{"id":"da523481-6c09-43fc-b073-4941b62cdc8f","arxiv_id":"2504.17567","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The inverse Kazhdan-Lusztig polynomials of paving matroids are log-concave, proved via a new real-rootedness result for a Hadamard product with (1+t)^n.","lead":"This paper proves that the inverse Kazhdan-Lusztig polynomial of any paving matroid has log-concave coefficients, confirming a 2021 conjecture for a large class of matroids. The proof works by showing a related polynomial has only real roots, a stronger property that implies log-concavity.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.3 rests on unverified discriminant inequalities; Lemmas 3.4 and 3.5 silently omit ranks d=3..6.","rationale":"Read as a proof, the paper's structural strategy is sound: the n-sequence argument in Theorem 1.3 correctly handles the non-linearity of B, and the interlacing framework is standard. The weak point is not the framework but the computational certificates. The reader already flags the Mathematica checks; I agree they are load-bearing. I add that the text's own range declarations (n≥3 in all four cells) leave d=3..6 untreated in Lemmas 3.4 and 3.5, which are needed for every rank. This is an internal incompleteness, not merely a reproducibility preference. It is likely fixable—for n=1 the polynomials are linear and for n=2 the discriminant is an explicit quadratic in k—but as written the proof is conditional. Hence I would keep the reader's CONDITIONAL verdict: accept only after the computations are made reproducible and extended to the small-rank cases.","tokens_in":11287,"tokens_out":27749,"duration_ms":251785,"concrete_test":"Re-execute the four Resolve computations from Lemmas 3.3–3.5 with full integer declarations (Element[n|m|d, Integers], n≥1) and add the d=3,4,5,6 cases for Lemmas 3.4 and 3.5; independently compute the discriminants of p1–p5 with CylindricalDecomposition in Mathematica or an exact CAD in Maple/Sage. If every call returns True and the small-d cases are verified (or proved by the elementary degree-1/2 argument), the concern is settled; a single False or a missing small-d counterexample would invalidate the main theorem.","verdict_should_be":"UNCHANGED","load_bearing_attack":"To get Theorem 1.3 the authors need the interlacing chain in Lemma 3.2. That chain uses Lemmas 3.4 and 3.5 for every d≥3, but both proofs reduce to Resolve[ForAll[...]] cells (In[6], In[9], In[12]) whose stated hypotheses are only n≥3, i.e. d≥7. No argument is given for d=3,4,5,6 in these lemmas; Lemma 3.3's small-d check is for a different pair and does not cover the differences appearing in Lemmas 3.4 and 3.5. In addition, the printed Mathematica inputs are partial pseudocode (e.g. n∑_{i=0}), with no explicit integer domain for n, m, d and no shipped notebook or certificate, so the claimed True outputs cannot be audited from the text. Since the only proof that these cubic and quadratic discriminants stay positive is the unverified computation, the real-rootedness of B(Q_M) is conditional on exactly those computations. If any omitted d or any n, m, d, k tuple gives a non-positive discriminant, the chain f0≪fm and fi−1≪fi breaks and Theorem 1.3 does not follow.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that the inverse Kazhdan-Lusztig polynomial of any paving matroid has log-concave coefficients with no internal zeros (Theorem 1.2). The proof goes through the stronger statement (Theorem 1.3) that the Hadamard product B(Q_M(t)) of Q_M(t) with (1+t)^n, where n is the degree of Q_M, is real-rooted. Using the explicit formula for uniform matroids and the Ferroni--Nasr--Vecchi decomposition of paving matroids into stressed hyperplanes, the authors reduce real-rootedness to interlacing statements for the polynomials B(Q_{U_{m,d}}) and B(Q_{U_{h,d}}-Q_{U_{h,d-1}}). The interlacing is established through Hermite-Kakeya-Obreschkoff and multiplier-sequence arguments, with a final n-sequence transfer that handles the nonlinearity of B on polynomials of different degrees. The central interlacing lemmas depend on three Mathematica Resolve[ForAll] computations asserting positivity of certain discriminants.","tokens_in":11626,"tokens_out":16006,"duration_ms":147091,"significance":"If the computational ingredients are made fully rigorous and reproducible, this is a substantial result: it confirms the Gao--Xie log-concavity conjecture for the broad class of paving matroids and introduces a plausible real-rootedness strengthening (Conjecture 1.4) with a clean proof strategy. The use of n-sequences to handle the degree dependence of the Hadamard operator is elegant, and the explicit Wronskian sign computations in Lemmas 3.3--3.5 are checkable and