REVIEW 3 major objections 4 minor 10 references
A note on the ultra log-concavity of matroid intersection
T0 review · 3 major / 4 minor · reviewed 2026-08-28 · deepseek-v4-flash
Pith's one-line read The paper proves that common independent set counts of any two matroids form an ultra log-concave sequence, and that this fails for three matroids.
desk verdict A genuinely new, cleanly proved extension of Mason's conjecture to intersections of two matroids, with a correct counterexample for three; worth a serious referee despite being a short note. 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 argument rests on Lorentzian polynomials, a class of polynomials whose coefficient sequences behave like those of stable polynomials. A recent theorem identifies $M^\natural$-concave functions exactly with those whose generating polynomial $Z_{\nu,q}$ is Lorentzian, and another theorem characterizes differential operators that preserve Lorentzianity. The paper combines these into a single Lorentzian polynomial $$R_{\nu_1,\nu_2,q}=\sum_{S_2\subseteq S_1\subseteq E} $q^{{-\nu_1(S_1)-\nu_2(S_2)}}$$x^{{S_1\setminus S_2}}$$y^{{|S_2|}}$$z^{{|E\setminus S_1|}}$,$$ obtained by applying a Lorentzianity-preserving differential operator to a Lorentzian polynomial. Restricting variables in $R$ produces the bivariate generating polynomials whose coefficients are the sequences in question, and two-variable Lorentzian polynomials have ultra log-concave coefficient sequences.
What would settle it
Enumerate all pairs of matroids on a small ground set (say $n=6$), compute $I_k$, the number of common independent sets of size $k$, and test the normalized inequality $I_k/\binom{n}{k}\cdot I_{k+2}/\binom{n}{k+2}\le(I_{k+1}/\binom{n}{k+1})^2$ for every $k$; a single violation would refute Corollary 1 and therefore the central theorem.
Extended reading notes
Core claim
The central claim is that the family of functions whose weighted subset counts are ultra log-concave is closed under sums: the sum of two $M^\natural$-concave functions, called an $M^\natural_2$-concave function, still produces an ultra log-concave sequence of weighted subset counts. In matroid terms, for two matroids $M_1,M_2$ on an $n$-element ground set, the numbers $I_k$ of common independent sets of size $k$ satisfy $I_k/\binom{n}{k}\cdot I_{k+2}/\binom{n}{k+2}\le(I_{k+1}/\binom{n}{k+1})^2$. The positive result stops at two matroids: the paper constructs three partition matroids where the common independent sets consist of all subsets of a three-element set plus $m$ singletons, giving $m+3$ one-element sets but only one three-element set, so log-concavity fails once $m\ge7$.
Load-bearing premise
The proof leans on two recent theorems about when a polynomial stays in a special 'Lorentzian' class under a certain operation; if either theorem has an unstated condition that fails here, the main result collapses.
Editorial extensions
If this is right
- For any two matroids on the same ground set, the number of common independent sets of size $k$ is ultra log-concave, hence log-concave and unimodal in $k$.
- The same conclusion holds for the counts $J_k$ of nested pairs $I_1\subseteq I_2$ with $|I_2\setminus I_1|=k$, and for the counts $H_k$ of disjoint independent-set pairs covering $k$ ground elements.
- The weighted statement for $M^\natural$-concave functions gives a full family of inequalities parametrized by $q\in(0,1]$, not just the unweighted matroid case.
- The three-partition-matroid example marks a sharp boundary: intersections of three matroids can violate even ordinary log-concavity.
Reading between the lines
- Beyond the paper's statements, ultra log-concavity of common independent set counts may make the uniform distribution over common independent sets satisfy sharp concentration bounds, since log-concave sequences have known extremal properties; the paper does not draw this consequence.
- The operator-based proof suggests a closure principle: whenever a class of discrete functions is governed by a Lorentzian generating polynomial, adding two members of the class may preserve the ultra log-concavity property through the same differential-operator mechanism; testing this on other $M$-convex-type classes could determine how far beyond matroids it reaches.
- A testable extension is whether a third matroid with a special structure, for example all three laminar, restores ultra log-concavity, or whether the $q$-weighted inequality has limits as $q$ approaches $0$ that recover known valuated-matroid inequalities.
