REVIEW 3 major objections 5 minor 2 references
Further analysis on the second frequency of union-closed set families
T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash
Pith's one-line read This paper proves that a union-closed family whose second-most frequent element appears in at most one third of the sets, and which is not a near-2-cube, must have between 81 and 113 sets and all minimal 2-good sets of size 4.
desk verdict Real progress on Nagel's k=2 conjecture, but the proof's load-bearing LP bounds are asserted rather than verified. 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 linear program $L_0$ built from a minimal 2-good set $S$ is the carrying device. For each $T\subseteq S$, the variable $q_T$ counts the sets of $\mathcal{F}$ whose intersection with $S$ is exactly $T$. The constraints are $q_\emptyset\le 2$ (only $\emptyset$ and $\{1\}$ can miss $S$), $q_T\ge 1$ for every $T$ (from minimality), $\sum_{T\subseteq S}q_T=m$, and $\sum_{T\ni y}q_T\le m/3$ for each $y\in S$ (from the second-frequency bound). Solving $L_0$ gives $m\ge 45$ for $|S|=4$ and $m\ge 70.5$ for $|S|=5$. An element $y$ of $S$ is covered by an element $x$ if every witness set $F_y$ with $F_y\cap S=\{y\}$ contains $x$, and flexible if a witness can be chosen with or without $x$. The three lemmas about flexible and covered elements add inequalities to $L_0$, and each subcase is settled by solving the augmented program; the quoted optima are the outputs of these computations.
What would settle it
Exhibit a union-closed family $\mathcal{F}$ with $45\le |\mathcal{F}|\le 80$ that is not a near-2-cube and whose second-most frequent element appears in at most one third of the sets; such a family would disprove part (i) of Theorem 1.3. Alternatively, produce a feasible solution to the paper's stated linear program for $|S|=4$, $|C|\ge 3$ with $m<81$, which would contradict the claimed infeasibility.
Extended reading notes
Core claim
The central claim is Theorem 1.3: if $\mathcal{F}$ is union-closed, $|\bigcup\mathcal{F}|\ge 2$, $\mathcal{F}$ is not a near-2-cube, and $f_2(\mathcal{F})\le 1/3$, then $81\le |\mathcal{F}|\le 113$ and every minimal 2-good set of $\mathcal{F}$ has size $4$. The proof analyzes a minimal 2-good set $S$ and splits into the cases $|S|=4$ and $|S|=5$. It records, for each $T\subseteq S$, the number $q_T$ of sets in $\mathcal{F}$ meeting $S$ exactly in $T$, and it shows that 2-goodness and the frequency bound force certain linear inequalities on the $q_T$ and $m=|\mathcal{F}|$. Additional lemmas, using an element $x$ outside $S\cup\{1\}$ that can be chosen flexibly, add more inequalities depending on how many elements of $S$ are covered by $x$. Solving the resulting linear programs gives $m\ge 81$ in the surviving cases, $m\ge 114$ in others, and infeasibility when $|S|=4$ and at least three elements are covered; the case $|S|=5$ is ruled out entirely. The conclusion is that the open interval $[45,113]$ shrinks to $[81,113]$, and the minimal 2-good sets of any possible counterexample are uniform of size $4$.
Load-bearing premise
The numerical bounds and the infeasibility of some cases come from linear programs whose specifications and solutions are not derived in the paper; if those computations are wrong, the theorem's claimed range could change.
Editorial extensions
If this is right
- Every non-near-cube union-closed family with $45\le |\mathcal{F}|\le 80$ has $f_2(\mathcal{F})>1/3$, so the $k=2$ second-frequency conjecture holds throughout that lower range.
- Any potential non-near-cube counterexample must have between 81 and 113 sets and every minimal 2-good set of size 4; no such counterexample can have a minimal 2-good set of size 5.
- The search range for counterexamples is cut from $45$–$113$ down to $81$–$113$, and the remaining candidates must satisfy a strong uniformity condition.
- Combined with the earlier result, the $k=2$ second-frequency bound $f_2(\mathcal{F})\ge 1/3$ holds for all non-near-cube families with $|\mathcal{F}|\le 80$ or $|\mathcal{F}|\ge 114$.
