{"id":"352cd039-2483-48cc-b7de-addce03c8d39","arxiv_id":"1908.08154","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For random cosine polynomials with palindromic coefficient blocks of length ℓ, the expected number of real zeros equals (2n/√3)Kℓ + O(n^{2/3}), with an explicit double-integral constant Kℓ.","lead":"This paper calculates how many times a random cosine curve, built from blocks of coefficients that are mirror images, crosses zero on average. It shows the expected count is the classical value times a constant slightly above 1 for blocks longer than one coefficient, and it gives an explicit formula for that constant.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2.1 is not proved for the stated parameter range: the proof assumes n−ℓ odd and later n even, leaving complementary parities (e.g., ℓ=2, n even) without an argument; and for ℓ=1 the claimed n/√3 contradicts the deterministic-zero count n+n/√3 in eq (1).","rationale":"The reader's REJECT verdict is justified. The parity restriction is not cosmetic: the identity E[N(0,π)] = 2E[N(0,π/2)] is used to pass from the half-interval Kac-Rice integral to the full zero count, and it depends on paired frequencies having the same parity, i.e., on n−ℓ being odd. The paper never treats n−ℓ even; for ℓ=2 this leaves all even n unproved. The ℓ=1 case is worse: eq (1), cited in the paper, gives a leading term n from deterministic zeros, so the theorem's n/√3 cannot hold for N_n(0,2π). This is an internal inconsistency, not merely a disagreement with prior results. The paper does contain plausible and potentially repairable material: the explicit double-integral constant Kℓ, the Jensen-based proof that Kℓ>1 for ℓ≥2, and the Riemann-sum machinery are all substantive. But they do not establish Theorem 2.1 as stated. A concrete re-derivation for the omitted parity, or an explicit exclusion of ℓ=1 and restriction to the parity assumed, would be needed before the claim could be accepted. Thus the reader's verdict should remain REJECT (major revision), not acceptance.","tokens_in":21318,"tokens_out":36510,"duration_ms":330009,"concrete_test":"Independently re-derive eq (11) for a case with n−ℓ even, taking ℓ=2 and n=2ℓm (so n even, n−ℓ even), by applying the Kac-Rice formula over (0,2π) without invoking the π/2 symmetry. If the leading constant is not (2n/√3)K2 with the same double-integral K2, or if parity-dependent terms survive, Theorem 2.1 as stated fails; if it is recovered, the gap is one of exposition and the theorem should at minimum be restricted to the parity subsequence actually proved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is Theorem 2.1, for fixed ℓ and all n=2ℓm+r. In the proof, the sentence \"For the sake of simplicity of our computations, we assume that n−ℓ is odd\" is used to obtain symmetry of A(x), C(x), B(x) about x=π/2, hence E[N(0,π)] = 2E[N(0,π/2)] and later E[N(0,2π)] = 4E[N(0,π/2)]. Then, in the Riemann-sum step, the paper says \"for simplicity we let n be even.\" Together these two assumptions restrict the proof to n with n−ℓ odd and n even, i.e., to a parity subsequence. For a fixed ℓ≥2, infinitely many allowed n are not covered; for instance, ℓ=2 forces n odd for n−ℓ odd, so every even n is outside the proof. No argument is supplied for n−ℓ even, where the paired basis functions have opposite parity and the π/2 symmetry fails. The stated O(n^{2/3}) asymptotic for all n therefore rests on an unproved subsequence reduction. In addition, the theorem is literally false at ℓ=1: the paper's own eq (1) says palindromic coefficients give E[N_n(0,2π)] = n + n/√3 + O(n^{3/4}) because of n deterministic zeros, while Theorem 2.1 with K1=1/2 gives n/√3. The proof's use of Kac-Rice is also invalid at the zeros of the deterministic factor, where A(x)=0. These limitations are acknowledged in the text (\"we assume n−ℓ is odd,\" \"for simplicity we let n be even,\" and the comment around (1)) but never reconciled with the theorem's unrestricted statement.