [4]
-
[5]
F. Bencs and P. Buys. Optimal zero-free regions for the independence polynomial of bounded degree hypergraphs.Random Structures & Algorithms, 66(4):e70018, 2025. 16, 18, 19
work page 2025
-
[6]
P. Bennett and T. Bohman. A note on the random greedy independent set algorithm.Random Structures & Algorithms, 49(3):479–502, 2016. 8, 15
work page 2016
-
[7]
I. Bez´ akov´ a, A. Galanis, L. A. Goldberg, H. Guo, and D. Stefankovic. Approximation via correlation decay when strong spatial mixing fails.SIAM Journal on Computing, 48(2):279–349, 2019. 14
work page 2019
-
[8]
T. Bohman and P. Keevash. The early evolution of the H-free process.Inventiones mathematicae, 181(2):291–336, 2010. 43
work page 2010
-
[9]
S. Chatterjee and S. S. Varadhan. The large deviation principle for the Erd˝ os-R´ enyi random graph. European Journal of Combinatorics, 32(7):1000–1017, 2011. 4
work page 2011
-
[10]
Z. Chen, K. Liu, N. Mani, and A. Moitra. Strong spatial mixing for colorings on trees and its algorithmic applications. In2023 IEEE 64th Annual Symposium on Foundations of Computer Science (FOCS), pages 810–845. IEEE, 2023. 14
work page 2023
-
[11]
A. Coja-Oghlan and W. Perkins. Belief propagation on replica symmetric random factor graph models. Annales de l’institut Henri Poincare D, 5(2):211–249, 2018. 15
work page 2018
-
[12]
A. Dembo and A. Montanari. Ising models on locally tree-like graphs.The Annals of Applied Probability, 20(2):565–592, 2010. 15
work page 2010
- [13]
- [14]
-
[15]
M. Dyer, A. Sinclair, E. Vigoda, and D. Weitz. Mixing in time and space for lattice spin systems: A combinatorial view.Random Structures & Algorithms, 24(4):461–479, 2004. 13
work page 2004
-
[16]
P. Erd˝ os, D. Kleitman, and B. Rothschild. Asymptotic enumeration ofK n-free graphs.Colloquio Inter- nazionale sulle Teorie Combinatorie (Rome, 1973), (17):19–27, 1973. 2
work page 1973
-
[17]
A. Galanis, D. ˇStefankoviˇ c, and E. Vigoda. Inapproximability of the partition function for the antifer- romagnetic Ising and hard-core models.Combinatorics, Probability and Computing, 25(4):500–559, July
-
[18]
A. Galanis, D. Stefankovic, E. Vigoda, and L. Yang. Ferromagnetic Potts model: Refined #-BIS-hardness and related results.SIAM Journal on Computing, 45(6):2004–2065, 2016. 15
work page 2004
- [19]
-
[20]
D. Gamarnik and D. Katz. Correlation decay and deterministic FPTAS for counting list-colorings of a graph. InProceedings of the eighteenth annual ACM-SIAM Symposium on Discrete Algorithms (SODA), pages 1245–1254, 2007. 14
work page 2007
-
[21]
Georgii.Gibbs measures and phase transitions, volume 9
H.-O. Georgii.Gibbs measures and phase transitions, volume 9. Walter de Gruyter, 2011. 13
work page 2011
-
[22]
A. Gerschenfeld and A. Montanari. Reconstruction for models on random graphs. In48th Annual IEEE Symposium on Foundations of Computer Science (FOCS’07), pages 194–204. IEEE, 2007. 15
work page 2007
-
[23]
H. Gl¨ ockner. Implicit functions from topological vector spaces to banach spaces.Israel Journal of Math- ematics, 155(1):205–252, 2006. 27
work page 2006
- [24]
-
[25]
S. Janson. Poisson approximation for large deviations.Random Structures & Algorithms, 1(2):221–229,
-
[26]
4, 8 52 NON-EXISTENCE PROBABILITIES AND LOWER TAILS IN THE CRITICAL REGIME
- [27]
-
[28]
S. Janson and L. Warnke. The lower tail: Poisson approximation revisited.Random Structures & Algo- rithms, 48(2):219–246, 2016. 4
work page 2016
-
[29]
M. Jenssen, W. Perkins, and A. Potukuchi. On the evolution of structure in triangle-free graphs.Advances in Mathematics, 480:110499, 2025. 2
work page 2025
-
[30]
M. Jenssen, W. Perkins, A. Potukuchi, and M. Simkin. Lower tails for triangles inside the critical window. arXiv preprint arXiv:2411.18563, 2024. 3, 4, 14, 44
-
[31]
M. Jenssen, W. Perkins, A. Potukuchi, and M. Simkin. Sampling, counting, and large deviations for triangle-free graphs near the critical density. InFoundations of Computer Science (FOCS 2024), 2024. 14
work page 2024
-
[32]
Z. Kelley and R. Meka. Strong bounds for 3-progressions. In2023 IEEE 64th Annual Symposium on Foundations of Computer Science (FOCS), pages 933–973. IEEE, 2023. 5
work page 2023
-
[33]