Reading between the lines
- The theorem does not settle whether a family in $[81,113]$ with uniform minimal 2-good sets actually exists; settling that existence is the natural next step.
- Because the numeric bounds come from computer-assisted linear programming, a machine-checkable certificate for each linear program would make the conclusion verifiable without trusting an external script.
- The same $q_T$ linear-programming method may extend to higher $k$, where the threshold is $1/(2^{k-1}+1)$, to shrink the remaining intervals for the multielement frequency conjecture as well.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies Nagel's k=2 conjecture for union-closed families when the second frequency f2(F) is at most 1/3. Following the Das-Wu theorem, which leaves only the range 45 <= |F| <= 113 open, the author defines a linear program L0 whose variables count how many sets of F intersect a fixed minimal 2-good set S in each subset T of S. Three lemmas, based on choices of an x-flexible element and covered elements, add extra constraints to L0. Solving the strengthened LPs for |S|=4 and |S|=5 yields lower bounds on |F| and an infeasibility. The main theorem concludes that any non-near-2-cube family with f2(F) <= 1/3 must have 81 <= |F| <= 113 and every minimal 2-good set of size 4. The combinatorial lemmas are expository and mostly coherent, but the numerical bounds that eliminate all cases are asserted without derivations or certificates in the text.
Significance. If the theorem is correct, it is a useful finite reduction for the k=2 case of Nagel's conjecture: any potential counterexample outside the near-cube case would have size in [81,113] and all minimal 2-good sets of size 4. The LP-plus-lemmas strategy is natural and the three lemmas are, in the main, carefully argued. However, the load-bearing numerical claims are outsourced to a companion repository: every cell of Figure 1 and the m>=129 bound inside Lemma 2.2 are stated as outputs of solving LPs without presenting dual certificates, primal solutions, or solver transcripts. Since an error in any one of these LP computations would invalidate Theorem 1.3, the manuscript as submitted is not self-contained at exactly the point where correctness matters most.
major comments (3)
- [Figure 1 and Sections 4-5] The bounds m>=81, m>=114, m>=118.5, m>=115.5, m>=122, the infeasibility for s=4, |C|>=3, and the similar bound m>=129 used inside the proof of Lemma 2.2 are asserted as results of solving L0 with additional constraints, but no derivation, dual certificate, or solver output is included in the manuscript. The companion GitHub repository is cited, but the submitted text contains none of the actual LP solutions or certificates. Because Theorem 1.3 depends on these numerical bounds to eliminate every case, the proof is not verifiable as written. Please include, for each cell of Figure 1, an exact LP solution or a dual certificate, or append the complete solver input and output, and do the same for the m>=129 claim in Lemma 2.2.
- [Section 3, proof of Lemma 2.2] In the |S|=4 subcase, the proof dismisses the possibility that S+x-b-c is 2-good with the sentence 'then it means that there is a 2-good set with size 3, which implies f2(F)>1/3.' No proof or reference is given for this implication. In the |S|=5 subcase, the proof states that 'Consider the LP L0 along with qb+qc+qbc+28-3 <= m/3, we get m >= 129' with no derivation. Both statements are load-bearing: they are used to show that S+x-b-c is not 2-good for any b,c in C, which is an essential step in Lemma 2.2. Please supply a proof or citation for the size-3 claim and a certificate for the LP bound.
- [Section 3, Lemma 2.4 and its use in Sections 4.2 and 5.2] The proof of Lemma 2.4 assumes that S is a minimal 2-good set maximizing incidence among all minimal 2-good s-sets of F, and this assumption is then used in Corollaries 4.1 and 5.1. The assumption is introduced only in the heuristic section and is not stated in Theorem 1.3 or in the formal setup of Sections 4 and 5. The proof should state explicitly that, for each s in {4,5}, the argument fixes a minimal 2-good s-set S of maximum incidence, and that this choice entails no loss of generality because if any minimal 2-good s-set exists, a maximum-incidence one exists. As written, the reader cannot immediately see that the contradiction obtained for the chosen S rules out all minimal 2-good s-sets.
minor comments (5)
- [Throughout] The notation for the second frequency is inconsistent: the introduction uses F2(F), while Theorem 1.3 and the rest of the paper use f2(F). Please unify the notation.
