Edge-connectivity and LLY curvature of hypergraphs

Qing Xia

REVIEW 2 major objections 4 minor 13 references

Edge-connectivity and LLY curvature of hypergraphs

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper proves that every locally finite connected simple r-uniform linear hypergraph with r≥3 and nonnegative Lin–Lu–Yau curvature has edge-connectivity equal to its minimum incidence degree, and shows by construction that dropping…

desk verdict New hypergraph rigidity result worth a round of revisions; Theorem 1.4 is misstated and the counterexample family has a repairable inequality bug. read the letter →

arxiv 2608.06029 v1 pith:I4QSSH2L submitted 2026-08-06 math.CO math.PR

Qing Xia This is my paper · ORCID

classification math.COmath.PR MSC 05C6505C4053C21

keywords Lin–Lu–YaucurvatureOllivier–Riccihypergraphedge-connectivitylinearTian–Zhaorandomwalk2-sectionWassersteindistance

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Ricci curvature bounds usually force global rigidity in discrete spaces. For graphs, nonnegative Lin–Lu–Yau curvature is known to make edge-connectivity equal the minimum degree. This paper carries that result over to hypergraphs: in every locally finite connected simple r-uniform linear hypergraph with r≥3, nonnegative Lin–Lu–Yau curvature implies λ(H)=δ(H). The key is that for such hypergraphs the Tian–Zhao random walk agrees with the lazy random walk on the 2-section, so curvature estimates transfer from graphs. The paper also constructs, for every r≥3 and t≥2, finite connected simple nonlinear r-uniform hypergraphs with positive curvature and λ(H)=δ(H)−t, showing linearity is essential.

What carries the argument

The machinery is the Lin–Lu–Yau curvature of hypergraphs defined through the Tian–Zhao random walk, which first chooses a hyperedge incident to a vertex and then chooses another vertex in that hyperedge. The load-bearing identity is (2.16): for r-uniform linear hypergraphs this walk coincides with the simple lazy random walk on the 2-section [H]_2, so κ^H_LLY(x,y)=$κ^{{[H]_2}}$_LLY(x,y). This lets the proof import the graph-theoretic curvature estimates of Liu–Xia. The other central tool is the combinatorial inequality Theorem 1.4, which bounds the smallest crossing neighborhood in a bipartite linear hypergraph and supplies the contradiction in the cut argument. Theorem 2.1, stating that the idleness function is linear on [1/2,1], reduces LLY curvature to κ_{1/2} and is used in the nonlinear construction.

What would settle it

One concrete way to settle Theorem 1.3 is to search for a locally finite connected simple r-uniform linear hypergraph with r≥3, nonnegative Lin–Lu–Yau curvature on every adjacent pair, and an edge cut with fewer than δ(H) hyperedges; any such example refutes the theorem. A more direct check is to compute both sides of (2.16) on a small linear hypergraph, since that identity is the bridge the proof depends on.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.3: a locally finite connected simple r-uniform linear hypergraph with r≥3 and nonnegative Lin–Lu–Yau curvature must satisfy edge-connectivity equals minimum incidence degree, λ(H)=δ(H). The proof follows the Liu–Xia cut strategy and rests on a hypergraph analogue of their combinatorial inequality: any bipartite linear r-uniform hypergraph that is not a bipartite star contains a crossing adjacent pair whose combined neighborhood is small. A separate star case is handled by a sharper estimate. Conversely, Theorem 1.6 shows the linearity hypothesis is not an artifact: for every r≥3 and t≥2 there is a finite connected simple nonlinear r-uniform hypergraph with positive curvature and λ(H)=δ(H)−t, so the gap can be prescribed arbitrarily.

Load-bearing premise

The whole argument for Theorem 1.3 rests on the identity that for uniform linear hypergraphs the Tian–Zhao random walk is exactly the lazy simple random walk on the 2-section, so hypergraph curvature equals graph curvature; if this identification failed, the transferred cut estimates would have no basis.

