{"id":"ed6029fb-e6df-4670-9e03-a77f8d7c7f8e","arxiv_id":"2607.16832","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Ideal Gårding polynomials — a convexity-enhanced subclass of Gårding polynomials — are characterized by concavity and Lorentzian conditions, and univariate members are modeled by Pitman–Stanley polytope volumes.","lead":"Fang and Ma introduce 'ideal Gårding polynomials,' a class of polynomials with convex positivity regions that sits between stable and Gårding polynomials. They prove a structure theorem connecting convexity, log-concavity, and Lorentzian polynomials, with a geometric model for the univariate case via Pitman–Stanley polytopes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unproven binary-relation dichotomy (Thm 9.6) is load-bearing: it supports the linear preserver theorem, which supplies closure properties used in the proof of Theorem 1.3.","rationale":"The reader's verdict of CONDITIONAL is appropriate. The main structural proofs—universal quotient concavity, polarization, and Lorentzian homogenization—are detailed and internally consistent. However, the proof of Theorem 1.3 relies on Theorem 11.1, which in turn relies on the multi-affine linear preserver theorem, whose proof depends on Section 9's binary-relation results that are omitted as 'identical to [24]'. This is a genuine gap in the manuscript's self-containedness, and it is load-bearing for the central claim. I did not find an internal inconsistency in the main analytical arguments (Theorems 6.3, 6.6, 7.3, 8.1, 12.5), so the conditionality rests on the need to supply the missing foundational proofs, not on a demonstrated counterexample. The concrete test would settle whether the omitted results are true as stated or require additional hypotheses; until then, CONDITIONAL remains the right verdict.","tokens_in":29932,"tokens_out":37292,"duration_ms":320645,"concrete_test":"Write out the full proof of Theorem 9.6 and Prop. 9.10 by transcribing the corresponding arguments from [24], verifying at each step that the relevant components C_{∂^α h} and C_{∂^α g} remain convex. In particular, test the dichotomy on the minimal edge case h(x,y)=y-x (f=1, g=-x): if this degree-1 polynomial is ideal but satisfies neither alternative, the theorem needs a qualification excluding constant f or deg h≥2. Then verify Prop. 9.10 for a small univariate example such as f=x+1, g_1=x^2+1, g_2=x^2+x+1, checking directly that f≺≺g_1+g_2. If any step uses convexity not implied by the hypotheses, the linear preserver theorem and Theorem 1.3's closure route have a real gap.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The weakest load-bearing point is an omitted proof, not a single equation. Section 9 defines ideal domination (◀) and ideal position (≺≺), and Theorem 9.6 gives a dichotomy for h(x,y)=f(x)y+g(x)∈I. The authors state: 'The proof ... is identical to similar results in [24]. We skip the proof.' This theorem, together with Prop. 9.10, is used in Lemma 10.1, which is the core of the multi-affine linear preserver theorem (Thm 10.2). Theorem 10.2 is then used in Theorem 11.1 to prove closure under positive affine pullback (11.1(2)), specialization (11.1(3)), and directional derivatives (11.1(4)). These closure properties are invoked in the proof of Theorem 1.3: (2)⇒(1) uses pullback closure; (3)⇒(1) uses specialization; Prop. 12.2 uses (1) and (3). Thus if Theorem 9.6 or Prop. 9.10 is false, or even requires extra hypotheses (e.g., excluding degree 1 or constant f), the linear preserver theorem and hence Theorem 11.1 collapse, and Theorem 1.3 is unsupported. Moreover, the dichotomy is not a trivial restatement of [24]: the Gårding domination relation in [24] does not incorporate the convexity of quotients that defines ◀ and ≺≺. The 'identical' claim must be checked: convexity must survive every step of the [24] proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces ideal Gårding polynomials, a subclass of Gårding polynomials whose distinguished Gårding components are required to be convex and remain convex under all partial derivatives. The main structure theorem (Theorem 1.3) asserts that this condition is equivalent to: being a pullback of a multi-affine ideal Gårding polynomial; having a polarization that is multi-affine ideal Gårding; having every derivative quotient (∂^α f)^{1/deg} concave on the corresponding component; having log ∂^α f concave; and having all localized homogenizations H_{x0}∂^α f Lorentzian. The paper also gives a universal univariate model via Pitman–Stanley polytopes and monotone root sequences, establishes quotient concavity, proves closure and polarization theorems, and develops a linear preserver theory. The main analytic engine is the concavity of the universal quotient, transferred fibrewise in Section 7, which then powers the polarization theorem and the Lorentzian characterization.","tokens_in":30381,"tokens_out":18478,"duration_ms":162390,"significance":"If the main