appear correct. The nonnegativity of coefficients from Braden--Huh--Matherne--Proudfoot--Wang is used correctly, and the proof of Theorem 1.3 is not circular with respect to Conjecture 1.4. The main obstacles are verification gaps in the computational lemmas rather than flaws in the overall architecture.","major_comments":[{"comment":"The proofs of Lemmas 3.3--3.5 rest entirely on the Mathematica commands In[1] through In[12], whose printed forms are not reproducible. As printed, In[1] contains an expression like \"n∑_{i=0}\" rather than a valid Sum command, and none of the Resolve[ForAll[...]] calls declares n, m, h, d as integers, even though the factorizations (t+1)^{n-3} and the binomial sums are only meaningful for integer n. No notebook, code file, or independent certificate (e.g., a Sturm sequence or an explicit discriminant factorization) is supplied. Since the positivity of the discriminants of p1, p3, p4, and p5 is the only support for the interlacing conclusions, the proof of Theorem 1.3 is conditional on unverifiable computation. The authors should provide executable code with explicit integer domains, or replace each Resolve output by a checkable analytic argument.","section":"Lemmas 3.3--3.5"},{"comment":"Both Lemma 3.4 and Lemma 3.5 are stated for all d ≥ 3, but the Resolve commands only cover the range n ≥ 3, which corresponds to d ≥ 7. Lemma 3.3 at least mentions a direct Mathematica verification for d = 3, 4, 5, 6, but Lemmas 3.4 and 3.5 contain no such discussion. Thus the interlacing statements for d ≤ 6 are unproved in the text. Since Lemma 3.2 and hence Theorem 1.3 require the chain f0 ≪ fm and f_{i-1} ≪ f_i for every d ≥ 3, these omitted cases are load-bearing and must be supplied, either by explicit finite checks or by an argument showing they are automatically covered by the same Wronskian sign and real-rootedness of the relevant linear and quadratic polynomials.","section":"Lemmas 3.4 and 3.5"}],"minor_comments":[{"comment":"The sentence \"It follows from Lemmas 3.2, 3.3, and 3.4\" should read \"It follows from Lemmas 3.3, 3.4, and 3.5.\"","section":"Proof of Lemma 3.2"},{"comment":"In the display after the formula for B(Q_{U_{m,d}}(t)) + k B(Q_{U_{m,d}}(t) - Q_{U_{m,d-1}}(t)), the symbol p3 should be p5.","section":"Lemma 3.5"},{"comment":"The notation \"Γ = {γ_i}^n_{k=0}\" has mismatched indices; it should be \"Γ = {γ_i}_{i=0}^n.\"","section":"Definition 2.2"},{"comment":"The step \"By Theorem 2.5, we conclude that the coefficients of Q_M(t) form a floor((d-1)/2)-sequence\" is very terse. It would help to spell out that the degree-n Hadamard product of Q_M has only real roots, hence the coefficient sequence of Q_M is an n-sequence, and then applying this n-sequence to (1+t)^{deg Q_M} yields the desired B(Q_M(t)).","section":"Theorem 1.3, proof"},{"comment":"The proof's claim that a linear combination af(t)+bg(t) of two degree-n real-rooted polynomials has degree n or n-1 is false; for example, f(t) = (t+1)^2 and g(t) = t^2 + 2t are both real-rooted with nonnegative coefficients, but f - g = 1 has degree 0. The lower bound on deg B(Q_M(t)) is therefore not established by the given argument.","section":"Proposition 4.3"}],"recommendation":"major_revision","confidential_remarks":"The overall proof strategy is sound and the paper addresses a significant open conjecture, but the central interlacing lemmas currently rest on opaque computer algebra with no reproducible code and with genuinely missing small-degree cases. These issues are fixable: the authors should provide a notebook or certificates and complete the d = 3,4,5,6 checks. I do not see circularity or misuse of Conjecture 1.4; the main concern is verification, not mathematical architecture."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good to see a substantial step on the Gao-Xie conjecture. The paper proves log-concavity of inverse Kazhdan-Lusztig polynomials for all paving matroids, and does it via a stronger statement: the Hadamard product B(Q_M) is real-rooted. That is a genuinely new conjecture (1.4) and a sensible strengthening, since the raw Q_M is not always real-rooted. The machinery — interlacing, multiplier sequences, n-sequences, Wronskian sign — is appropriate, and the reduction to the uniform matroid formula of Gao-Xie plus the Ferroni-Nasr-Vecchi paving formula is clean. The higher-order Turán corollary via Mařík is a nice bonus.