Editorial extensions

If this is right

For any such hypergraph, no edge cut can be smaller than the minimum incidence degree; the only way to disconnect is to cut at least δ(H) hyperedges.
The graph rigidity theorem of Liu–Xia now applies to an entire hypergraph class via the 2-section, so other graph curvature rigidity results may transfer to uniform linear hypergraphs.
In the nonlinear setting, positive curvature imposes no lower bound on edge-connectivity beyond the trivial one, and the gap δ(H)−λ(H) can be arbitrarily large.
The infinite nonlinear example with nonnegative curvature and gap one shows that even nonnegative curvature does not force equality outside the linear class.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural next step, not taken in the paper, is to test whether the theorem extends to linear hypergraphs with varying edge sizes when the edge sizes are bounded; the proof uses the uniform size r only through the degree scaling d^{[H]_2}_x=(r−1)d^H_x.
The nonlinear examples suggest that the curvature–connectivity link breaks down through large overlaps inside hyperedges; defining an overlap parameter and proving a curvature bound in terms of it could give a refined statement.
Because the linear case reduces exactly to 2-section curvature, any future sharpening of the Liu–Xia graph estimates would immediately sharpen Theorem 1.3.
A computational survey of small 3-uniform linear hypergraphs could test whether the bipartite-star case in the proof is essential or merely an artifact of the estimates.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

The paper has a real new result: Theorem 1.3 extends Chen–Liu–You / Liu–Xia rigidity to r-uniform linear hypergraphs (r≥3) with nonnegative LLY curvature, and the proof is not just a transcription because hyperedge cuts do not correspond to graph cuts under the 2-section. The combinatorial inequality in Theorem 1.4 is the load-bearing tool, and the transfer of graph curvature estimates via the identity (2.16) is plausible for linear uniform hypergraphs. The nonlinear counterexample family in Theorem 1.6 is also a genuine contribution: it shows the gap λ=δ−t can be prescribed under positive curvature.

That said, there are two technical problems you need to know about before trusting the statements.

First, Theorem 1.4 as printed is false. The statement puts the whole expression over 2: min |S| < ((m+1)(r−1)−n)/2 − 1. The proof, in both cases, establishes the weaker bound min |S| < (m+1)(r−1) − n/2 − 1. The counterexample is a single 4-uniform linear hyperedge on {x,z,y,w} with partition {x,z} | {y,w}. Then |E|=1, |V|=4, the stated RHS is 0 and min |S|=2. So the theorem needs restating. Fortunately Theorem 1.3's proof uses the weaker version (equation (4.5)), so the main rigidity argument can survive after the statement is corrected. But as it stands the advertised inequality is not what is proved.

Second, in the proof of Theorem 1.6, Case 2, the line "mu_x(u) ≥ 1/(2(n-1))" is false. From their definitions, the best uniform lower bound is roughly 1/[2((n−1)+(r−1)/L)], which is below 1/(2(n−1)). The argument only needs mu_x(u) > 1/(4n), and that does follow from (5.4) for the stated parameter range, so the curvature claim is repairable. Still, as written the inequality is wrong.

The other soft spot is the reliance on two self-cited preprints ([13] for the idleness theorem and [9] for the graph curvature estimates). They are not restated, so the hypergraph results inherit any errors there. That is normal practice but worth flagging for a referee.

Verdict: the core is solid and the ideas are right, but no refereed paper should carry a false theorem statement. Send to a serious referee after the author corrects Theorem 1.4 and the finite-case inequality. This deserves referee time.

Referee Report

2 major / 4 minor

Summary. The paper studies edge-connectivity versus Lin-Lu-Yau curvature for hypergraphs under the Tian-Zhao random walk. The main result, Theorem 1.3, asserts that every locally finite connected simple r-uniform linear hypergraph with r >= 3 and nonnegative Lin-Lu-Yau curvature satisfies lambda(H) = delta(H), extending the Liu-Xia graph rigidity theorem. The proof uses a reduction of the hypergraph walk to the simple random walk on the 2-section and a new combinatorial inequality for bipartite linear hypergraphs, stated as Theorem 1.4. A second main result, Theorem 1.6, constructs, for every r >= 3 and t >= 2, a finite connected simple nonlinear r-uniform hypergraph with positive Lin-Lu-Yau curvature and lambda(H) = delta(H) - t, showing that linearity is essential. The paper also includes an example of a nonlinear 3-uniform hypergraph with nonnegative curvature and lambda = delta - 1.

