{"id":"a075fe5d-3f29-4fbd-853e-4b7aea6869ad","arxiv_id":"2607.25023","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"A robust OR polynomial yields Schrijver quasi-tensorization ϑ'(⊠Gi)≤∏ϑ'(Gi)^{C log r log ϑ'(Gi)} and improves Rr(k) to exp(−Ω(k/(r^9(log r)^6))) r^{rk}.","lead":"The paper builds a robust OR polynomial that stitches one-coordinate PSD certificates into OR-type constraints, then uses it for two bounds. It quasi-tensorizes the Schrijver number and sharpens the r-dependence in multicolor Ramsey upper bounds.","discovery_kind":"new_method","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"No significant objection identified. Both main proofs verify internally; the one genuine load-bearing input is the external BBC+26 moment-positivity lemma (reader's pick). A minor, repairable bookkeeping slip exists in the Lemma 3.3 verification chain but does not touch the conclusion.","rationale":"The reader's verdict (ACCEPT, high confidence) matches my own reading. The paper's central claims rest on three pillars: (i) the internal robust-OR-polynomial construction, which I verified line by line (Lemma 1.6 appendix proof, Lemma 2.2 sinc-quadrature approximation with correct variation and truncation bounds, PSD composition argument, degree bookkeeping); (ii) the external BBC+26 moment-positivity lemma, correctly identified by the reader as the weakest assumption — but it is a refereed, properly isolated input used in the same way as in the source, so it does not raise correctness risk beyond \"the cited literature is correct\"; (iii) the parametrized rerun of the BBC+26 book algorithm (Lemma 3.3/A.2), where the only flaw I found is a single false intermediate inequality in the final parameter check whose endpoint conclusion nevertheless holds directly. Lemma A.1 (parametrized Lemma 2.2 of BBC+26) is stated without proof, but its derivation transfers verbatim since C_r, β_r enter only through the geometric lemma's conclusion. The AI-use disclosure concerns exactly this bookkeeping, and my independent check of the key inequalities (page-set bound, reservoir bound, threshold choices) found them consistent apart from the noted slip. No ad hominem concerns, no circularity beyond the acknowledged citation chain, no manufactured objections: this is a clean methods paper with a correctable presentation bug. Verdict unchanged.","tokens_in":23048,"tokens_out":12408,"duration_ms":391022,"concrete_test":"Verify the Lemma A.3 hypothesis numerically/symbolically with the paper's own parameters: with β_r = (4A_r(1+rC_r))^{-1}, A_r = 2r^r, C_r = C_1 r log r, μ_r = A·K_1 r²(log 2r)², p_r = 1/r − 2ε_r ∈ [1/(2r), 1], check that log(r/β_r) ≤ μ_r/(16 log(μ_r²/p_r)) holds for all r ≥ 2 (e.g., sweep r = 2..10^4 in arbitrary precision, plus the asymptotic comparison C_4 r log(2r) vs. (A K_1/160) r² log(2r)). If it holds for explicit A, K_1, the write-up needs only a one-line fix (drop the false intermediate step through M_r log(2r)); if it failed, the k-threshold k ≥ C_0 r^{14}(log 2r)^{12} would need to grow and Theorem 1.5's exponent range would weaken.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I traced both proofs end to end and could not find a load-bearing flaw. Theorem 1.1: the dual-certificate Lemma 2.1 is elementary and correct; the compressor T_i(x)=(1+xF_i(x²)/H_i(x²))^L−1 lands in [−1,−1+(1/8)^L] on nonedge coordinates because Z_j(x,y)≤−1 there and |Z_j(x,y)|≤2λ_j−1 follows from the 2×2 minor of the PSD matrix Z_j−J (valid for λ_j≥2, matching the ϑ′≥2 hypothesis); even L keeps T_i≥−1 on all other coordinates, which is all Lemma 1.6 needs; the nonnegative-coefficient expansion (8) plus Schur's theorem gives K−γJ⪰0; the degree bookkeeping D_i≤L(2d_i+1)=O(log r·log ϑ′(G_i)) yields the stated exponent. Lemma 1.3: the Bessel compressor's Taylor nonnegativity, the [−1,−1+(8er)^{-1}] compression for x≤−1 (Fact 3.1, standard