{"id":"4782374e-ff57-4728-a448-1614f7c20d65","arxiv_id":"2507.09793","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A vector-valued BKK theorem: generic zeros of a torus-invariant vector-valued Laurent polynomial are counted by the mixed volume of virtual polytopes delta_i - delta_{i-1}, and the associated mixed volumes satisfy an Alexandrov-Fenchel inequality extended to polymatroids.","lead":"Generalizes the classical BKK theorem to systems of vector-valued Laurent polynomial equations, counting solutions via mixed volumes of virtual polytopes built from an invariant subspace. The same framework yields new Alexandrov-Fenchel inequalities for these counts, with an extension to arbitrary polymatroids via matroid Hodge theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sign error in Lemma 5.6 and Eq. (11) invalidates the induction proof of the central BKK theorem: under the paper's minimum support-function convention, the cube example gives RHS = -n! while LHS = n!, so Theorem 5.1 is not established as written.","rationale":"The reader's conditional verdict is correct, but for a different reason than the one flagged. The most load-bearing issue is not the polymatroid extension in Section 9; it is a sign error in the proof of the paper's headline result, Theorem 5.1 (equivalently Theorem 1.14). The paper consistently uses the minimum convention for support functions. Under that convention, the standard mixed-volume integration formula acquires a minus sign: integrating h_1...h_n over a complete toric variety gives n! times the mixed volume of the corresponding virtual polytopes, but expressing it as a sum over primitive rays with the minimum support values as in Lemma 5.6 gives the negative of that. Testing on the unit cube shows Lemma 5.6 fails: LHS n!, RHS -n!. The same missing minus sign propagates into equation (11), where (9), Lemma 5.5 and (10) imply a negative sum, so the induction cannot conclude |Y(f)| = n! MVol. A rank-1 example makes the failure concrete: the displayed (11) would count -1 solutions. This is not merely a cosmetic typo: the proof of the central theorem is internally inconsistent as written. The theorem may well be true, and the fix (inserting a global minus sign in Lemma 5.6 and (11), or switching to the maximum support-function convention throughout) appears straightforward, but the current manuscript does not establish it. The polymatroid gap identified by the reader is real but secondary: it concerns an advertised extension whose proof is a sketch, while the sign error affects the main vector-valued BKK theorem. Since the paper's overall conclusion should remain conditional pending these corrections, the reader's verdict is unchanged.","tokens_in":29686,"tokens_out":16316,"duration_ms":184979,"concrete_test":"Compute both sides of Lemma 5.6 for P1=...=Pn=[0,1]^n using the paper's minimum support-function convention on the normal fan: for n=1 the RHS is -1 and the LHS is 1; for n=2 the RHS is -2 and the LHS is 2. If the inequality persists, the lemma is false as stated. Then re-derive equation (11) from (9), Lemma 5.5, and (10): the missing minus sign yields a negative right-hand side in the rank-1 example L=C*1+C*x, contradicting the count |Y(f)|=1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central theorem (Theorem 5.1 / Theorem 1.14) is proved by induction using Lemma 5.6, but Lemma 5.6 is false with the paper's minimum convention for support functions (Definition 1.6, Theorem 1.10). Take P1=...=Pn=[0,1]^n and let Sigma be the normal fan. For every positive coordinate ray xi=e_i, the support value h_n(xi)=0 and P_i^xi has (n-1)-volume 1; for every negative ray xi=-e_i, h_n(xi)=-1 and P_i^xi has (n-1)-volume 1. The sum in Lemma 5.6 is n*0 + n*(-1) = -n, so the RHS equals -(n-1)! * n = -n!, while the LHS is n! * 1 = n!. For n=1 this is -1 vs 1, for n=2 it is -2 vs 2. The same sign loss appears in equation (11): combining (9), Lemma 5.5 and (10) gives |Y(f)| = -(n-1)! * sum_xi h_n(xi) MVol(...), not the displayed positive expression. Indeed, for the rank-1 example L = C*1 + C*x with A={0,1}, equation (11) as written yields -1, while the true count is 1. Thus the induction step of the proof does not close as stated; a minus sign must be inserted in Lemma 5.6 and (11), or the min/max convention changed consistently.