Pith. sign in

REVIEW 1 major objections 5 minor 34 references

The Frankl--Tokushige product conjectures for $r$-cross-intersecting families

T0 review · 1 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper proves the Frankl–Tokushige product conjecture: for r-cross-intersecting uniform families with levels at most (r−1)n/r, the product of normalized sizes is at most the product of the levels over n, attained by a common 1-star.

desk verdict A serious, largely convincing resolution of both Frankl–Tokushige product conjectures; the main chain is sound, but two small gaps (unproved weaker-hypothesis claim in Remark 3.2 and the endpoint passage in Lemma 3.3) should be fixed before publication. read the letter →

arxiv 2607.21589 v1 pith:MZAKE57A submitted 2026-07-23 math.CO math.CAmath.FA

Fan Chang , Hong Liu , Miao Liu This is my paper
classification math.COmath.CAmath.FA MSC 05D05
keywords r-cross-intersectingfamiliesFrankl–Tokushigeproductconjectureuppershadowstarextremalconfigurationbiasedmeasureordered-partitioncouplingsettheorybound
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

This paper establishes the Frankl–Tokushige product conjecture for r-cross-intersecting uniform families. The statement is that whenever one selects r families—each family consisting of k_i-element subsets of an n-element set, with every choice of one set from each family sharing at least one element—the product of their normalized sizes is at most (k_1/n)⋯(k_r/n). The corresponding levels of a common one-point star attain this bound, so the star is the extremal shape over the full allowed range k_i≤(r−1)n/r. The same sharp bound is transferred to biased product measures on all subsets, for biases at most (r−1)/r. A sympathetic reader should care because this closes a conjecture that had only been settled in equal-level and special asymmetric cases, and it does so with a method that reduces the intersection condition to a single additive coupling bound.

What carries the argument

The key machinery is the ordered-partition coupling (Lemma 3.1), which turns r-cross-intersection into a family of additive inequalities over all level vectors summing to (r−1)n, together with a star-calibrated upper-shadow comparison (Lemma 3.4) whose induction step is a two-point inequality (Lemma 3.3). The comparison profile Φ_{s,t}(z) is a piecewise power function interpolating between z^{log s/log t} and its complement; the two-point inequality says that after restricting a family along one coordinate and recombining the two sections, the profile is preserved. This gives a smooth directed isoperimetric estimate normalized so that a 1-star is the extremal shape, and the final analytic st

What would settle it

Search all pairs 0≤x≤y≤1 for small adjacent slice parameters (e.g., n=5, k=2, ℓ=3, and n=4, k=1, ℓ=2) and test whether the two-point inequality (3) holds; a single violation would invalidate Lemma 3.3 and the induction. Independently, an exhaustive check over all r-cross-intersecting uniform families for small n (say n=6, r=3, k_i∈{0,1,2,3,4}) would test the product bound directly.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.1: for r≥2 and 0≤k_i≤(r−1)n/r, any r-cross-intersecting families F_i⊆binom([n],k_i) satisfy the product inequality ∏μ_{k_i}(F_i)≤∏k_i/n, with equality attained by the corresponding levels of a fixed 1-star. The argument has three linked parts. First, an ordered random partition of [n] shows that for any target levels ℓ_i with sum (r−1)n, the densities satisfy ∑μ_{ℓ_i}(F_i)≤r−1; this is the only place the cross-intersection hypothesis enters. Second, a star-calibrated upper-shadow comparison, proved by induction through a two-point inequality, controls the density of a family raised to a target level relative to the star density. Third, an analytic theorem for o

Load-bearing premise

The load-bearing premise is the two-point comparison inequality (Lemma 3.3) used in the shadow induction; if it failed for any adjacent slice parameters, including the endpoint cases k=1 and ℓ=n−1, the star-calibrated upper-shadow bound and the analytic hypothesis of the final theorem would no longer be available.