\n\nThe soft spots are real but not fatal to the approach. First, the proof of the three key lemmas (3.3–3.5) reduces to a Mathematica Resolve[ForAll] claiming certain discriminants are positive. No code or certificate is shipped, and the printed inputs are fragmented (the n∑ notation, undeclared integer domains). For a computer-assisted proof, that is not auditable. Second, Lemmas 3.4 and 3.5 only cover d≥7: the Resolve hypotheses say n≥3, i.e., d≥7, and unlike Lemma 3.3 there is no small-d check for d=3,4,5,6. The interlacing chain in Lemma 3.2 uses these lemmas for every d≥3, so as written the proof has a gap for small rank. It is probably easy to fill — those cases have degree at most 2 — but it needs to be written down. There are also minor typos (the reference to 'Lemmas 3.2, 3.3, and 3.4' in the proof of Lemma 3.2 should include 3.5, and the labels p3/p5 are mixed).\n\nThe main logic from the lemmas to Theorem 1.3 is sound, including the trick of showing the coefficient sequence is an n-sequence from real-rootedness of B at maximal degree, which handles the degree-drop issue gracefully. The citation pattern is fine.\n\nThis deserves a serious referee. Send it to peer review, but ask the authors for reproducible computational support — a notebook, code, or an analytic proof of the discriminant inequalities — and for a complete treatment of d=3..6 in Lemmas 3.4 and 3.5. If those are supplied, the paper should be a solid contribution.","headline":"Strong result for paving matroids, but the proof hinges on unverified Mathematica discriminants and misses small-rank cases in two lemmas.","tokens_in":12064,"tokens_out":9569,"would_cite":true,"duration_ms":85857,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05A20","05B35","33F10","26C10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any paving matroid, the inverse Kazhdan-Lusztig polynomial is log-concave with no internal zeros.","keywords":["log-concavity","inverse Kazhdan-Lusztig polynomial","paving matroid","real-rooted polynomials","interlacing","multiplier sequence","Hadamard product","Newton's inequalities"],"falsifier":"Run the same three discriminant checks with $n,m,d$ explicitly declared integers (or search the parameter range numerically) and find any triple with a non-positive discriminant; alternatively, compute $B(Q_M(t))$ for a concrete paving matroid and find a non-real root.","tokens_in":11083,"feed_emoji":"📐","tokens_out":5980,"duration_ms":50290,"temperature":0.7,"pith_summary":"This paper proves a real-rootedness theorem that yields log-concavity: for every paving matroid, the Hadamard product of its inverse Kazhdan-Lusztig polynomial with $(1+t)^n$ has only real roots. Newton's inequalities then force the coefficients to form a log-concave sequence with no internal zeros, resolving a 2021 conjecture for the paving family. The proof reduces the problem to three explicit interlacing relations among uniform-matroid polynomials, with the needed sign of the Wronskian checked at $t=0$ and three cubic discriminants verified by computer algebra. The result also implies higher-order Turán inequalities for these coefficients.","feed_headline":"Paving matroid inverse Kazhdan-Lusztig polynomials are log-concave","feed_subtitle":"Real-rootedness of the Hadamard twist forces Newton's inequalities on the coefficients.","key_machinery":"The gearing is the Hadamard operator $B$: $B(\\sum a_i t^i)=\\sum \\binom{n}{i} a_i t^i$, which turns coefficient inequalities into root-location statements. Interlacing of real-rooted polynomials (via Hermite-Kakeya-Obreschkoff and Wronskian sign checks) and multiplier sequences or $n$-sequences (Pólya-Schur, Craven-Csordas) carry the proof. The concrete engine is the paving-matroid decomposition $Q_M(t)=Q_{U_{m,d}}(t)-\\sum_h \\lambda_h (Q_{U_{h,d}}(t)-Q_{U_{h,d-1}}(t))$; interlacing is proved by checking Wronskians at $t=0$ and by computer-assisted positivity of discriminants of three cubic polynomials $p_1,p_3,p_5$.","core_discovery":"The central claim is Theorem 1.3: for a paving matroid $M$ of rank $d$, the degree-$n$ polynomial $B(Q_M(t))$ obtained by multiplying the coefficient of $t^i$ by $\\binom{n}{i}$ has only real roots. Since the coefficients of $Q_M(t)$ are nonnegative and form an $n$-sequence with respect to this operator, the algebraic characterization of $n$-sequences transfers real-rootedness to $B(Q_M(t))$; Newton's inequalities then give Theorem 1.2, the log-concavity of $Q_M(t)$ with no internal zeros. The argument uses the decomposition of $Q_M(t)$ for paving matroids as a nonnegative combination of differences of uniform-matroid inverse Kazhdan-Lusztig polynomials and proves interlacing of the corresponding $B$-transforms.","pith_inferences":["The Hadamard-twist method may extend to non-paving families, since the proof only needs a uniform-type decomposition plus interlacing of $B$-transforms; checking the few small non-paving matroids with known inverse Kazhdan-Lusztig polynomials would be a cheap first test.","If the more general conjecture holds, the