J_0 decay), and the exp(O(log r)√(x+1)) growth for x≥−1 all check out, and the layer-cake contradiction with C′_r=C_r+1 is correct, including the λ=−1 escape case where the event is E itself for every i. The genuinely load-bearing external input is BBC+26 Lemma 3.2 moment positivity E[∏Z_i^{a_i}]≥0, used verbatim to get EP(Z)≥0 (termwise swap is fine since X is finite so Z is bounded). This is a refereed JAMS input, properly isolated, used exactly as in the source paper — so it is a dependence, not a weakness. One real but minor slip: in the proof of Lemma 3.3 the displayed chain \"log(r/β_r) ≤ M_r log(2r) ≤ μ_r/(16 log(μ_r²/p_r))\" has a false second inequality (its justification lower-bounds μ_r²/p_r where an upper bound is needed; numerically the middle term exceeds the right side by a factor ~log²(2r)). However the endpoint inequality needed for Lemma A.3 holds directly, since log(r/β_r)=O(r log r) is far below μ_r/(16 log(μ_r²/p_r))=Θ(r² log r). Fixable in one line; does not affect Theorem 1.5.","agreement_with_reader":"agree"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper introduces a \"robust OR polynomial\" Q_r(t)=∏(1+t_i)−∏(1+(1−1/r)t_i), which has nonnegative coefficients and detects the event that some coordinate lies in [−1,−1+η], and uses it to compose one-coordinate PSD certificates across OR constraints via Schur products. Two applications are given. First (Theorem 1.1), a quasi-tensorization theorem for the Schrijver number: for graphs with ϑ′(G_i)≥2, ϑ′(⊠_i G_i) ≤ ∏_i ϑ′(G_i)^{C log r·log ϑ′(G_i)}, proved by building a dual-feasible kernel from the one-coordinate Schrijver dual solutions passed through a sinc-quadrature sign compressor (Lemma 2.2) and composed with Q_r. This yields quasipolynomial bounds M_r(n) ≤ (2n)^{C_0 r log r log(2n)} for r-way acute-free families (Corollary 1.2). Second (Lemma 1.3, Theorem 1.5), an improvement of the geometric lemma of Balister et al. (BBC+26), replacing C_r=Θ(r^{3/2}) by C_r=O(r log r) via a Bessel-function compressor with nonnegative Taylor coefficients and exp(O(log r)√(x+1)) growth; fed into the unchanged BBC+26 book algorithm, this gives R_r(k) ≤ exp(−ck/(r^9(log 2r)^6)) r^{rk} for k ≥ C_0 r^{14}(log 2r)^{12}, improving the previous exp(−Ω(k/r^{12}))r^{rk}.","tokens_in":23553,"tokens_out":11364,"duration_ms":374832,"significance":"If correct, the results are strong. The improvement of the multicolor Ramsey exponent from r^{12} (a JAMS breakthrough from 2024/26) to r^9(log r)^6 is substantial progress on a high-profile problem, and it is achieved by a clean modular substitution: the book algorithm is untouched, only the geometric input is improved. The Schrijver quasi-tensorization theorem appears to be the first general statement of its kind and the quasipolynomial bound on M_r(n) answers (up to the exponent) a problem circulated by the BBC+26 authors. Strengths worth naming: the proofs are complete and largely self-contained; the one external load-bearing input (moment positivity, BBC+26 Lemma 3.2) is properly isolated and used verbatim; the certificate construction is parameter-free in the sense that all constants are absolute and nothing is fitted; and the framework itself (compressor + nonnegative-coefficient OR polynomial + Schur products) is a reusable method likely to find further applications. I verified the main chains independently: the dual-certificate Lemma 2.1, the compressor bounds in Theorem 1.1 (including the 2×2-minor bound |Z_j(x,y)| ≤ 2λ_j−1 that requires ϑ′≥2), the expansion (8) giving K−","major_comments":[{"comment":"Proof of Lemma 3.3 (Appendix A.2, p. 24) and Lemma A.3 hypothesis (p. 21): the verification of the hypothesis of Lemma A.3 contains wrong-direction inequalities, and as stated Lemma 3.3's hypotheses do not imply Lemma A.3's condition. Specifically: (i) Lemma A.3 assumes log(r/β_r) ≤ μ_r/(16 log(μ_r²/p_r)), but the