Editorial extensions

If this is right

  • The common 1-star is extremal over the full uniform range k_i≤(r−1)n/r; no mixed construction can beat the product of the star's levels.
  • The biased version holds: for any r-cross-intersecting families of all subsets with p_i≤(r−1)/r, the product of their p_i-biased measures is at most p_1⋯p_r.
  • Both parameter ranges are best possible—crossing the threshold (r−1)n/r or (r−1)/r makes the full level or large-subset family violate the inequality.
  • The proof reproduces the earlier equal-level Frankl–Tokushige theorem as a special case, using only the additive coupling.
  • The coupling extends formally to r-cross t-intersecting families for t≥2, with a missing shadow-comparison step identified as the open part.

Reading between the lines

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

  • Editorial extension: the proof's reduction suggests a transfer principle—any families satisfying the level-sum constraints, even without an intersection condition, should obey the same product bound; one could test this by constructing non-intersecting families that satisfy the constraints.
  • Editorial extension: a stability theorem should hold near the star: families whose normalized product is close to (k_1⋯k_r)/n^r should be close to a common 1-star in normalized symmetric difference, since the shadow comparison is strict away from the star.
  • Editorial extension: the ordered-prefix pivot used to choose target levels might be adapted to other unbalanced product problems in extremal set theory, where unequal parameters block the simple AM–GM step.
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.

Referee Report

1 major / 5 minor

Summary. The paper proves the Frankl–Tokushige product conjectures for r-cross-intersecting families. Theorem 1.1 states that for r ≥ 2, 0 ≤ k_i ≤ (r−1)n/r, and r-cross-intersecting families F_i ⊆ C([n], k_i), one has ∏ μ_{k_i}(F_i) ≤ ∏ k_i/n, with equality attained by the corresponding levels of a common 1-star. Corollary 1.2 transfers the statement to p_i-biased product measure for p_i ≤ (r−1)/r. The proof has three main ingredients: an ordered-partition coupling giving sharp additive inequalities at every critical level vector (Lemma 3.1); a star-calibrated upper-shadow comparison (Lemma 3.4) proved by induction on n, with the induction step reduced to a two-point inequality (Lemma 3.3); and an analytic optimization theorem (Theorem 4.1) that converts the asymmetric additive constraint into the desired product bound. The paper also recovers the equal-level theorem of Frankl–Tokushige and discusses extensions to r-cross t-intersection.

Significance. If the proof is completed, this settles two long-standing conjectures in full generality, and the method is attractive: Lemma 3.1 isolates essentially all combinatorial content in a sharp linear constraint, the upper-shadow comparison is a slice analogue of Bellman-function arguments, and the analytic theorem is a nontrivial replacement for the symmetric AM–GM step. The equality cases and parameter ranges are discussed carefully, and the biased theorem follows from a clean lifting argument. The result would be a major contribution to extremal set theory. However, the proof has a load-bearing gap in the endpoint cases of the two-point inequality, as detailed below; once that is repaired, the paper should be acceptable.

major comments (1)
  1. [§5, Lemma 3.3; used in §3.4, Eq. (5) and §4, Eq. (7)] The proof of Lemma 3.3 begins by saying that the cases k=1 and ℓ=n−1 are 'the pointwise limits in (2)' and that the inequalities below pass to those limits. This is not a proof. The endpoint profiles Φ_{0,t} and Φ_{s,1} are discontinuous at z=1, and the arguments inside the Φ terms in (3) can equal 1 in the boundary cases, so pointwise convergence alone does not automatically preserve the inequality. These endpoint cases are not decorative: in the induction in Lemma 3.4, a pivot family may have k_j=1, and the target level may be qn=n−1, so Lemma 3.3 is invoked exactly at k=1 or ℓ=n−1. If (3) fails at any boundary point, Lemma 3.4(5), and hence the pivot estimate (7) used in Theorem 1.1, is unproved. Please supply a direct verification of the two endpoint cases, or state and prove Lemma 3.3 for real k,ℓ with a rigorous limit argument that handles the discontinuities at z=1 explicitly.
minor comments (5)
  1. [§3.2, Remark 3.2] The 'more importantly' statement that Theorem 1.1 remains valid under a weaker hypothesis (the critical-sum constraint replacing cross-intersection) is asserted without proof. It is not used later, but as written it is a substantive claim. Either provide a proof or explicitly label it as a conjecture/expected consequence.
  2. [§5, proof of Lemma 3.3, Case 3] Several algebraic identities are stated as 'direct substitution' without derivation, e.g. the two displayed identities following the Case 3 split and the penultimate equality in Case 2. They are verifiable, but a line or two of explanation would help the reader and reduce the risk of a hidden sign error.
  3. [§4, proof of Theorem 4.1] The definition p = 1−q + (j−2)/r is easy to misread as 1−q + j − 2/r. Use a displayed equation with parentheses: p = 1 − q + (j−2)/r.
  4. [§1, first paragraph] Typo: 'level vectors level vectors' is duplicated. Also, the phrase 'the exact bound is attained by the corresponding levels of a common 1-star' is repeated in the abstract and introduction; this is fine but should be checked for redundancy.
  5. [§2, Eq. (2)] The endpoint profiles Φ_{0,t} and Φ_{s,1} are defined by pointwise limits but the limits are not computed. Since these profiles are used in the shadow induction, a short explanation of the convergence and the interpretation as empty/full sections would be helpful.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the derivation is self-contained, with only an unproven endpoint-limit passage flagged as a correctness concern, not a circular step.