theorems are correct, the paper substantially extends the structure theory of Gårding polynomials while imposing convexity, placing the class strictly between real-stable and Lorentzian polynomials. The universal univariate model via Pitman–Stanley polytopes is elegant and yields concrete volume interpretations. The paper contains no fitted parameters and makes falsifiable, precise structural claims. The self-contained proof of Theorem 6.3 (concavity of the Pitman–Stanley quotient) is a clear contribution. However, the manuscript relies heavily on the companion paper [24] for foundational facts, and several load-bearing results in Section 9 are asserted without proof under the claim that the proofs are 'identical to [24]'. Because the relations ◀ and ≺≺ are defined using convexity of quotients, this transfer is not automatic and must be verified. The paper is plausible and likely fixable, but the current manuscript leaves an essential part of the proof infrastructure unverified.","major_comments":[{"comment":"The proofs of Theorem 9.6 and Lemmas/Propositions 9.7–9.10 are omitted, with the statement that they are identical to results in [24]. This is not a routine transfer: the relations ◀ and ≺≺ are defined here using convexity of quotients ∂^α f/∂^α g and g/f on the relevant Gårding components, a condition not present in the domination relation of [24]. Theorem 9.6 is used in Lemma 10.1 to obtain h11≺≺h10 and h11≺≺h01, and Proposition 9.10 is then used to sum them; this feeds Theorem 10.2, Theorem 11.1, and ultimately Theorem 1.3. A proof, or at least a precise step-by-step reduction showing that the [24] arguments preserve quotient convexity, is required. As written, a central pillar of the paper is unverified.","section":"§9, Theorems 9.6–9.10"},{"comment":"Theorem 12.1(4) states that every localized homogenization H_{x0}∂^α f is Lorentzian for every x0∈C_{∂^α f}. The proof says '(1)⇒(4) is due to Theorem 12.5', but Theorem 12.5 is stated only for Hf with f∈I_+. The missing reduction is: translate by x0 and use Lemma 3.1 to obtain g_x0(t)=∂^α f(x0+t)∈I_+, and then H_{x0}∂^α f = H(g_x0). This step should be made explicit. In the proof of Theorem 12.5 the sentence 'It remains only to treat A direct Taylor expansion' is incomplete; moreover, the Hessian is checked only at (0,0), so the author should explicitly note that the relevant quadratic polynomials are homogeneous and therefore have constant Hessian. Without that remark, the verification of the Lorentzian Hessian condition on the whole positive orthant is not apparent.","section":"§12, Theorem 12.5 / Theorem 12.1(4)"},{"comment":"The equality C_{∂^i_y f} = {(x,y) : x∈D_i, y>r_i(x)} is asserted in a single sentence. The preceding Corollary 2.4 only shows that each fibre ∂^i_y f(x,·) is univariate Gårding; it does not by itself identify the global distinguished component as the epigraph of r_i over D_i. Since the convexity of r_i is used in Theorem 7.3 and hence in the polarization theorem (Theorem 8.1), the authors should either prove this fibrewise description or cite a specific statement in [24] that does so.","section":"§7, Lemma 7.2"}],"minor_comments":[{"comment":"When ∂_i f ≡ 0, the expression C_{∂_i f} is not defined. The intersection should be over those i with ∂_i f not identically zero, or an explicit convention for C_0 should be introduced.","section":"Definitions 1.1 and 2.1"},{"comment":"The conclusion is written as ∫_{a+b}^{α+β} dz/h(z) ≤ max(...), but the proof and application use ∫_{α+β}^{a+b} dz/h(z). The bounds are reversed in the statement; please correct.","section":"Lemma 6.1"},{"comment":"The orientation of the integral in the formula for ψ(t,a) is unclear as printed: it appears to be ∫_{a1}^t, but the subsequent reasoning uses ∫_t^{a1}. Please clarify the sign/orientation.","section":"Eq. (6.6), §6.2"},{"comment":"The domain notation b ∈ R_{>0} × Γ+_{d-2} is not consistent with the coordinates b_i = r_{i-1} − r_i, which may vanish on the boundary of U_d^{(1)}. It should likely be R_{>0} × R_{\\ge0}^{d-2}, with continuity used to extend concavity.","section":"Theorem 6.6, second case"},{"comment":"The convention '0≺≺g≺≺0' seems to be a typo; it presumably means 0≺≺g and g≺≺0 for any g∈I, but as written it is confusing.","section":"Definition 9.4"},{"comment":"The sentence 'It remains only to treat A direct Taylor expansion at the origin gives' is grammatically incomplete and should be rewritten as a proper claim with a displayed equation and an explicit statement that the Hessian of a homogeneous quadratic is constant.","section":"Theorem 12.5, proof"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious contribution if the omitted proofs can be supplied. Given the heavy reliance on [24], the editor may wish to verify that [24] is available and that its statements used here are indeed correct as cited. The Section 9 gap is the main risk; it is not merely a matter of exposition. The remaining issues are local and fixable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"New class, plausible, but the main equivalence leans on an unproven dichotomy in Section 9, and the paper also depends on the companion preprint [24] for its base theory. The referee should be sent in, but told to check those points.