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the expected number of real zeros in (0,2π) of a random cosine polynomial V_n(x)=∑_{j=0}^n a_j cos(jx) whose Gaussian coefficients are arranged into palindromic blocks of fixed length ℓ. The main result, Theorem 2.1, claims that for every fixed ℓ and every n=2ℓm+r in the stated range, E[N_n(0,2π)] = (2n/√3) K_ℓ + O(n^{2/3}), where K_ℓ is an explicit double integral, K_ℓ>1 for ℓ≥2, and K_1=1/2. The proof uses the Kac–Rice formula, asymptotic expansions of A(x), B(x), C(x), a Riemann-sum evaluation of an oscillatory integral, and numerical evaluations of K_ℓ in Table 1. The paper also explicitly recalls in eq. (1) that the ℓ=1 palindromic-coefficient case has n deterministic zeros, giving E[N_n(0,2π)] = n + n/√3 + O(n^{3/4}).","tokens_in":21690,"tokens_out":10558,"duration_ms":150291,"significance":"If the main claim were valid, it would give an exact leading constant for a natural dependent-coefficient model and would quantify the increase over the classical i.i.d. case 2n/√3. The constant K_ℓ is defined as an integral with no fitted parameters, and the numerical table is reproducible from the formula, which are strengths. However, the theorem as stated is not supported: the ℓ=1 case is contradicted by the paper's own eq. (1), and the proof is carried out only under parity restrictions that exclude infinitely many n (and, for ℓ=2, all n). There is also an unsupported localization step involving F_ℓ. These are load-bearing issues, so the current manuscript should not be accepted.","major_comments":[{"comment":"As stated, Theorem 2.1 is false for ℓ=1. For ℓ=1 the coefficients are simply palindromic, and eq. (1) of the paper gives E[N_n(0,2π)] = n + n/√3 + O(n^{3/4}) because of n deterministic zeros. Since K_1=1/2, Theorem 2.1 instead predicts the leading term n/√3, omitting the deterministic term n. The remark after the theorem and the sentence after eq. (11) repeat this omission, and the Kac–Rice argument is formally inapplicable at the deterministic zeros, where A(x)=0; for ℓ=1 the excluded set F_ℓ only removes a neighborhood of x=0, not the n zeros of cos(nx/2). The theorem must either exclude ℓ=1 or state the correct ℓ=1 asymptotics.","section":"Section 2, Theorem 2.1 and eq. (1)"},{"comment":"The proof is explicitly restricted in two places: it assumes 'n−ℓ is odd' to obtain symmetry of A, B, C about x=π/2, and later says 'for simplicity we let n be even' in the Riemann-sum step. These two assumptions together cover only the subsequence with ℓ odd and n even. The theorem, however, claims the result for every fixed ℓ and every n in the stated range. For example, when ℓ=2, the condition n−ℓ odd forces n odd while the Riemann-sum step requires n even, so no n is covered; when ℓ is odd and n is odd, neither condition holds. No argument is supplied for the complementary parities, where the π/2 symmetry and the reduction of I_ℓ(n) to half intervals fail. The O(n^{2/3}) asymptotic is therefore established only on a parity-restricted subsequence, not for all n as claimed.","section":"Proof of Theorem 2.1, parity reduction"},{"comment":"The step 'E[N_n(F_ℓ)] = O(n^{1-a}) by Lemma 3.1' is not justified. Lemma 3.1 concerns only the interval (0, n^{-a}), whereas F_ℓ also contains intervals centered at iπ/ℓ for i=1,...,[ℓ/2]. The behavior of V_n near those points is not equivalent to its behavior near 0, and a separate localization estimate would be needed. Since the subsequent Kac–Rice evaluation is performed only on E_ℓ, this gap affects the error term of the main asymptotic.","section":"Proof of Theorem 2.1, before eq. (11)"}],"minor_comments":[{"comment":"The text says 'sec(ℓx)=O(n^a)' where the subsequent division by sin(ℓx) indicates that csc(ℓx) is meant; please correct the notation.","section":"Proof of Theorem 2.1, near eq. (7)"},{"comment":"The statement that g(s,t) is 'symmetric about the line ⟨π/2,π/2,r⟩' is unclear; the needed symmetry of g under the relevant reflection in (s,t) should be stated explicitly.","section":"Proof of Theorem 2.1, symmetry statement"},{"comment":"The proofs of Lemmas 3.1 and 3.2 are not included and are attributed to the unpublished preprint [23]; please either include proofs or give a published reference, since these lemmas are used in the main proof.","section":"Lemmas 3.1 and 3.2"}],"recommendation":"reject","confidential_remarks":"I recommend rejection because the main theorem as stated is not established: the ℓ=1 case contradicts eq. (1), the parity restrictions exclude infinitely many valid n, and the localization step involving F_ℓ is unsupported. A substantially revised version that restricts or corrects the ℓ=1 statement, treats the complementary parities, and supplies the missing localization estimate could be reconsidered."