full rationale

The paper's derivation chain is self-contained and does not reduce to its inputs. Lemma 3.1 derives the additive constraint sum μ_ℓ_i(F_i) ≤ r−1 directly from a random ordered partition coupling; the cross-intersection hypothesis is used there and nowhere else, and the lemma is proved in the text. The upper-shadow comparison (Lemma 3.4) is proved by induction using the explicit comparison profile Φ_{s,t} defined in (1), together with the two-point inequality Lemma 3.3. The profile is introduced by definition in the paper and is not imported from the authors' prior work as an unverified ansatz; the self-citation [7] appears only in Remark 3.2 as motivation and is not load-bearing. Theorem 4.1, which turns the additive constraints into the product bound, is a separate analytic theorem proved in Section 6; its hypothesis is not merely a restatement of the conclusion. Corollary 1.2 follows from Theorem 1.1 by the standard Dinur–Safra/Frankl–Tokushige lifting argument, so there is no fitted parameter being renamed as a prediction. The only flagged issue is not circularity: the proof of Lemma 3.3 states, 'The cases k = 1 and ℓ = n−1 are the pointwise limits in (2); the inequalities below pass to those limits,' and no ε-δ or monotone-convergence argument is supplied for that limit passage. This is an omitted rigor step that could affect Lemma 3.4 and hence Theorem 1.1 if the endpoint convention failed, but it is not a reduction of the theorem to its own assumptions or to a self-citation. Accordingly, the circularity score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central proof is self-contained. No parameters are fitted to data; the only external inputs are standard analytic facts. The paper's previous equal-bias result [7] is cited only for context, not used as a black box.

assumptions (3)
  • standard math Elementary binomial identities and the uniform normalization of slice densities (Lemma 2.1, double-counting argument).
    Used throughout to relate μ_k(A), section densities, and level sizes; standard double counting.
  • standard math Standard one-variable calculus facts: monotone l'Hôpital's rule, convexity, and monotonicity of comparison functions in Section 6.
    Invoked in Lemma 6.1 and in the proof of Theorem 4.1; these are ordinary analytic tools, not new assumptions.
  • domain assumption Upper shadows preserve r-cross-intersection.
    Used without comment in the proof of Theorem 1.1 when F_j is replaced by its upper shadow; elementary because each shadow set contains a member of the original family.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Frankl--Tokushige product conjectures for $r$-cross-intersecting families." pith.science (2026). https://pith.science/paper/MZAKE57A

@misc{pith2026260721589,
  author       = {Pith},
  title        = {Pith review of: The Frankl--Tokushige product conjectures for $r$-cross-intersecting families},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MZAKE57A}},
  note         = {Machine review of arXiv:2607.21589}
}
abstract