F. P. Kelly. Stochastic models of computer communication systems.Journal of the Royal Statistical Society. Series B (Methodological), pages 379–395, 1985. 13
work page 1985
-
[34]
D. J. Kleitman and K. J. Winston. On the number of graphs without 4-cycles.Discrete Mathematics, 41(2):167–172, 1982. 8
work page 1982
-
[35]
Y. Kohayakawa, T. Luczak, and V. R¨ odl. OnK 4-free subgraphs of random graphs.Combinatorica, 17(2):173–213, 1997. 4
work page 1997
-
[36]
D. Koller and N. Friedman.Probabilistic Graphical Models: Principles and Techniques. MIT Press, 2009. 12
work page 2009
-
[37]
G. Kozma and W. Samotij. Lower tails via relative entropy.The Annals of Probability, 51(2):665–698,
- [38]
-
[39]
L. Li, P. Lu, and Y. Yin. Correlation decay up to uniqueness in spin systems. InProceedings of the twenty-fourth annual ACM-SIAM symposium on Discrete algorithms, pages 67–84. SIAM, 2013. 14
work page 2013
- [40]
-
[41]
J. Liu, A. Sinclair, and P. Srivastava. Fisher zeros and correlation decay in the Ising model.Journal of Mathematical Physics, 60(10), 2019. 14
work page 2019
-
[42]
J. Liu, C. Wang, Y. Yin, and Y. Yu. Phase transitions via complex extensions of Markov chains. In Proceedings of the 57th Annual ACM Symposium on Theory of Computing, pages 903–914, 2025. 14
work page 2025
-
[43]
T. Luczak. On triangle-free random graphs.Random Structures & Algorithms, 16(3):260–276, 2000. 3, 4
work page 2000
-
[44]
W. Mantel. Problem 28.Wiskundige Opgaven, 10(60-61):320, 1907. 2
work page 1907
-
[45]
M. Mezard and A. Montanari.Information, physics, and computation. Oxford University Press, 2009. 10, 12
work page 2009
-
[46]
F. Mousset, A. Noever, K. Panagiotou, and W. Samotij. On the probability of nonexistence in binomial subsets.Annals of Probability, 48(1):493–525, 2020. 2, 8
work page 2020
- [47]
- [48]
-
[49]
D. Saxton and A. Thomason. Hypergraph containers.Inventiones mathematicae, 201(3):925–992, 2015. 3, 4, 5, 8, 15
work page 2015
-
[50]
S. Shao and Y. Sun. Contraction: A unified perspective of correlation decay and zero-freeness of 2-spin systems.Journal of Statistical Physics, 185:1–25, 2021. 14
work page 2021
-
[51]
A. Sinclair, P. Srivastava, and M. Thurley. Approximation algorithms for two-state anti-ferromagnetic spin systems on bounded degree graphs.Journal of Statistical Physics, 155(4):666–686, 2014. 14
work page 2014
-
[52]
A. Sly. Computational transition at the uniqueness threshold. InProceedings of the Fifty-first Annual IEEE Symposium on Foundations of Computer Science, FOCS 2010, pages 287–296. IEEE, 2010. 13
work page 2010
- [53]
- [54]
-
[55]
D. Stark and N. Wormald. The probability of non-existence of a subgraph in a moderately sparse random graph.Combinatorics, Probability and Computing, 27(4):672–715, 2018. 2
work page 2018
-
[56]
E. Szemer´ edi. On sets of integers containing no k elements in arithmetic progression.Acta Arith, 27(199- 245):2, 1975. 5
work page 1975
-
[57]
S. C. Tatikonda and M. I. Jordan. Loopy belief propagation and Gibbs measures. InProceedings of the Eighteenth conference on Uncertainty in artificial intelligence, pages 493–500, 2002. 12
work page 2002
-
[58]
D. Weitz. Counting independent sets up to the tree threshold. InProceedings of the Thirty-Eighth Annual ACM Symposium on Theory of Computing, STOC 2006, pages 140–149. ACM, 2006. 13
work page 2006
-
[59]
N. C. Wormald. The perturbation method and triangle-free random graphs.Random Structures & Algo- rithms, 9(1-2):253–270, 1996. 2
work page 1996
-
[60]
J. S. Yedidia, W. T. Freeman, Y. Weiss, et al. Understanding belief propagation and its generalizations. Exploring artificial intelligence in the new millennium, 8(236-239):0018–9448, 2003. 12
work page 2003
-
[61]
S. Zhang. Hypergraph independence polynomials with a zero close to the origin.Combinatorics, Proba- bility and Computing, pages 1–5, 2025. 14
work page 2025
-
[62]
Y. Zhao. On the lower tail variational problem for random graphs.Combinatorics, Probability and Com- puting, 26(2):301–320, 2017. 4 AppendixA.Proof of Lemma 8.4 Proof of Lemma 8.4.Recall that we define the operatorF=F (k) c,ζ on the set of bounded measurable functionsf: [0,1]→R >0 by F f(t) =cexp − ζ αk kX ℓ=1 Z w(t,ℓ) 0 Y i∈Iℓ f(t+is)ds , whereα k...
work page 2017
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.