Formalized claims in Lean
-
Claim #1: The central claim is that the family of functions whose weighted subset counts are ultra log-concave is closed under sums: the sum of two $M^\natural$-concave functions, called an $M^\natural_2$-concave function, still produces an ultra log-concave sequence of weighted subset counts. In matroid terms, for two matroids $M_1,M_2$ on an $n$-element ground set, the numbers $I_k$ of common independent
/-- @claim 1 The central claim is that the family of functions whose weighted subset counts are ultra log-concave is closed under sums: the sum of two $M^\natural$-concave functions, called an $M^\natural_2$-concave function, still produces an ultra log-concave sequence of weighted subset counts. In matroid terms, for two matroids $M_1,M_2$ on an $n$-element ground set, the numbers $I_k$ of common independent -/ def central_claim : Prop :=
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proves that for two M^natural-concave functions ν1, ν2 on a finite ground set E and any real q with 0<q≤1, the weighted counts I_{q,ν;k} of subsets of size k, where ν=ν1+ν2 and the weight of S is q^{-ν(S)}, form an ultra log-concave sequence. This generalizes Mason's conjecture to the intersection of two matroids (Corollary 1). A second main theorem proves the analogous ultra log-concavity for the weighted counts J_{ν1,ν2,q;k} of pairs S1⊆S2 with |S2\S1|=k, yielding further corollaries for pairs of matroids. The paper also supplies a counterexample with three partition matroids showing that the statement fails for intersections of three matroids.
Significance. If the cited theorems hold, the main result is a clean and meaningful extension of the recent proofs of Mason's conjecture from matroids and M^natural-concave functions to sums of two M^natural-concave functions, including matroid intersections. The proof is elegant: Theorem 5 constructs a Lorentzian polynomial by applying a Lorentzian-preserving differential operator to a product of Lorentzian polynomials, then extracts the desired sequences by restrictions. The counterexample for three partition matroids is simple, explicit, and convincing. The paper is transparent about relying on two recent preprints, and the internal derivation is logically consistent given those inputs. The main uncertainty is the status and exact hypotheses of the cited results, particularly Theorem 4 of [AMCD+26] at the endpoint q=1 and for functions taking -∞ values.
major comments (3)
- [Section 2.2.1 (Theorem 4) and Section 3 (Lemma 2, Theorem 5)] The paper states Theorem 1 for all 0<q≤1, but it never defines the value of q^{-ν(S)} when ν(S)=-∞. For q<1 the standard convention q^{-∞}=0 works, but for q=1 the expression 1^{-∞} is not defined in the manuscript. Since Lemma 2 applies Theorem 4 at q=1, the proof of Theorem 1 at the endpoint q=1 is not justified as written. Please either state an explicit convention (e.g., q^{-∞}=0 for all 0<q≤1) and confirm that the cited Theorem 4 of [AMCD+26] holds under this convention, or restrict Theorem 1 and the related statements to 0<q<1. Corollary 1 is unaffected because a matroid valuation can be used with any q<1.
- [Section 3, proof of Theorem 2] The proof says 'restrict R_{ν1,ν2,q} to x_i=x_j (for all i,j∈E) and y=z, resulting in Σ_{S1⊆S2⊆E} ...'. This is not what R_{ν1,ν2,q} gives: R is defined as a sum over S2⊆S1, so the restriction produces a sum over S2⊆S1 with weight q^{-ν1(S1)-ν2(S2)}. The definition of J in Theorem 2 instead uses S1⊆S2, placing ν1 on the smaller set. The theorem is true, but the proof must be corrected by applying Theorem 5 to the swapped pair (ν2,ν1) or by defining R with the opposite inclusion, then renaming the sets.
- [Section 1, Corollary 3] The derivation of Corollary 3 is ambiguous. Taking ν2(S) to be 0 if E\S∈M2 and -∞ otherwise pairs a set S with its complement E\S, so the union of the two independent sets is the entire ground set. If H_k counts pairs of disjoint independent sets covering k elements, the construction gives H_k = J_{ν1,ν2; n-k} via Theorem 2, not directly via the stated 'if instead' sentence. The manuscript should state which theorem is being used and spell out the index shift, or the wording of H_k must be clarified.
minor comments (4)
- [Section 1, Theorem 1 statement] The expression ν=ν1+ν2 should specify the convention for addition with -∞ (for example, (-∞)+a=-∞). This is standard in discrete convex analysis but should be stated for completeness.
- [Section 3, proof of Theorem 5] In the expansion after 'apply the operator P_{ν,q}(∂_x,∂_y)', the function should be denoted P_{ν2,q} to avoid ambiguity, since the operator is built from ν2.
- [Section 1, Example 1] The phrase '(see )' contains an empty cross-reference; either fill in the target or delete it.
- [Section 2.2, Theorem 3] The statement of Theorem 3 is telegraphic: the definition of the operator N and the role of κ are explained only tersely. A short clarifying sentence would help readers who are not specialists in Lorentzian operator theory.