\n\nWhat is actually new: the recursive definition of ideal Gårding polynomials, the six-way structure theorem (Theorem 1.3), the polarization theorem (8.1), the universal Pitman–Stanley model of univariate Gårding polynomials, and the quotient concavity of Ψ_d/q_d. The proofs of the fibre quotient concavity (7.3) and the polarization theorem are worked out in detail and read convincingly. The Lorentzian homogenization argument in Section 12 is also substantial. This is a real step beyond the authors' earlier paper [24], and the examples (real stable polynomials, eigen-polynomials of nonnegative matrices) make the class worth having.\n\nWhere it is soft. Section 9 is the problem. Theorem 9.6 — the dichotomy for h(x,y)=f(x)y+g(x) to be ideal — is asserted with \"proof identical to similar results in [24]… we skip the proof.\" Proposition 9.10 is likewise dismissed as \"essentially the same.\" These are not cosmetic omissions. Lemma 10.1 needs Proposition 9.10; Theorem 10.2 (the multi-affine linear preserver) needs Lemma 10.1; Theorem 11.1's closure properties need Theorem 10.2; and Theorem 1.3's equivalences (1)↔(2) and (1)↔(3) go through Theorem 11.1. So the skipped proofs are load-bearing, and the stress-test point is good: the ideal domination relation includes convexity of quotients, which is not present in the analogous Gårding domination in [24]. The \"identical\" claim has to be checked, not taken on faith.\n\nThere is also the companion issue: Definition 2.1, Theorem 2.3, Theorem 2.5, and the universal representation are taken from [24], an unpublished preprint. A referee will need to verify the foundations there too. That does not make the paper wrong, but it makes it not self-contained.\n\nBottom line: the analytic core (Sections 5–8, 12) is genuinely new and mostly proved; the algebraic closure part (Sections 9–11) rests on omitted proofs that are central. I would send it to a careful referee rather than desk-reject, but the referee should be explicitly told to check Section 9 and the dependence on [24]. If those hold, this is a useful paper for the Lorentzian/convex-geometry and Hessian-PDE audiences.","headline":"Introduces a plausible new class between stable and Gårding with a substantial structure theorem, but the main equivalence leans on an unproven dichotomy in Section 9 and on the companion preprint [24].","tokens_in":30778,"tokens_out":5180,"would_cite":true,"duration_ms":45705,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A Gårding polynomial is 'ideal' when every derivative has a convex positivity component—and this paper proves that condition is equivalent to concavity, log-concavity, and Lorentzian localized homogenization.","keywords":["ideal Gårding polynomials","Gårding polynomials","real stable polynomials","Lorentzian polynomials","Pitman–Stanley polytopes","polarization","concavity","linear preservers"],"falsifier":"Numerically test the key superadditivity inequality of Theorem 6.3: for random positive a,b in Γ+_d with d=3,4, check whether Ψ_d(a+b) ≥ Ψ_d(a) + Ψ_d(b), where Ψ_d = Vol(Ξ_d)/Vol_{d−1}(Ξ_d,Δ_1). A single violation would destroy the concavity of the universal quotient and with it the polarization theorem.","tokens_in":29888,"feed_emoji":"📐","tokens_out":6358,"duration_ms":52504,"temperature":0.7,"pith_summary":"The paper introduces ideal Gårding polynomials, the subclass of Gårding polynomials whose distinguished positivity components (Gårding components) are convex and remain so under all partial derivatives. Its main theorem says that for a degree-d Gårding polynomial f, the following are equivalent: f is ideal; every nontrivial derivative satisfies (∂^α f)^(1/deg) concave on its component; every nontrivial derivative is log-concave there; and every localized homogenization is Lorentzian. The paper proves that ideal Gårding polynomials are preserved under polarization, satisfy closure properties under positive affine maps, products, and directional derivatives, and admit a linear preserver theory. It also gives a universal model for univariate Gårding polynomials via monotone root sequences and Pitman–Stanley polytopes, which yields quotient concavity and Newton–Maclaurin-type inequalities. A sympathetic reader should care because this places a strictly larger class than real stable polynomials inside the Lorentzian framework while retaining its robust structure theory.","feed_headline":"Convex Gårding polynomials keep polarization and Lorentzian structure","feed_subtitle":"The convex subclass preserves polarization and linear preserver theory while lying inside the Lorentzian class.","key_machinery":"Central machinery: the universal quotient q_d(br)=p_d/∂_-p_d