natural quantity the proof actually uses is μ_r·log(1/δ_r)/16 with the logarithm in the numerator — note log(1/δ_r)=log(μ_r²/p_r); the subsequent displayed bounds log(r/β_r) ≤ μ_r/(16 log(1/δ_r)) and C_r√(λ_{0,r}+1) ≤ √2 μ_r/(8 log(1/δ_r)) both have log(1/δ_r) in the denominator, whereas direct computation from λ_{0,r}=(μ_r log(1/δ_r)/(8C_r))² gives C_r√(λ_{0,r}+1) ≤ √2 μ_r log(1/δ_r)/8, with the logarithm in the numerator. (ii) Consequently, in Lemma 3.3 the chain log(r/β_r) ≤ M_r log(2r) ≤ μ_r/(16 log(μ_r²/p_r)) has a false second inequality: with μ_r=AM_r and p_r≥1/(2r), log","section":"§A.2, proof of Lemma 3.3 and Lemma A.3"}],"minor_comments":[{"comment":"Notation collision: the letter A denotes the absolute Bessel cutoff of Fact 3.1 (used throughout §3.1) and is reused as the large absolute constant in the proof of Lemma 3.3 (p. 23). Please rename one of them.","section":"§3, Fact 3.1 vs. proof of Lemma 3.3"},{"comment":"In the proof of Theorem 1.1, the identification λ_i = ϑ′(G_i) (dual optimum equals primal) is used implicitly when concluding (2λ_i)^{D_i} ≤ ϑ′(G_i)^{C log r log ϑ′(G_i)}; a one-line reminder that strong duality holds for the Schrijver SDP [Sch79] would help readers. Similarly, the step |Z_j(x,y)| ≤ 2λ_j−1 from Z_j−J ⪰ 0 via the 2×2 principal minor deserves one displayed line, since it is where the hypothesis ϑ′(G_i) ≥ 2 enters.","section":"§2, proof of Theorem 1.1"},{"comment":"In the proof of Lemma A.3, the displayed equality η_r^{2rt_r} ≥ exp(−μ_r r t_r/(2 log(1/δ_r))) = (μ_r²/p_r)^{−μ_r r t_r/2} has a false equality sign (exp(−a/log b) ≠ b^{−a}); the correct relation is '≥' via the numerator computation — see major comment. Please correct the display.","section":"§A.2, proof of Lemma A.3"},{"comment":"Conjecture 4.1 is said to be 'supported by numerical evidence' (correct order of M_2(n) seems to be n^{2+o(1)}), but no data or description of the computation is given. Either include a brief description or soften the claim.","section":"§4, Conjecture 4.1"},{"comment":"The abstract states the Ramsey bound with (log r)^6 while Theorem 1.5 uses (log(2r))^6; harmless, but please harmonize. Also, the absolute constants C, C_0, c in Theorems 1.1/1.5 are not estimated; a remark on whether the methods give explicit (if large) values would be useful.","section":"Abstract / §1"},{"comment":"The hypothesis ϑ′(G_i) ≥ 2 of Theorem 1.1 excludes graphs with ϑ′<2 (e.g., complete graphs); a short remark that such factors can be handled separately (or absorbed into the product bound) would preempt a natural reader question.","section":"§1, Theorem 1.1 statement"}],"recommendation":"minor_revision","confidential_remarks":"The improvement over the recent JAMS result of Balister et al. is genuine and the paper is well within scope. The authors disclose that ChatGPT 5.5 Pro was used to verify proofs, 'especially the parameter bookkeeping in the proof of Lemma 3.3'; it is worth noting to the authors (not as a criticism) that the one real error I found is precisely in that bookkeeping chain — the wrong-direction log inequalities flagged in my major comment. The core arguments (quasi-tensorization, Bessel compressor, layer-cake extraction) I verified by hand and believe to be correct. No concerns about citation practice or novelty; the dependence on BBC+26 is explicit and appropriately credited."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The thing worth knowing is that they built a reusable composition tool—a robust OR polynomial with nonnegative coefficients, paired with sign/Bessel compressors—and used it for two genuine advances: the first general quasi-tensorization of the Schrijver number under strong products, and a sharpened geometric lemma that improves the r-dependence in the BBC+26 Ramsey bound from r^12 to r^9 (log r)^6.