We settle the uniform and biased product conjectures of Frankl and Tokushige for $r$-cross-intersecting families. Let $r\geq2$, let $0\leq k_i\leq(r-1)n/r$, and let $\mathcal{F}_i\subseteq\binom{[n]}{k_i}$ be $r$-cross-intersecting. We prove the sharp inequality $$\prod_{i=1}^r\frac{|\mathcal{F}_i|}{\binom{n}{k_i}}\leq \prod_{i=1}^r\frac{k_i}{n},$$ with equality attained by the corresponding levels of a common $1$-star. As a consequence, we obtain the analogous $p_i$-biased measure theorem for $0\leq p_i\leq(r-1)/r$, $$ \prod_{i=1}^r\mu_{p_i}(\mathcal{F}_i)\leq \prod_{i=1}^r p_i.$$The main difficulty is that unequal parameters do not determine a single common target level; instead, the target levels $\ell_1,\ldots,\ell_r$ must satisfy $\sum_{i=1}^r \ell_i=(r-1)n$. We overcome this asymmetry in three steps. An ordered-partition coupling gives a sharp additive inequality for every such choice of target levels. A star-calibrated upper-shadow inequality relates the density of a family on its original level to the density of its upper shadow on a suitably chosen target level; it is proved by induction on $n$, with the induction step reduced to a two-point inequality. Finally, an analytic inequality shows that the resulting asymmetric additive estimate implies the required product bound. Perhaps surprisingly, the coupling captures all the combinatorial information of cross-intersection, reducing the remainder of the proof to an analytic argument.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 4 linked inside Pith

  1. [7]

    Chang, H

    F. Chang, H. Liu, and M. Liu. A sharpp-biased product bound forr-cross-intersecting families.arXiv preprint arXiv:2606.15618, 2026

    arXiv 2026

  2. [20]

    Frankl and N

    P. Frankl and N. Tokushige. Onr-cross intersecting families of sets.Combin. Probab. Comput., 20(5):749–752, 2011

    2011

  3. [1]

    Ahlswede and L

    R. Ahlswede and L. H. Khachatrian. The complete intersection theorem for systems of finite sets.European J. Combin., 18(2):125–136, 1997

    1997

  4. [2]

    Beltran, P

    D. Beltran, P. Ivanisvili, and J. Madrid. On sharp isoperimetric inequalities on the hypercube. arXiv preprint arXiv:2303.06738, 2023. 19

    arXiv 2023

  5. [3]

    Beltran, P

    D. Beltran, P. Ivanisvili, J. Madrid, and L. Patil. Optimal Young’s convolutions inequality and its reverse form on the hypercube.arXiv preprint arXiv:2507.06115, 2025

    arXiv 2025

  6. [4]

    Berger and Y

    A. Berger and Y. Zhao.K4-intersecting families of graphs.J. Combin. Theory Ser. B, 163:112–132, 2023

    2023

  7. [5]

    S. G. Bobkov and F. Götze. Discrete isoperimetric and Poincaré-type inequalities.Probab. Theory Related Fields, 114(2):245–277, 1999

    1999

  8. [6]

    P. Borg. The maximum product of weights of cross-intersecting families.J. Lond. Math. Soc. (2), 94(3):993–1018, 2016

    2016

Show all 34 references
  1. [8]

    Dinur and S

    I. Dinur and S. Safra. On the hardness of approximating minimum vertex cover.Ann. of Math. (2), 162(1):439–485, 2005

    2005

  2. [9]

    Domelevo, P

    K. Domelevo, P. Ivanisvili, S. Petermichl, and A. Volberg. Dimension-free bounds for Riesz transforms on the Hamming cube via a Bellman function.arXiv preprint arXiv:2606.20289, 2026

    2026 arXiv

  3. [10]

    Durcik, P

    P. Durcik, P. Ivanisvili, and J. Roos. Sharp isoperimetric inequalities on the Hamming cube near the critical exponent.arXiv preprint arXiv:2407.12674, 2024

    2024 arXiv

  4. [11]

    Durcik, P

    P. Durcik, P. Ivanisvili, J. Roos, and X. Xie. Sharp isoperimetric inequalities on the Hamming cube II: The critical exponent.arXiv preprint arXiv:2602.20462, 2026

    2026

  5. [12]

    Ellis, Y

    D. Ellis, Y. Filmus, and E. Friedgut. Triangle-intersecting families of graphs.J. Eur. Math. Soc. (JEMS), 14(3):841–885, 2012

    2012

  6. [13]

    Ellis, E

    D. Ellis, E. Friedgut, and H. Pilpel. Intersecting families of permutations.J. Amer. Math. Soc., 24(3):649–682, 2011

    2011

  7. [14]