Circularity Check
No significant circularity: Theorem 1 follows from external Lorentzian-preservation and M#-concavity theorems; no step reduces to its own conclusion.
full rationale
The derivation is self-contained given the cited theorems. Theorem 5 constructs R_{nu1,nu2,q} by applying the differential operator P_{nu2,q}(partial_x,partial_y) to y^{|E|} Z_{nu1,q}(x,z); Lorentzianity of the operator is established in Lemma 2 through the external operator-preservation criterion of Ross, Suss and Wannerer (Theorem 3), and Lorentzianity of the input is obtained from the external characterization of Ardila-Mantilla et al. (Theorem 4). No parameter is fitted, no quantity is defined in terms of the target inequality, and no result of the present paper is cited as its own justification. The new content, the M2#-concave extension, is not assumed anywhere in the proof. The cited theorems are from other author groups, so there is no load-bearing self-citation chain. The only nontrivial caveat is the paper's silence on the convention for q^{-nu(S)} when nu(S)=-infinity and q=1; that is a correctness or hypothesis-checking concern about the externally cited Theorem 4, not a circularity in the paper's own derivation. Since the central claim for matroid intersections can be read with q<1, the endpoint issue does not affect the main corollary. Thus no circular step is present.
Assumptions & free parameters
assumptions (4)
- standard math Lorentzian polynomials are closed under restriction to a variable set to zero, substitution of one variable for another, and multiplication (Proposition 2 from [BH20]).
- standard math Theorem 3 of Ross, Suss, and Wannerer [RSW23]: the differential operator P(d) preserves the Lorentzian property iff N(x^kappa P(x^{-1})) is Lorentzian.
- standard math Theorem 4 of Ardila-Mantilla et al. [AMCD+26]: nu is M^natural-concave iff Z_{nu,q} is Lorentzian for every q<=1.
- standard math Matroid valuation functions are M^natural-concave.
Cite this review
Pith. "Pith review of A note on the ultra log-concavity of matroid intersection." pith.science (2026). https://pith.science/paper/4UMDZXTU
@misc{pith2026260823262,
author = {Pith},
title = {Pith review of: A note on the ultra log-concavity of matroid intersection},
year = {2026},
howpublished = {\url{https://pith.science/paper/4UMDZXTU}},
note = {Machine review of arXiv:2608.23262}
}
abstract
In 1971, Mason conjectured that the numbers of independent sets of fixed size in a matroid constitute an ultra log-concave sequence. In 2020, this conjecture was proven by Br\"and\'en and Huh and independently by Anari, Liu, Gharan and Vinzant. Recently, this result was extended to $M^\natural$-concave functions. In this note, we make the next step by proving it for $M_2^\natural$-concave functions. This shows the same property for the intersection of any pair of matroids (which itself may not be a matroid). Furthermore, we show that this can not be further extended to the intersections of three matroids by including a counterexample of partition matroids.
Reference graph
Works this paper leans on
-
[1]
Dually Lorentzian polynomials , author=. arXiv preprint arXiv:2304.08399 , year=
- [2]
-
[3]
arXiv preprint arXiv:2601.02547 , year=
Tree metrics and log-concavity for matroids , author=. arXiv preprint arXiv:2601.02547 , year=
-
[4]
Mathematical Programming , year =
Murota, Kazuo , title =. Mathematical Programming , year =. doi:10.1007/BF02680565 , url =
-
[5]
2006 , publisher=
Matroid theory , author=. 2006 , publisher=
2006
-
[6]
Anari, Nima and Liu, Kuikui and oveis gharan, Shayan and Vinzant, Cynthia , year =. Log-concave polynomials III: Mason’s ultra-log-concavity conjecture for independent sets of matroids , journal =
-
[7]
Unimodal conjectures and Motzkin's theorem , author=. Combinatorial Math. Inst. , pages=. 1971 , publisher=
work page 1971
-
[8]
M-Convex Function on Generalized Polymatroid , urldate =
Kazuo Murota and Akiyoshi Shioura , journal =. M-Convex Function on Generalized Polymatroid , urldate =
Show all 10 references
-
[9]
arXiv preprint arXiv:2412.02051 , year=
Postnikov--Stanley polynomials are Lorentzian , author=. arXiv preprint arXiv:2412.02051 , year=
-
[10]
International Mathematics Research Notices , volume =
Hafner, Elena S and Mészáros, Karola and Vidinas, Alexander , title =. International Mathematics Research Notices , volume =. 2024 , month =. doi:10.1093/imrn/rnae058 , url =
2024 doi
Reviewed August 28, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.