log-concavity conjecture would follow for all matroids, and the real-rooted operator $B$ would give a route to higher-order Turán inequalities beyond paving.","The paper's positivity method, combined with the proposed degree conjecture, suggests that paving matroids could serve as a test bed for strict positivity of Kazhdan-Lusztig polynomial coefficients."],"forward_implications":["For every paving matroid, the coefficients of $Q_M(t)$ form a log-concave sequence with no internal zeros (Theorem 1.2).","The same coefficients satisfy the higher-order Turán inequalities, by Corollary 1.5 and Mařík's theorem.","Every coefficient of $Q_M(t)$ below the top degree is positive, with the top coefficient possibly zero (Proposition 4.3).","The real-rootedness of $B(Q_M(t))$ supplies a template for the more general conjecture: if $B(Q_M)$ is real-rooted for a matroid, log-concavity follows automatically."],"supporting_citations":[{"why":"Introduces the inverse Kazhdan-Lusztig polynomial and supplies the explicit uniform-matroid formula used throughout.","marker":"[10]"},{"why":"Gives the paving-matroid decomposition into uniform-polynomial differences, the backbone of the proof.","marker":"[9]"},{"why":"Establishes nonnegativity of the coefficients, needed to apply the n-sequence characterization.","marker":"[1]"},{"why":"Supplies the interlacing lemmas and Wronskian sign criterion used to chain the interlacing relations.","marker":"[17]"},{"why":"Supplies the algebraic characterization of n-sequences, transferring real-rootedness from the combination back to B(Q_M(t)).","marker":"[5]"},{"why":"Gives the Pólya-Schur multiplier-sequence theorem underlying the real-rootedness of auxiliary polynomials.","marker":"[14]"},{"why":"Provides the Hermite-Kakeya-Obreschkoff theorem used to convert real-rooted linear combinations into alternating roots.","marker":"[2]"}],"fun_headline_variants":["Paving matroid inverse Kazhdan-Lusztig polynomials log-concave","Hadamard twist proves log-concavity for paving matroid polynomials","Real-rooted twist forces log-concavity for paving matroids","Log-concavity confirmed for paving matroid inverse KL polynomials","Paving matroids: inverse KL polynomials log-concave"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the computer algebra output correctly proves strict positivity of the three cubic discriminants for all allowed integers; if that output is faulty or the domains were not specified as integers, the interlacing chain and both theorems fall.","fun_headline_variants_meta":{"raw":{"variants":["Paving matroid inverse Kazhdan-Lusztig polynomials log-concave","Hadamard twist proves log-concavity for paving matroid polynomials","Real-rooted twist forces log-concavity for paving matroids","Log-concavity confirmed for paving matroid inverse KL polynomials","Paving matroids: inverse KL polynomials log-concave"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000898,"raw_usage":{"total_tokens":3806,"prompt_tokens":820,"completion_tokens":2986,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":436,"completion_tokens_details":{"reasoning_tokens":2890}},"tokens_in":436,"tokens_out":2986,"duration_ms":19531,"temperature":1.0,"reasoning_tokens":2890,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:37:26.965083+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same three discriminant checks with $n,m,d$ explicitly declared integers (or search the parameter range numerically) and find any triple with a non-positive discriminant; alternatively, compute $B(Q_M(t))$ for a concrete paving matroid and find a non-real root.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the inverse Kazhdan-Lusztig polynomial and supplies the explicit uniform-matroid formula used throughout."},{"cited_title":"Ferroni, G","cited_arxiv_id":null,"evidence_quote":"Gives the paving-matroid decomposition into uniform-polynomial differences, the backbone of the proof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the interlacing lemmas and Wronskian sign criterion used to chain the interlacing relations."},{"cited_title":"Craven and G","cited_arxiv_id":null,"evidence_quote":"Supplies the algebraic characterization of n-sequences, transferring real-rootedness from the combination back to B(Q_M(t))."},{"cited_title":"Pólya and J","cited_arxiv_id":null,"evidence_quote":"Gives the Pólya-Schur multiplier-sequence theorem underlying the real-rootedness of auxiliary polynomials."},{"cited_title":"Brändén, On linear transformations preserving the Pó lya frequency property, Trans","cited_arxiv_id":null,"evidence_quote":"Provides the Hermite-Kakeya-Obreschkoff theorem used to convert real-rooted linear combinations into alternating roots."}],"review_version":1}