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper generalizes the BKK theorem from systems of scalar Laurent polynomials to T-invariant subspaces L of vector-valued Laurent polynomials. For a rank-r subspace L, the authors define a characteristic sequence of polytopes (Δ_1,...,Δ_r) and an r-valued support function h_L, and they prove that for generic f in L the zero set Y(f) is computed by n!MVol_n(h_L). They interpret h_L as the equivariant Chern roots of an associated toric vector bundle, use this to derive Alexandrov-Fenchel type inequalities for mixed volumes of the characteristic polytopes, and finally attempt to extend these inequalities to arbitrary polymatroids via Lorentzian fans. Applications to hyperplane arrangements and matroid classes are also given.","tokens_in":29981,"tokens_out":12526,"duration_ms":143856,"significance":"If the gaps identified below are repaired, the vector-valued BKK theorem is a substantial and natural generalization of Bernstein-Kushnirenko-Khovanskii theory, with clean combinatorial answers in terms of characteristic polytopes and multi-valued support functions. The toric vector bundle construction and the Chern-root interpretation of h_L are elegant, and the applications to linear subspaces and matroids are valuable. The paper is mostly written in a classical proof style, with detailed treatment of truncations and non-degeneracy, but it does not include machine-checked proofs or reproducible code; the main load-bearing issues are a sign error in the inductive mixed-volume formula and two proof gaps in the genericity and polymatroid arguments.","major_comments":[{"comment":"The inductive mixed-volume formula is false under the minimum convention for support functions adopted in Section 2. For example, take P_1=...=P_n=[0,1]^n and let Sigma be the normal fan of the cube. For each positive ray xi=e_i, h_n(xi)=0; for each negative ray xi=-e_i, h_n(xi)=-1. Each face P_i^xi has (n-1)-volume 1, so the right-hand side of Lemma 5.6 equals (n-1)! * (n*0 + n*(-1)) = -n!, while the left-hand side is n! MVol_n(P_1,...,P_n)=n!. The same sign loss appears in Eq. (11): combining (9), Lemma 5.5 and (10) yields |Y(f)| = -(n-1)! * sum_rho h_n(xi) MVol(...), not the displayed positive expression. Since the induction proof of Theorem 5.1 closes only through this formula, the proof of the vector-valued BKK theorem does not close as written. The fix is local: replace h_n(xi) by -h_n(xi) in Lemma 5.6 and Eq. (11), or switch consistently to the maximum convention throughout the mixed-volume formula.","section":"Section 5, Lemma 5.6 and Eq. (11)"},{"comment":"The proof uses the assertion that 'every semi-algebraic set either contains a Zariski open or is contained in a Zariski closed'. This assertion is false; for instance, a real hyperplane in C^n is a semi-algebraic set that is neither contained in a proper complex Zariski closed set nor contains a nonempty complex Zariski open set. The argument needs a proof that the locus U_xi of xi-nondegenerate f is constructible, and then a density argument; without this, the Zariski-openness of L-non-degeneracy is not established. This genericity statement is used in Theorem 1.11 and in the 'generic f' statement of Theorem 1.14, so the gap is load-bearing.","section":"Section 4.2, proof of Theorem 4.8(c)"},{"comment":"The proof assumes without proof that every complete smooth fan is Lorentzian, that products of Lorentzian fans are Lorentzian, and that smooth refinements of Lorentzian fans are Lorentzian. In particular, the first assertion is not a consequence of the normal-fan examples in [AHK18] as cited: complete smooth fans need not be projective, so they are not covered by the projective normal fan case. No reference or proof is supplied for these closure properties in the form used here. Since the Alexandrov-Fenchel inequality for arbitrary polymatroids is concluded by applying the Lorentzian property to the refined fan Sigma_tilde, the extension to non-representable polymatroids is not established as written. The authors should either prove or cite these properties, or restrict the claim accordingly.","section":"Section 9, proof of Theorem 9.3 (Theorem 1.19)"},{"comment":"The final assertion of Theorem 1.14 ('for any f in L, the number of isolated solutions, counted with multiplicity, is <= n! MVol_n(h_L)') is not proved in Section 5. Theorem 5.1 proves only the equality for L-non-degenerate f, and Remark 5.2 concerns finiteness with multiplicity for a weaker non-degeneracy condition. A specialization or limiting argument is needed to justify the bound for arbitrary f, and it should be included if this part of the theorem is retained.","section":"Theorem 1.14, upper-bound statement"}],"minor_comments":[{"comment":"Equation (16) appears to be missing the square on the left-hand side: the correct Alexandrov-Fenchel form, as stated in Theorem 1.17, is MVol(h_{L1⊕L2⊕L3})^2 >= MVol(h_{L1⊕L1⊕L3}) MVol(h_{L2⊕L2⊕L3}).","section":"Section 8, Corollary 8.3, Eq. (16)"},{"comment":"The reference list entry for [BKK76] is merged into the [BCF23] entry and is missing its title and page details; the title 'Newton polyhedra' is appended to [BCF23] instead