Significance. If Theorem 1.3 is correct, it is a natural and nontrivial extension of the Liu-Xia theorem to uniform linear hypergraphs, and Theorem 1.6 provides explicit, checkable constructions showing that the linearity assumption cannot be dropped. The cut-curvature framework and the explicit computation of curvature in the nonlinear examples are concrete and appropriate for the problem. However, the paper currently contains a false statement of the central combinatorial inequality (Theorem 1.4) and a false intermediate inequality in the proof of Theorem 1.6, so the results are not ready in their present form. The main rigidity theorem may still be correct after the statement of Theorem 1.4 is corrected to match the proof, and the flaw in the proof of Theorem 1.6 appears locally repairable, but both issues must be fixed before acceptance.

major comments (2)

[Section 1, Theorem 1.4] The displayed inequality (1.3) is false as stated. For H consisting of one linear 4-uniform hyperedge on {x,z,y,w} with partition X={x,z}, Y={y,w}, we have |E|=1, |V|=4, H is not a bipartite star with respect to this partition, and the right-hand side of (1.3) equals ((2)(3) - 4)/2 - 1 = 0. Yet every adjacent cross pair x,y has S^H_xy = {z,w}, so min |S^H_xy| = 2, contradicting (1.3). The proof in Section 3 does not establish (1.3): Case 1 on page 8 proves the bound |S^H_xy| < (m+1)(r-1) - n/2 - 1, and Case 2 proves the same weaker bound by contradiction with threshold (3.1). The correct theorem is the weaker bound, which is also the version used later in Eq. (4.5) in the proof of Theorem 1.3. The statement of Theorem 1.4 must be corrected, or the proof must be extended to the stated stronger bound.
[Section 5, proof of Theorem 1.6, Case 2] The proof contains a false inequality in the step after (5.10). The text claims that for u in A\{x}, mu_x(u) is at least L/[2(D+1)(r-1)] and then, using (5.4), that mu_x(u) >= 1/[2(n-1)]. This implication is incorrect. From (5.4), L/[2(D+1)(r-1)] = D/[2(n-1)(D+1)], which is strictly smaller than 1/[2(n-1)] whenever D > 0. For example, r=4, t=2 gives n=6, D=10, L=6, and for u in A\A_i the value mu_x(u)=6/(2*11*3)=1/11, which is less than 1/10. The needed conclusion mu_x(u) > mu_b(u) = 1/(4n) still follows from the direct inequality 2nD > (n-1)(D+1), so the gap is repairable, but the proof as written contains an invalid step in a load-bearing argument.

minor comments (4)

[Section 4, proof of Theorem 4.1] The final sentence of the proof says 'we obtain (4.1) and (4.1)'; this should be '(4.1) and (4.2)'.
[Section 5, proof of Theorem 1.6, Case 1] The symbol E_A is defined as a family of hyperedges but is then used as a vertex set in the phrases 'Both mu_x and mu_y are supported on EA' and 'any two distinct vertices of EA are adjacent'; the intended set is A union {b}.
[Introduction, Theorem 1.1] The attribution 'Chen-Liu-You [9, Theorem 1.1]' appears to cite reference [9], which is the Liu-Xia paper; the intended citation should be [3].
[Abstract] The word 'ypergraph' in the abstract should be 'hypergraph'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the hypergraph rigidity result is a genuine extension built on parameter-free prior graph and idleness lemmas, not on its own conclusion.

full rationale

The main derivation reduces hypergraph Lin-Lu-Yau curvature to graph curvature on the 2-section via Eq. (2.16): 'the hypergraph walk agrees exactly with the simple lazy random walk on [H]_2, and kappa^H_alpha(x,y)=kappa^{[H]_2}_alpha(x,y)'. It does not identify the hypergraph edge-connectivity lambda(H) with the graph edge-connectivity lambda([H]_2), so Theorem 1.3 is not a renaming of the graph theorem. The cut estimates are imported from Liu-Xia [9, Theorems 4.1 and 4.2] and the idleness linearity from Xia [13, Theorem 1.4 and Corollary 1.5]; these are parameter-free statements whose assumptions do not include the hypergraph rigidity conclusion, so under the review rules they are independent support rather than a circular self-citation chain. Theorem 1.4 is proved by a self-contained counting argument (with one appeal to the external graph lemma [9, Theorem 1.11]), and Section 5 checks the nonlinear construction directly. There is a separate correctness gap, not a circularity gap: the printed bound (1.3) has an extra outer division by 2, while the proof establishes the weaker inequality (m+1)(r-1)-|V|/2-1; the main proof of Theorem 1.3 uses the weaker proved version, so this concerns the statement of Theorem 1.4 rather than the circularity of the derivation.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No free parameters are fitted. The paper's contribution rests on standard optimal transport facts and on two prior articles by the same author(s). The central new inequality is proved directly, and the counterexample family is explicitly constructed rather than postulated.