- [Lemma 2.3 statement] The formula '(2s - 2s-1-|C|)' has lost its superscripts and should read '2^s - 2^{s-1-|C|}'. Similar superscript problems appear elsewhere in the typeset text.
- [Section 4, first line] The sentence 'Now we assume that S = {a,b,c,d} and x, ax in F S+x' is garbled and should be rewritten as a proper statement that x and a∪{x} are elements of the ground set or that certain sets are in F.
- [Section 3, proof of Lemma 3.1] The claims 'sum_{a in S} q_a >= 40 for |S|=5' and 'q_y >= 8 for each y in S, |S|=4' are stated as immediate consequences of the constraints of L0. A short derivation, or a reference to a prior lemma proving these LP bounds, would greatly improve readability.
- [Reproducibility] The companion GitHub repository is a useful supplement, but a URL is not a substitute for a permanent archival record. If the journal permits, please include the LP solver input/output or certificates as supplementary material, and consider archiving the repository in a stable venue.
Circularity Check
No significant circularity: the LP bounds are outputs of a well-defined LP whose constraints follow from f2(F) <= 1/3, and the cited companion theorem is an independent prior result.
full rationale
The derivation chain is not circular. The central reduction in Sections 4 and 5 is: from f2(F) <= 1/3 one derives linear inequalities on the quantities q_T (Lemma 2.2 through 2.4, together with the elementary inequalities q_∅ <= 2, q_T >= 1, and sum_{y in T} q_T <= m/3), and then solves the linear program L0 with these constraints. The target bounds (m >= 81, m >= 114, infeasibility, etc.) are outputs of the LP, not inputs; the family size m and the counts q_T are defined from the family independently of the theorem's conclusion. The use of Theorem 1.2 from the companion Das-Wu paper is a citation to an independent prior theorem whose stated assumptions (m <= 44 or m >= 114, or k >= 3) do not include the target statement; although one author overlaps, this is not an unverified self-citation chain that forces the result. The only notable weakness is that the numerical LP computations in Figure 1 and inside Lemma 2.2 are asserted rather than carried out in the text; that is a reproducibility/correctness concern, not circularity. I found no equation in which a fitted parameter is renamed as a prediction, and no definition that presupposes the conclusion.
Assumptions & free parameters
assumptions (6)
- domain assumption F is union-closed, contains the empty set, and 1 is the most abundant element.
- domain assumption The hypothesis f2(F) <= 1/3.
- domain assumption Das-Wu Theorem 1.2 is correct.
- domain assumption The LP solver returns exact optima for the finite programs L0 and its variants.
- ad hoc to paper S is a minimal 2-good set maximizing incidence among minimal 2-good s-sets.
- standard math Basic set algebra and inclusion-exclusion.
Cite this review
Pith. "Pith review of Further analysis on the second frequency of union-closed set families." pith.science (2026). https://pith.science/paper/ODXMZPJP
@misc{pith2026241203863,
author = {Pith},
title = {Pith review of: Further analysis on the second frequency of union-closed set families},
year = {2026},
howpublished = {\url{https://pith.science/paper/ODXMZPJP}},
note = {Machine review of arXiv:2412.03863}
}
abstract
The Union-Closed Sets Conjecture, also known as Frankl's conjecture, asks whether, for any union-closed set family $\mathcal{F}$ with $m$ sets, there is an element that lies in at least $\frac{1}{2}\cdot m$ sets in $\mathcal{F}$. In 2022, Nagel posed a stronger conjecture that within any union-closed family whose ground set size is at least $k$, there are always $k$ elements in the ground set that appear in at least $\frac{1}{2^{k-1}+1}$ proportion of the sets in the family. Das and Wu showed that this conjecture is true for $k\geq 3$ and $k=2$ if $|\mathcal{F}|$ is outside a particular range. In this companion paper, we analyse further when $\mathcal{F}$ fails Nagel's conjecture for $k=2$ via linear programming.
Figures
Reference graph
Works this paper leans on
Reviewed August 11, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.