for univariate Gårding polynomials, realized via Pitman–Stanley polytopes as the volume/mixed-volume ratio Ψ_d. The paper proves Ψ_d is concave (equivalently superadditive) on Γ+_d and q_d is concave and nonincreasing in each root coordinate. This concavity-plus-monotonicity is then transported fibrewise: for ideal f(x,y), each root r_i(x) of ∂_y^i f is convex, hence f/∂_y f = q_d(y, r_0(x),…, r_{d-1}(x)) is concave on C_{∂_y f}. That quotient concavity is the load-bearing step for polarization.","core_discovery":"The core discovery is that imposing recursive convexity on Gårding components does not dismantle the Gårding structure theory. Theorem 1.3 equates ideality (all C_{∂^α f} convex) with: (∂^α f)^(1/deg) concave on C_{∂^α f}; log ∂^α f concave there; and H_{x0}∂^α f Lorentzian. The technical engine is a universal model of univariate Gårding polynomials as volumes of Pitman–Stanley polytopes; its universal quotient q_d is concave and coordinatewise decreasing, and this fibrewise yields the quotient concavity f/∂_y f used to prove polarization.","pith_inferences":["If Theorem 1.3(6) holds as stated, ideal Gårding polynomials form a unified source of fully nonlinear elliptic operators of Hessian type in which ellipticity and convexity coexist; the quotient concavity of f/∂_y f may directly yield PDE estimates for subequations modeled on these polynomials.","The Pitman–Stanley volume model suggests a probabilistic reading: univariate Gårding polynomials are generating functions of nested simplices, and the quotient concavity may correspond to Brunn–Minkowski-type inequalities for conditional volumes that are not available for general convex bodies.","The binary-relation results (ideal domination and ideal position) are asserted to be identical to those in the companion paper; supplying those proofs would make the linear preserver theorem fully self-contained and could give a convexity-aware analogue of interlacing for hyperbolic polynomials.","A testable algorithmic consequence of the equivalence (1)⇔(6): given a Gårding polynomial, one could numerically verify ideality by checking M-convex support plus the Hessian condition for every localized homogenization."],"forward_implications":["Every positive real stable polynomial is ideal Gårding, and ideal Gårding polynomials are preserved under polarization and strictly positive affine pullbacks (Theorem 8.1 and Theorem 11.1).","Localized homogenizations of ideal Gårding polynomials are Lorentzian, so the class supplies new Lorentzian examples, including eigen-polynomials of nonnegative matrices and their M-matrix variants.","Newton–Maclaurin-type inequalities hold for all Gårding polynomials: directional derivative sequences satisfy the refined bound (D_v^j f)^2 ≥ (ℓ-j+1)/(ℓ-j) (D_v^{j-1} f)(D_v^{j+1} f), recovering the classical Newton–Maclaurin inequality when f is an elementary symmetric polynomial.","A linear operator whose symbol is an ideal Gårding polynomial with nonnegative coefficients preserves ideal Gårding polynomials, in both multi-affine and general settings."],"fun_headline_variants":["Ideal Gårding: convex components, same polarization power","Recursive convexity doesn't break Gårding polarization","Convex Gårding and Lorentzian: still polarizable","Universal Pitman-Stanley model for convex Gårding","Quotient concavity proves polarization for ideal Gårding"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The recursive definition of Gårding polynomials and the structural facts inherited from the authors' companion paper—component condition, preservation under positive affine maps and polarization, the Rayleigh property, and the monotone-root-sequence representation of univariate Gårding polynomials—are taken as given; the quotient concavity and polarization theorems build directly on them.","fun_headline_variants_meta":{"raw":{"variants":["Ideal Gårding: convex components, same polarization power","Recursive convexity doesn't break Gårding polarization","Convex Gårding and Lorentzian: still polarizable","Universal Pitman-Stanley model for convex Gårding","Quotient concavity proves polarization for ideal Gårding"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001039,"raw_usage":{"total_tokens":4173,"prompt_tokens":676,"completion_tokens":3497,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":420,"completion_tokens_details":{"reasoning_tokens":3411}},"tokens_in":420,"tokens_out":3497,"duration_ms":21839,"temperature":1.0,"reasoning_tokens":3411,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T19:49:50.649771+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically test the key superadditivity inequality of Theorem 6.3: for random positive a,b in Γ+_d with d=3,4, check whether Ψ_d(a+b) ≥ Ψ_d(a) + Ψ_d(b), where Ψ_d = Vol(Ξ_d)/Vol_{d−1}(Ξ_d,Δ_1). A single violation would destroy the concavity of the universal quotient and with it the polarization theorem.","supporting_citations":[],"review_version":1}