assumptions (5)

domain assumption Theorem 2.1 from Xia [13]: the idleness function of an adjacent pair is concave and piecewise affine with at most three pieces and affine on [1/2,1], giving kappa_LLY(x,y)=2*kappa_{1/2}(x,y).
Used throughout to replace LLY curvature by 1/2-curvature and to equate signs; the proof is not reproduced and comes from a preprint by the present author.
domain assumption Liu-Xia [9, Theorems 4.1 and 4.2]: graph cut-curvature estimates, including the bipartite star estimate.
Theorem 4.1 transfers these to hypergraphs via the 2-section; they are load-bearing for the cut argument and come from a preprint co-authored by the present author.
domain assumption Liu-Xia [9, Theorem 1.11]: in a bipartite graph that is not a star, min over cross adjacent x,y of |S_xy| is at most |E|-|V|/2.
Used in Theorem 1.4, Case 1, to bound the cross-neighbor graph; again imported from the same prior preprint.
standard math Bourne-Cushing-Liu-Muench-Peyerimhoff [2, Lemma 4.1]: existence of an optimal transport plan with maximal diagonal entries.
Used to lower bound the Wasserstein distance in the curvature positivity check for Example 5.1.
standard math Circuit rank formula beta(G)=m-n+c(G) and nonnegativity of beta(B(H)) for connected H.
Used in Theorem 1.4 to bound n by m(r-1)+1 in the connected case.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Edge-connectivity and LLY curvature of hypergraphs." pith.science (2026). https://pith.science/paper/I4QSSH2L

@misc{pith2026260806029,
  author       = {Pith},
  title        = {Pith review of: Edge-connectivity and LLY curvature of hypergraphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I4QSSH2L}},
  note         = {Machine review of arXiv:2608.06029}
}

abstract

Chen, Liu, and You \cite{ChenLiuYou2025} proved that a locally finite connected graph with positive Lin--Lu--Yau curvature has edge-connectivity equal to its minimum degree. Liu and Xia \cite{LiuXia2026} subsequently showed that the same conclusion holds for every finite connected graph with nonnegative Lin--Lu--Yau curvature and classified all infinite exceptions. We investigate the corresponding problem for the random-walk curvature of hypergraphs introduced by Tian and Zhao \cite{TianZhao2025}. We formulate a hypergraph analogue of the combinatorial inequality used by Liu and Xia \cite{LiuXia2026} and use it to study edge cuts in uniform linear hypergraphs. Our first main result asserts that every locally finite connected $r$-uniform linear hypergraph, $r\ge 3$, with nonnegative Lin--Lu--Yau curvature has edge-connectivity equal to its minimum incidence degree. The linearity assumption is essential. In particular, for every $r\ge 3$ and every integer $t\ge 2$, we construct a finite connected simple nonlinear $r$-uniform ypergraph with positive Lin--Lu--Yau curvature such that its edge-connectivity is $t$ less than its minimum degree. Consequently, in the nonlinear setting the gap between minimum degree and edge-connectivity can be arbitrarily large even under strictly positive curvature.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 12 canonical work pages

[13]

Idleness Functions for Ollivier-Ricci Curvature on Hypergraphs

Q. Xia, Idleness functions for Ollivier–Ricci curvature on hypergraphs, arXiv:2608.01970, 2026. School of Mathematical Sciences, University of Science and Technology of China, 96 Jinzhai Road, Hefei 230026, Anhui Province, China Email address:xq0420@mail.ustc.edu.cn

work page Pith review arXiv 2026
[9]

Edge-connectivity and non-negative Lin-Lu-Yau curvature

S. Liu and Q. Xia, Edge-connectivity and non-negative Lin–Lu–Yau curvature, arXiv:2508.20950v2, 2026

work page Pith review arXiv 2026
[1]

Asoodeh, T

S. Asoodeh, T. Gao, and J. A. Evans, Curvature of hypergraphs via multi-marginal optimal transport, in2018 IEEE Conference on Decision and Control, IEEE, 2018, pp. 1180–1185

work page 2018
[2]

D. P. Bourne, D. Cushing, S. Liu, F. Muench, and N. Peyerimhoff, Ollivier–Ricci idleness functions of graphs,SIAM J. Discrete Math.32(2018), no. 2, 1408–1424

work page 2018
[3]

K. Chen, S. Liu, and Z. You, Connectivity versus Lin–Lu–Yau curvature,Int. Math. Res. Not. IMRN(2025), no. 19, Article rnaf303

work page 2025
[4]

Coupette, S

C. Coupette, S. Dalleiger, and B. Rieck, Ollivier–Ricci curvature for hypergraphs: a unified framework, inThe Eleventh International Conference on Learning Representa- tions, 2023

work page 2023
[5]

Diestel, Graph theory,Graph Theory, 5th ed., Springer, Berlin, 2017

R. Diestel, Graph theory,Graph Theory, 5th ed., Springer, Berlin, 2017

work page 2017
[6]

Eidi and J

M. Eidi and J. Jost, Ollivier Ricci curvature of directed hypergraphs,Sci. Rep.10 (2020), Article 12466

work page 2020

Show all 13 references

[7]

Ikeda, Y

M. Ikeda, Y. Kitabeppu, Y. Takai, and T. Uehara, Coarse Ricci curvature of hyper- graphs and its generalization,Theoret. Comput. Sci.930(2022), 1–23

2022
[8]

Y. Lin, L. Lu, and S.-T. Yau, Ricci curvature of graphs,Tohoku Math. J.63(2011), no. 4, 605–627

2011
[10]

Ollivier, Ricci curvature of Markov chains on metric spaces,J

Y. Ollivier, Ricci curvature of Markov chains on metric spaces,J. Funct. Anal.256 (2009), no. 3, 810–864

2009
[11]

Tian and L

Y. Tian and L. Zhao, Lin–Lu–Yau Ricci curvature on hypergraphs, arXiv:2507.04109, 2025

2025 arXiv
[12]

Villani,Topics in Optimal Transportation, Graduate Studies in Mathematics, vol

C. Villani,Topics in Optimal Transportation, Graduate Studies in Mathematics, vol. 58, American Mathematical Society, Providence, RI, 2003

2003

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [13]

Idleness Functions for Ollivier-Ricci Curvature on Hypergraphs

Q. Xia, Idleness functions for Ollivier–Ricci curvature on hypergraphs, arXiv:2608.01970, 2026. School of Mathematical Sciences, University of Science and Technology of China, 96 Jinzhai Road, Hefei 230026, Anhui Province, China Email address:xq0420@mail.ustc.edu.cn

work page Pith review arXiv 2026

[2] [9]

Edge-connectivity and non-negative Lin-Lu-Yau curvature

S. Liu and Q. Xia, Edge-connectivity and non-negative Lin–Lu–Yau curvature, arXiv:2508.20950v2, 2026

work page Pith review arXiv 2026

[3] [1]

Asoodeh, T

S. Asoodeh, T. Gao, and J. A. Evans, Curvature of hypergraphs via multi-marginal optimal transport, in2018 IEEE Conference on Decision and Control, IEEE, 2018, pp. 1180–1185

work page 2018

[4] [2]

D. P. Bourne, D. Cushing, S. Liu, F. Muench, and N. Peyerimhoff, Ollivier–Ricci idleness functions of graphs,SIAM J. Discrete Math.32(2018), no. 2, 1408–1424

work page 2018

[5] [3]

K. Chen, S. Liu, and Z. You, Connectivity versus Lin–Lu–Yau curvature,Int. Math. Res. Not. IMRN(2025), no. 19, Article rnaf303

work page 2025

[6] [4]

Coupette, S

C. Coupette, S. Dalleiger, and B. Rieck, Ollivier–Ricci curvature for hypergraphs: a unified framework, inThe Eleventh International Conference on Learning Representa- tions, 2023

work page 2023

[7] [5]

Diestel, Graph theory,Graph Theory, 5th ed., Springer, Berlin, 2017

R. Diestel, Graph theory,Graph Theory, 5th ed., Springer, Berlin, 2017

work page 2017

[8] [6]

Eidi and J

M. Eidi and J. Jost, Ollivier Ricci curvature of directed hypergraphs,Sci. Rep.10 (2020), Article 12466

work page 2020