\n\nWhat is new is the framework itself. ϑ' is tight on the hypercube graph but not multiplicative; they bypass that by dual certificates, a sinc-quadrature rational compressor that lands non-edges near -1, and the OR polynomial (5) whose nonnegative coefficients keep the Schur product PSD. The degree bookkeeping is clean and yields the stated quasipolynomial Mr(n) bound. On the Ramsey side they replace the earlier test function with a Bessel compressor whose growth is only exp(O(log r)\\sqrt(x+1)), compose with the same OR polynomial, and feed the improved Cr = O(r log r) straight into the unchanged book algorithm. Both proofs check end-to-end; the only external load-bearing input is the BBC+26 moment-positivity lemma, which is properly isolated and already refereed.\n\nSoft spots are minor and proportionate. Absolute constants are non-explicit (as usual). There is a one-line bookkeeping slip in the verification chain for Lemma 3.3—the displayed inequality log(r/\\beta r) \\le Mr log(2r) \\le µr/(16 log(µr^{2}/pr)) has a false second step—but the endpoint inequality needed for the argument holds directly by a larger margin, so Theorem 1.5 is unaffected. The Ramsey half is an incremental improvement of a 2026 breakthrough rather than a new paradigm; the acute-free half is the more conceptual contribution. Citation pattern is appropriate.\n\nThis is for people who work on SDP bounds for codes, spherical codes, or multicolor Ramsey. A serious referee will want to see the constants cleaned and the bookkeeping line fixed, but the math is solid enough to send out. I would engage with it and would cite the quasi-tensorization result.","headline":"Clean methods paper: a reusable nonnegative OR polynomial + compressors gives the first quasi-tensorization of Schrijver numbers and a real r-exponent improvement on the BBC+26 multicolor Ramsey bound.","tokens_in":24653,"tokens_out":556,"would_cite":true,"duration_ms":10674,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C55","05D10","90C22","05C69"],"pacs":[],"model":"grok-4.5","headline":"A single polynomial recipe turns one-coordinate PSD certificates into OR certificates, yielding quasi-tensorization of the Schrijver number and tighter multicolor Ramsey bounds.","keywords":["Schrijver number","strong product","acute-free families","multicolor Ramsey numbers","OR polynomials","positive semidefinite kernels","geometric lemma","book algorithm"],"falsifier":"Exhibit maps σ1,…,σr for which some nonnegative multi-index has negative inner-product moment, or compute ϑ′ of an explicit strong product of small Schrijver graphs and check whether it exceeds the claimed product-of-powers bound.","tokens_in":24119,"feed_emoji":"📐","tokens_out":1032,"duration_ms":21817,"temperature":0.7,"pith_summary":"Many extremal problems ask that at least one of several coordinates satisfies a bad event. One-coordinate versions often have clean positive-semidefinite certificates, but those certificates do not automatically compose under an OR. This paper builds a robust OR polynomial with nonnegative coefficients that detects a bad coordinate while preserving positive-semidefiniteness after composition with compressors. Applied to the Schrijver number, the method gives a general quasi-tensorization bound under strong products, which controls the size of r-way acute-free families of vectors by a quasipolynomial in n. The same tool improves the geometric input to an existing book algorithm for multicolor Ramsey numbers, shaving the r-dependence in the exponential error term. A sympathetic reader cares because the method turns a structural obstruction—non-multiplicativity of a tight SDP bound—into a usable approximate product rule and feeds a better local lemma into a known global algorithm.","feed_headline":"OR polynomials quasi-tensorize Schrijver and cut Ramsey r-loss","feed_subtitle":"One nonnegative polynomial turns one-coordinate PSD certificates into product bounds and a sharper