of appearing in the [BKK76] entry.","section":"References"},{"comment":"The title in the manuscript header reads 'VECTOR-V ALUED LAURENT POLYNOMIAL EQUATIONS' but should read 'VECTOR-VALUED LAURENT POLYNOMIAL EQUATIONS'.","section":"Header"},{"comment":"The displayed formula in Corollary 8.2 has an extra closing parenthesis: '..., P_{n-r}))' should be '..., P_{n-r})'.","section":"Section 8, Corollary 8.2"},{"comment":"There is a duplicated word in Remark 2.1: 'Some authors define the the support function' should be 'Some authors define the support function'.","section":"Remark 2.1"},{"comment":"The notation 'Y f' in Theorem 5.4 should be 'Y(f)' for consistency with the rest of the paper.","section":"Theorem 5.4"}],"recommendation":"major_revision","confidential_remarks":"The sign error in Lemma 5.6/Eq. (11) is local and fixable, and the overall BKK framework is plausible. The larger concern is the proof of Theorem 9.3: the unproved Lorentzian-fan closure properties and the constructibility gap in Theorem 4.8(c) require real mathematical work, not just editing. The paper should also clarify its novelty relative to the recent work of Esterov on engineered complete intersections, which is acknowledged but not carefully delimited. I recommend major revision rather than rejection because the central combinatorial construction appears sound and the issues are repairable within the paper's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the core result is a genuine extension of BKK to vector-valued Laurent polynomial systems, and the paper is worth refereeing, but it needs two fixes before acceptance.\n\nWhat is new and good: Theorem 1.14, expressing the generic solution count as n! times the mixed volume of the characteristic polytopes Delta_1, Delta_2 - Delta_1, ..., is a real generalization. The route through toric vector bundles and the ring of conditions is clean, and identifying the equivariant Chern roots with the multi-valued support function h_L is well explained. The hyperplane-arrangement corollary reproduces known tropical/matroid results, and the authors say so; that is a consistency check, not a priority claim. Esterov's overlapping work is acknowledged directly.\n\nThe main soft spot: Lemma 5.6 and equation (11) carry the wrong sign under the paper's minimum convention for support functions. For the unit cube, the right side of Lemma 5.6 evaluates to -n!, and the rank-one example L = C*1 + C*x gives -1 in (11) while the true number of solutions is 1. So the induction in Theorem 5.1 does not close as written. The fix is a single minus sign (or a consistent switch to the maximum convention). I see no deeper obstruction, but the proof must be corrected before the theorem is established.\n\nThe second soft spot: the proof of Theorem 9.3 (polymatroid Alexandrov-Fenchel) asserts without proof that every complete smooth fan is Lorentzian, and that products and smooth refinements of Lorentzian fans are Lorentzian. Those properties are not in the cited literature as far as I know, and the first is doubtful without projectivity. The polymatroid extension depends entirely on this, so it is currently a sketch. The authors should either prove these closure properties or provide a reference.\n\nMinor: the proof of Theorem 4.10 uses a semi-algebraic claim that is false in the stated generality. That is not load-bearing for the counting theorem, but the argument should be rewritten.\n\nWho this is for: people in Newton polyhedra, toric vector bundles, and matroid Hodge theory. The main theorem is important enough, and the presentation solid enough, to deserve serious refereeing. My recommendation: send it to review, and require the sign fix and a justification of the Lorentzian-fan assertions before acceptance.","headline":"The vector-valued BKK theorem is likely right and worth taking seriously, but the central proof has a fixable sign error and the polymatroid extension rests on unproved Lorentzian-fan claims.","tokens_in":30546,"tokens_out":4720,"would_cite":true,"duration_ms":59550,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14M25","52B40"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that the number of solutions of a generic vector-valued Laurent polynomial system is $n!