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real news: for ℓ≥2 the paper derives a concrete double-integral constant Kℓ and shows it exceeds 1, so palindromic blocks genuinely increase the expected zero density over the i.i.d. case. That is a new, citable benchmark, and the proof machinery — Riemann-sum approximation plus bounded-variation estimates for oscillatory integrands — is substantive and looks coherent for the regime it actually covers. No fitted parameters; the table is honest numerical evaluation. The two lemmas imported from the author's preprint are auxiliary estimates, so the self-citation is not a problem.\n\nThe soft spots are in the gap between the theorem statement and the proof. Equation (1) in the introduction correctly says palindromic coefficients (ℓ=1) have n deterministic zeros, so the expected count is n + n/√3, not n/√3. But Theorem 2.1 with K1=1/2 gives n/√3, and the proof even says \"allowing ℓ=1 gives us n/√3.\" That is a direct contradiction, not a minor typo. The theorem should either exclude ℓ=1 or handle the deterministic-factor case separately.\n\nFor ℓ≥2, the proof assumes n−ℓ is odd to get the π/2 symmetry that turns E[N(0,π)] into 2E[N(0,π/2)], and later assumes n is even for the Riemann-sum split. For even ℓ those two assumptions cannot both hold; for odd ℓ they restrict to one parity of n. The theorem states a full O(n^{2/3}) asymptotic for all allowed n, but the proof never addresses the complementary parity. The \"for simplicity\" sentences acknowledge the restrictions in passing but never reconcile them with the stated result.\n\nNone of this kills the likely truth of the ℓ≥2, covered-parity theorem. But as written, the central claim is not supported, and the ℓ=1 case is false. The paper deserves a serious referee — it is not a desk reject — but the referee should send it back for major revision: fix or exclude ℓ=1, and either prove the missing parity or restrict the theorem to the parity actually treated.","headline":"New explicit constant for palindromic-block cosine polynomials, but Theorem 2.1 overclaims: false at ℓ=1 and only proved on a parity subsequence.","tokens_in":22239,"tokens_out":29293,"would_cite":false,"duration_ms":258122,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["30C15","26C10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Random cosine polynomials with coefficients arranged in palindromic blocks of fixed length $\\ell$ have expected real-zero count $\\frac{2n}{\\sqrt3}K_\\ell+O(n^{2/3})$, with $K_\\ell>1$ for $\\ell\\ge2$.","keywords":["random cosine polynomials","expected real zeros","palindromic blocks","dependent coefficients","Kac-Rice formula","Gaussian coefficients","asymptotic expansion"],"falsifier":"For a fixed $\\ell$ (say $\\ell=2$), evaluate the Kac-Rice integral numerically for $n$ with $n-\\ell$ odd and with $n-\\ell$ even, and compare both against $(2n/\\sqrt3)K_2+O(n^{2/3})$; if the even-parity sequence differs by more than the stated error, the theorem's claim for all $r$ fails. Independently, for $\\ell=1$, count all real zeros of the palindromic-coefficient polynomial; if the leading term is $n+n/\\sqrt3$ rather than $n/\\sqrt3$, the theorem's $\\ell=1$ reading omits the deterministic zeros.","tokens_in":21075,"feed_emoji":"📈","tokens_out":9787,"duration_ms":252662,"temperature":0.7,"pith_summary":"This paper studies a random cosine polynomial $V_n(x)=\\sum_{j=0}^n a_j\\cos(jx)$ whose Gaussian coefficients are no longer independent: they are grouped into palindromic blocks of a fixed length $\\ell$, so each block is repeated in reverse order on the other side of the coefficient sequence. The central result is that the expected number of real zeros in $(0,2\\pi)$ is $\\frac{2n}{\\sqrt3}K_\\ell+O(n^{2/3})$, where $K_\\ell$ is an explicit double integral that depends only on $\\ell$ and is strictly larger than 1 for every $\\ell\\ge2$. This means forcing palindromic-block correlation into the coefficients strictly increases the expected zero count compared with the classical i.i.d. case, by about 6.4% for $\\ell=2$ and by smaller amounts as $\\ell$ grows. The paper makes the constant $K_\\ell$ computable, so the prediction is concrete and testable for each block length.","feed_headline":"Palindromic blocks add real zeros to random cosine polynomials","feed_subtitle":"Expected zeros scale as (2n/√3)Kℓ, with Kℓ>1, shrinking toward the classical count as ℓ grows.","key_machinery":"The argument runs through the Kac-Rice formula, which expresses the expected number of real zeros of a Gaussian random function as $\\frac1\\pi\\int \\sqrt{AC-B^2}/A\\,dx$ in terms of the coefficient-level sums $A,B,C$. The key object is the block-ratio function $u_\\ell(s)=\\sin(\\ell s)/(\\ell\\sin s)$, which encodes the palindromic-block structure: after trigonometric simplification, the quantities $A(x)$, $B(x)$, and $C(x)$ for the block polynomial are approximated by $n$-scale expressions built from $u_\\ell$, leaving the integrand $\\sqrt{1+\\frac{3(1-u_\\ell(x)^2)}{(1+u_\\ell(x)\\cos(nx))^2}}$ on most of $[0,\\pi/2]$. The final step replaces the rapidly oscillating $\\cos(nx)$ by a Riemann sum over intervals of length $\\pi/n$, using a modified-sawtooth error estimate, to obtain the double-integral constant $K_\\ell$; the remaining terms are controlled by comparing with the known behavior of i.i.d. cosine polynomials near the endpoints.","core_discovery":"For a fixed block length $\\ell$, write $n=2\\ell m+r$ with $m\\in\\mathbb N$ and $r\\in\\{-1,\\dots,2\\ell-2\\}$, and let the coefficient vector consist of $2m$ blocks of length $\\ell$ arranged so that the first $m$ blocks are repeated in reverse order by the last $m$ blocks, with the leftover $r+1$ coefficients i.i.d. Gaussian. The paper's Theorem 2.1 asserts that the expected number of real zeros of $V_n$ in $(0,2\\pi)$ satisfies $E[N_n(0,2\\pi)] = \\frac{2n}{\\sqrt3}K_\\ell + O(n^{2/3})$, where $K_\\ell=\\frac{1}{\\pi^2}\\int_0^\\pi\\int_0^{\\pi/2}\\sqrt{1+\\frac{3(1-u_\\ell(s)^2)}{(1+u_\\ell(s)\\cos t)^2}}\\,ds\\,dt$ and $u_\\ell(s)=\\frac{\\sin(\\ell s)}{\\ell\\sin s}$. The constant is strictly greater than 1 for $\\ell\\ge2$, so palindromic blocks produce more expected real zeros than independent coefficients, and the size of the excess is read off from $K_\\ell$. For $\\ell=1$ the formula gives $K_1=1/2$ and a leading term $n/\\sqrt3$, which the paper connects to the earlier palindromic-coefficient result, with the caveat that the fully palindromic case carries $n$ additional deterministic zeros recorded in equation (1).","pith_inferences":["The theorem's $\\ell=1$ special case, taken literally, gives a leading term $n/\\sqrt3$ for fully palindromic coefficients, but the paper's own equation (1) says the actual expected count is $n+n/\\sqrt3$ because of deterministic zeros; a unified reading is that the double-integral formula captures only the random zeros, and a version covering $\\ell=1$ would need an extra deterministic $n$ term.","Because the proof assumes $n-\\ell$ is odd, the most direct check of the full theorem is to test the even-parity subsequence numerically; if it shows the same constant $K_\\ell$, a symmetric argument for the missing parity likely exists, and if not, the theorem as stated is too strong.","The numerical decrease of $K_\\ell$ toward 1 suggests a quantitative conjecture outside the paper: $K_\\ell-1$ should decay like a power of $1/\\ell$, and its leading exponent could be extracted from the double-integral formula."],"forward_implications":["For every fixed $\\ell\\ge2$, the expected zero count of these dependent polynomials is asymptotically $K_\\ell$ times the classical $2n/\\sqrt3$, with $K_\\ell>1$; for $\\ell=2$ and $\\ell=3$ the numerical constants are about 1.0642 and 1.0408.","The constant $K_\\ell$ is explicitly computable as a double integral, so the predicted excess can be compared with simulations or with exact Kac-Rice evaluations for any chosen block length.","If the theorem is correct, the classical $2n/\\sqrt3$ is not just the i.i.d. answer but a lower benchmark for this family of dependent structures: palindromic blocks raise it, while earlier pairwise-equal-block constructions keep it unchanged.","The factor $K_\\ell$ decreases toward 1 as $\\ell$ grows (the table gives $K_{2019}\\approx1.000046$), meaning long palindromic blocks behave almost like the independent-coefficient case."],"supporting_citations":[{"why":"Supplies the classical baseline: i.i.d. Gaussian cosine polynomials have $E[N_n(0,2\\pi)]=2n/\\sqrt3+O(n^{11/13}(\\log n)^{3/13})$, which the new result multiplies by $K_\\ell$.","marker":"[7]"},{"why":"Provides the palindromic-coefficient case ($\\ell=1$) that the block construction generalizes, with the $n+n/\\sqrt3$ expected-zero count once deterministic zeros are included.","marker":"[13]"},{"why":"Gives