geometric lemma.","key_machinery":"The robust OR polynomial Qr(t) = ∏(1+ti) − ∏(1+(1−1/r)ti), whose nonnegative coefficients detect a coordinate near −1 and, by Schur products, keep composed kernels positive semidefinite.","core_discovery":"There is a robust OR polynomial, built from elementary symmetric sums with nonnegative coefficients, that becomes strictly negative whenever any input lies near −1 while remaining controlled when all inputs stay at least −1. Composed with sign or Bessel compressors that map one-coordinate PSD kernels into that range, it produces dual feasible kernels for strong products and nonnegative-expectation test functions for vector-valued maps. Consequently the Schrijver number of a strong product is at most a product of the factors raised to C log r log ϑ′, and the multicolor Ramsey number satisfies Rr(k) ≤ exp(−Ω(k/(r9 (log r)6))) r^{rk} for large enough k.","pith_inferences":["If the degree of the sign compressor can be reduced below log ϑ′, the quasi-tensorization exponent could drop from quasipolynomial toward the pure product form conjectured for two factors.","The same OR polynomial may give approximate multiplicativity for other non-multiplicative SDP hierarchy numbers that differ from Lovász theta only by nonnegativity constraints.","Numerical checks that M2(n) is n^{2+o(1)} would support attacking the two-factor Schrijver conjecture via a tighter Delsarte LP rather than further polynomial engineering."],"forward_implications":["r-way acute-free families on the sphere or hypercube have size at most (2n)^{O(r log r log(2n))}, closing most of the gap from the trivial (2n)^r lower bound to a quasipolynomial.","Schrijver numbers of strong products admit a uniform quasi-multiplicative upper bound whenever each factor has ϑ′ ≥ 2.","The geometric lemma feeding the multicolor book algorithm improves from Cr = Θ(r^{3/2}) to Cr = O(r log r), yielding Rr(k) ≤ exp(−Ω(k/(r^9 (log r)^6))) r^{rk}.","The same compressor-plus-OR template applies to any OR-type extremal problem that already possesses one-coordinate PSD or moment certificates."],"fun_headline_variants":["Robust OR polynomials quasi-tensorize Schrijver number","OR framework gives Schrijver product bounds and sharper Ramsey","Quasi-tensorized Schrijver via OR polynomials cuts Ramsey r-loss","OR polynomials bound acute-free families and multicolor Ramsey","Schrijver quasi-tensorization sharpens R_r(k) r-dependence"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The Ramsey improvement rests on an inherited fact that every monomial moment of the inner-product coordinates is nonnegative, so any test function with nonnegative Taylor coefficients has nonnegative expectation.","fun_headline_variants_meta":{"raw":{"variants":["Robust OR polynomials quasi-tensorize Schrijver number","OR framework gives Schrijver product bounds and sharper Ramsey","Quasi-tensorized Schrijver via OR polynomials cuts Ramsey r-loss","OR polynomials bound acute-free families and multicolor Ramsey","Schrijver quasi-tensorization sharpens R_r(k) r-dependence"]},"model":"grok-4.5","effort":"low","cost_usd":0.005838,"raw_usage":{"total_tokens":1776,"prompt_tokens":1150,"num_sources_used":0,"completion_tokens":79,"cost_in_usd_ticks":58384000,"prompt_tokens_details":{"text_tokens":1150,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":547,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":1150,"tokens_out":79,"duration_ms":8151,"temperature":1.0,"reasoning_tokens":547,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T03:14:00.368726+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Exhibit maps σ1,…,σr for which some nonnegative multi-index has negative inner-product moment, or compute ϑ′ of an explicit strong product of small Schrijver graphs and check whether it exceeds the claimed product-of-powers bound.","supporting_citations":[],"review_version":1}