$ times a mixed volume of virtual polytopes determined by an associated subspace arrangement and matroid data.","keywords":["vector-valued Laurent polynomials","BKK theorem","mixed volume","toric vector bundles","multi-valued support functions","characteristic polytopes","Alexandrov-Fenchel inequality","polymatroids"],"falsifier":"Compute the three mixed volumes in the polymatroid Alexandrov-Fenchel inequality for a small nonrepresentable polymatroid, such as a rank-3 polymatroid on four elements with a valid rank function; if the product of the outer terms exceeds the square of the middle term, the theorem is false. Separately, for the main BKK formula, one could test a rank-2 invariant subspace on $(\\mathbb{C}^*)^2$ with explicitly chosen $E_\\alpha$ and count the solutions of a generic vector equation, comparing with $2! \\operatorname{MVol}(\\Delta_1, \\Delta_2-\\Delta_1)$.","tokens_in":29459,"feed_emoji":"🧮","tokens_out":8993,"duration_ms":90752,"temperature":0.7,"pith_summary":"The paper sets out to prove a vector-valued generalization of the BKK theorem. For a torus-invariant subspace $L$ of vector-valued Laurent polynomials on $(\\mathbb{C}^*)^n$ with rank $n$, it claims that a generic vector equation $f(x)=0$ has exactly $n! \\operatorname{MVol}_n(h_L)$ isolated nondegenerate solutions, where $h_L$ is an $n$-valued support function built from the subspace arrangement defining $L$. Equivalently, the count is $n! \\operatorname{MVol}(\\Delta_1, \\Delta_2-\\Delta_1, \\ldots, \\Delta_n-\\Delta_{n-1})$ for a sequence of lattice polytopes $\\Delta_i$ whose vertices come from admissible tuples of monomials. If true, the solution count of such systems depends only on combinatorial and matroid data, just as in the classical scalar case. The paper further claims an Alexandrov-Fenchel type inequality for these mixed volumes and an extension of that inequality to arbitrary polymatroids, which would give log-concavity results beyond representable subspace arrangements.","feed_headline":"A mixed volume counts zeros of vector-valued systems","feed_subtitle":"Extends the BKK theorem: generic solution counts depend only on matroid polytopes, not on coefficients.","key_machinery":"The load-bearing object is the multi-valued support function $h_L$: for each covector $\\xi$, $h_L(\\xi)$ is the multiset of critical levels of the filtration $E^\\xi_c = \\sum_{\\langle\\xi,\\alpha\\rangle \\leq c} E_\\alpha$, each level repeated according to its multiplicity. It has a canonical representation $h_1 \\leq \\cdots \\leq h_r$ by piecewise-linear functions, and the paper identifies $h_i$ with the support function of the virtual polytope $\\Delta_i-\\Delta_{i-1}$. The same function gives the equivariant Chern roots of a toric vector bundle $E_{L,\\Sigma}$ built from $L$ on a compatible toric compactification, so counts of solutions of $f=0$ become evaluations of a top equivariant Chern class. This bundle, together with the ring-of-conditions interpretation, carries the derivation of the BKK formula and of the Alexandrov-Fenchel inequality; the polymatroid version is obtained by replacing the subspace arrangement with the natural matroid of the polymatroid and applying Hodge-theoretic intersection theory on Lorentzian fans.","core_discovery":"The central discovery is that the enumerative geometry of generic vector-valued Laurent polynomial systems is controlled by a finite piecewise-linear object. Given $L = \\bigoplus_\\alpha E_\\alpha \\otimes x^\\alpha$, define for each covector $\\xi$ the subspace $E^\\xi_c = \\sum_{\\langle\\xi,\\alpha\\rangle \\leq c} E_\\alpha$; as $c$ varies this is a filtration, and the multiset of critical levels is an $r$-valued support function $h_L$. In the full-rank case $r=n$, the paper proves that for generic $f \\in L$ all points of $Y(f)$ are isolated and nondegenerate, that $|Y(f)| = n! \\operatorname{MVol}_n(h_L) = n! \\operatorname{MVol}_n(\\Delta_1, \\Delta_2-\\Delta_1, \\ldots, \\Delta_n-\\Delta_{n-1})$, and that any $f$ has at most this many isolated solutions counted with multiplicity. Here $\\Delta_i$ is the convex hull of sums $\\alpha_1+\\cdots+\\alpha_i$ over $i$-tuples for which one can choose $e_j \\in E_{\\alpha_j}$ linearly independent, and the $\\Delta_i-\\Delta_{i-1}$ are virtual polytopes. For rank $r \\leq n$, the class of $Y(f)$ in the ring of conditions of the torus is the product $\\prod_{i=1}^r [\\Delta_i-\\Delta_{i-1}]$, yielding mixed-volume formulas when extra scalar polynomial equations are added. The same framework yields an Alexandrov-Fenchel inequality for these mixed volumes and, via Hodge theory for matroids and natural matroid constructions, an extension to arbitrary polymatroids.","pith_inferences":["The toric-vector-bundle description suggests that finer invariants, such as Euler characteristics or Hodge numbers of $Y(f)$, should also be computable from the same multi-valued support function; the paper does not carry this out.","A testable extension would be to use the product formula $[Y(f)] = \\prod_i [\\Delta_i-\\Delta_{i-1}]$ as the definition of characteristic classes for arbitrary polymatroids, which could yield new log-concavity results beyond the representable case.","Because the paper notes that a polymatroid gives rise