    Ellis, N

    D. Ellis, N. Keller, and N. Lifshitz. Stability versions of Erdős–Ko–Rado type theorems via isoperimetry.J. Eur. Math. Soc. (JEMS), 21(12):3857–3902, 2019

    2019

  8. [15]

    Ellis, G

    D. Ellis, G. Kindler, and N. Lifshitz. Forbidden intersection problems for families of linear maps.Discrete Anal., pages Paper No. 19, 32, 2023

    2023

  9. [16]

    Erdős, C

    P. Erdős, C. Ko, and R. Rado. Intersection theorems for systems of finite sets.Quart. J. Math. Oxford Ser. (2), 12:313–320, 1961

    1961

  10. [17]

    Frankl and M

    P. Frankl and M. Deza. On the maximum number of permutations with given maximal or minimal distance.J. Combinatorial Theory Ser. A, 22(3):352–360, 1977

    1977

  11. [18]

    Frankl and Z

    P. Frankl and Z. Füredi. Beyond the Erdős-Ko-Rado theorem.J. Combin. Theory Ser. A, 56(2):182–194, 1991

    1991

  12. [19]

    Frankl and N

    P. Frankl and N. Tokushige. Weighted multiply intersecting families.Studia Sci. Math. Hungar., 40(3):287–291, 2003

    2003

  13. [21]

    Frankl and N

    P. Frankl and N. Tokushige. Invitation to intersection problems for finite sets.J. Combin. Theory Ser. A, 144:157–211, 2016. 20

    2016

  14. [22]

    Frankl and N

    P. Frankl and N. Tokushige.Extremal problems for finite sets, volume 86 ofStudent Mathematical Library. American Mathematical Society, Providence, RI, 2018

    2018

  15. [23]

    Friedgut

    E. Friedgut. On the measure of intersecting families, uniqueness and stability.Combinatorica, 28(5):503–528, 2008

    2008

  16. [24]

    Gupta, Y

    P. Gupta, Y. Mogge, S. Piga, and B. Schülke.r-cross t-intersecting families via necessary intersection points.Bull. Lond. Math. Soc., 55(3):1447–1458, 2023

    2023

  17. [25]

    G. Katona. A theorem of finite sets. InTheory of Graphs (Proc. Colloq., Tihany, 1966), pages 187–207. Academic Press, New York-London, 1968

    1966

  18. [26]

    Keller, N

    N. Keller, N. Lifshitz, D. Minzer, and O. Sheinfeld. Ont-intersecting families of permutations. Adv. Math., 445:Paper No. 109650, 28, 2024

    2024

  19. [27]

    J. B. Kruskal. The number of simplices in a complex. InMathematical optimization techniques, pages 251–278. Univ. California Press, Berkeley-Los Angeles, Calif., 1963

    1963

  20. [28]

    Kupavskii and D

    A. Kupavskii and D. Zakharov. Spread approximations for forbidden intersections problems. Adv. Math., 445:Paper No. 109653, 29, 2024

    2024

  21. [29]

    Matsumoto and N

    M. Matsumoto and N. Tokushige. The exact bound in the Erdős-Ko-Rado theorem for cross-intersecting families.J. Combin. Theory Ser. A, 52(1):90–97, 1989

    1989

  22. [30]

    S. Suda, H. Tanaka, and N. Tokushige. A semidefinite programming approach to a cross- intersection problem with measures.Math. Program., 166(1-2, Ser. A):113–130, 2017

    2017

  23. [31]

    Talagrand

    M. Talagrand. Isoperimetry, logarithmic Sobolev inequalities on the discrete cube, and Margulis’ graph connectivity theorem.Geom. Funct. Anal., 3(3):295–314, 1993

    1993

  24. [32]

    Tokushige

    N. Tokushige. Intersecting families—uniform versus weighted.Ryukyu Math. J., 18:89–103, 2005

    2005

  25. [33]

    Tokushige

    N. Tokushige. The maximum measure of non-trivial 3-wise intersecting families.Math. Program., 204(1-2, Ser. A):643–676, 2024

    2024

  26. [34]

    R. M. Wilson. The exact bound in the Erdős-Ko-Rado theorem.Combinatorica, 4(2-3):247– 257, 1984. 21

    1984

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.