the pairwise-equal-blocks setting and supplies Lemmas 3.1 and 3.2 used to bound contributions away from the main interval.","marker":"[23]"},{"why":"States the Kac-Rice formula in the form used to turn expected zeros into an integral of $\\sqrt{AC-B^2}/A$.","marker":"[22]"},{"why":"Supplies the integral identities (3.613(1), 1.341, 2.554) used repeatedly to evaluate trigonometric integrals in the proof.","marker":"[14]"},{"why":"Provides the Riemann-sum convergence method used to replace the oscillatory integral by the double-integral constant $K_\\ell$.","marker":"[5]"},{"why":"Gives Markov's inequality bounding $|u_\\ell(x)|\\le1$, which controls the integrand away from the endpoints.","marker":"[25]"},{"why":"Supplies Jensen's inequality used to show $K_\\ell>1$ for $\\ell\\ge2$.","marker":"[27]"}],"fun_headline_variants":["Palindromic blocks boost real zeros in random cosine polynomials","Block palindromes raise expected real zeros beyond i.i.d. case","Kℓ>1: palindromic blocks add to random cosine zero count","Palindromic coefficient blocks yield extra expected real zeros"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the main theorem assumes $n-\\ell$ is odd and never supplies an argument for the opposite parity, while the theorem is stated for all $n=2\\ell m+r$; the $\\ell=1$ case also drops the $n$ deterministic zeros that the paper itself acknowledges in equation (1).","fun_headline_variants_meta":{"raw":{"variants":["Palindromic blocks boost real zeros in random cosine polynomials","Block palindromes raise expected real zeros beyond i.i.d. case","Kℓ>1: palindromic blocks add to random cosine zero count","Palindromic coefficient blocks yield extra expected real zeros"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000301,"raw_usage":{"total_tokens":1829,"prompt_tokens":1134,"completion_tokens":695,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":750,"completion_tokens_details":{"reasoning_tokens":622}},"tokens_in":750,"tokens_out":695,"duration_ms":7222,"temperature":1.0,"reasoning_tokens":622,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:51:29.756156+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a fixed $\\ell$ (say $\\ell=2$), evaluate the Kac-Rice integral numerically for $n$ with $n-\\ell$ odd and with $n-\\ell$ even, and compare both against $(2n/\\sqrt3)K_2+O(n^{2/3})$; if the even-parity sequence differs by more than the stated error, the theorem's claim for all $r$ fails. Independently, for $\\ell=1$, count all real zeros of the palindromic-coefficient polynomial; if the leading term is $n+n/\\sqrt3$ rather than $n/\\sqrt3$, the theorem's $\\ell=1$ reading omits the deterministic zeros.","supporting_citations":[{"cited_title":"Dunnage, The number of real zeros of a random trigonometric polynomia l, Proc","cited_arxiv_id":null,"evidence_quote":"Supplies the classical baseline: i.i.d. Gaussian cosine polynomials have $E[N_n(0,2\\pi)]=2n/\\sqrt3+O(n^{11/13}(\\log n)^{3/13})$, which the new result multiplies by $K_\\ell$."},{"cited_title":"Farahmand and T","cited_arxiv_id":null,"evidence_quote":"Provides the palindromic-coefficient case ($\\ell=1$) that the block construction generalizes, with the $n+n/\\sqrt3$ expected-zero count once deterministic zeros are included."},{"cited_title":"Real zeros of random trigonometric polynomials with pairwise equal blocks of coefficients","cited_arxiv_id":"1905.13349","evidence_quote":"Gives the pairwise-equal-blocks setting and supplies Lemmas 3.1 and 3.2 used to bound contributions away from the main interval."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the Kac-Rice formula in the form used to turn expected zeros into an integral of $\\sqrt{AC-B^2}/A$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the integral identities (3.613(1), 1.341, 2.554) used repeatedly to evaluate trigonometric integrals in the proof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Riemann-sum convergence method used to replace the oscillatory integral by the double-integral constant $K_\\ell$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives Markov's inequality bounding $|u_\\ell(x)|\\le1$, which controls the integrand away from the endpoints."},{"cited_title":"Rudin, Real and Complex Analysis , McGraw-Hill (1974)","cited_arxiv_id":null,"evidence_quote":"Supplies Jensen's inequality used to show $K_\\ell>1$ for $\\ell\\ge2$."}],"review_version":1}