to a tropical vector bundle, the mixed-volume count may admit a purely tropical proof independent of the characteristic-0 setting, which would be a natural next step."],"forward_implications":["Generic members of $L$ define smooth subvarieties transverse to all torus orbits in any sufficiently refined smooth toric compactification, so $Y(f)$ has the same topological invariants for all $L$-nondegenerate $f$.","When $r \\leq n$, adding $n-r$ generic scalar Laurent equations with Newton polytopes $P_1,\\ldots,P_{n-r}$ makes the solution count $n! \\operatorname{MVol}(\\Delta_1, \\Delta_2-\\Delta_1, \\ldots, \\Delta_r-\\Delta_{r-1}, P_1, \\ldots, P_{n-r})$.","The hyperplane-arrangement corollary gives an explicit mixed-volume formula for BKK-type systems pulled back from complements of arbitrary hyperplane arrangements, recovering known expressions for the class of a linear space in the ring of conditions.","The Alexandrov-Fenchel inequality for these mixed volumes yields log-concavity statements: for fixed $L$ and varying convex bodies, the volume sequence satisfies $b^2 \\geq ac$.","The polymatroid version of the inequality gives nonnegative integer inequalities for combinatorial polymatroids that do not come from vector spaces."],"supporting_citations":[{"why":"Supplies the classical theorem being generalized: solution counts for generic Laurent systems are mixed volumes of Newton polytopes.","marker":"[BKK76]"},{"why":"Provides the toric compactification and nondegeneracy framework used to prove smoothness and transversality of generic zero sets.","marker":"[Kh77]"},{"why":"Gives the classification of toric vector bundles by compatible filtrations, used to define $E_{L,\\Sigma}$ and its equivariant Chern roots.","marker":"[Kly89]"},{"why":"Introduces multi-valued support functions and convex chains, the formalism in which $h_L$ and its mixed volume are defined.","marker":"[KhPu92]"},{"why":"Shows the mixed volume of an $n$-valued support function is well-defined, independent of the representation.","marker":"[Brion94]"},{"why":"Provides the inductive mixed-volume formula used in the proof of the vector-valued BKK theorem.","marker":"[Kh88]"},{"why":"Hodge theory for matroids is the engine used to extend the Alexandrov-Fenchel inequality to arbitrary polymatroids.","marker":"[AHK18]"},{"why":"Supplies the matroid tautological class technique and the linear-space class computation that the polymatroid proof adapts.","marker":"[BEST23]"},{"why":"Contains the equivalent known computation for the class of a linear space in the ring of conditions, which the hyperplane-arrangement corollary recovers.","marker":"[Huh]"}],"fun_headline_variants":["Matroids govern zero counts of Laurent systems","BKK extends to vector-valued polynomials","Virtual polytopes count generic solutions","Alexandrov-Fenchel inequality for matroids and polymatroids","A unified mixed volume for Laurent equations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The broadest new claim, the Alexandrov-Fenchel inequality for every polymatroid, relies on three properties of Lorentzian fans—every complete smooth fan is Lorentzian, products of Lorentzian fans are Lorentzian, and smooth refinements of Lorentzian fans are Lorentzian—that the proof asserts but does not prove and does not locate in the cited literature; if any of these properties fails, the polymatroid extension is unproved.","fun_headline_variants_meta":{"raw":{"variants":["Matroids govern zero counts of Laurent systems","BKK extends to vector-valued polynomials","Virtual polytopes count generic solutions","Alexandrov-Fenchel inequality for matroids and polymatroids","A unified mixed volume for Laurent equations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001165,"raw_usage":{"total_tokens":4894,"prompt_tokens":1087,"completion_tokens":3807,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":703,"completion_tokens_details":{"reasoning_tokens":3738}},"tokens_in":703,"tokens_out":3807,"duration_ms":34049,"temperature":1.0,"reasoning_tokens":3738,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:48:12.264851+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the three mixed volumes in the polymatroid Alexandrov-Fenchel inequality for a small nonrepresentable polymatroid, such as a rank-3 polymatroid on four elements with a valid rank function; if the product of the outer terms exceeds the square of the middle term, the theorem is false. Separately, for the main BKK formula, one could test a rank-2 invariant subspace on $(\\mathbb{C}^*)^2$ with explicitly chosen $E_\\alpha$ and count the solutions of a generic vector equation, comparing with $2! \\operatorname{MVol}(\\Delta_1, \\Delta_2-\\Delta_1)$.","supporting_citations":[],"review_version":1}