{"id":"6835e243-972d-4329-8ae4-2bb5edbb91df","arxiv_id":"2606.13056","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Any orthogonal polynomial family defined by Favard's theorem satisfies Q_{p+s}(x)=M(x)Q_p(x)+N(x)Q_{p-t}(x), with coefficients built from the standard recurrence coefficients.","lead":"This paper derives three-term recurrence formulas that jump by arbitrary step sizes for any family of orthogonal polynomials defined by the standard Favard recurrence. It gives explicit skip-by-2 formulas for Hermite, Gegenbauer, and Legendre polynomials and tests their numerical accuracy.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4's universal statement fails at zeros shared by Qp and Qp-t; paper's bias patch (67) forfeits the three-term form.","rationale":"The reader's verdict is CONDITIONAL, and the concern identified here is essentially the same one: the main theorem's validity depends on non-vanishing of a denominator, and the paper's special-case patches are not fully justified. The Legendre counterexample strengthens the reader's concern because it shows a concrete instance where Eq. (55) cannot hold with finite coefficients, even though the paper acknowledges this in §3.4 and Remark 3. The central derivation is algebraically sound on a dense open set, and the explicit formulas in Section 5 check out, so the appropriate disposition is not rejection but a qualified theorem statement. Thus the existing CONDITIONAL verdict is the right one, and no change is needed.","tokens_in":21084,"tokens_out":8993,"duration_ms":87447,"concrete_test":"Compute the Legendre example singled out in Remark 3: p=5, s=3, t=2, x=0. Using Eq. (15) or explicit values, verify L8(0)=35/128, L5(0)=0, L3(0)=0. Then Eq. (55) would require 35/128 = M(0)·0 + N(0)·0, impossible for finite M,N. This directly falsifies the universal form of Theorem 4. A secondary check: re-derive Eq. (55) symbolically for this case; the denominators are zero, so the theorem needs an x-restricted statement or a bias term to be correct.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 4 (Eq. 55) asserts that every Favard-defined family satisfies Q_{p+s}(x)=M(x)Q_p(x)+N(x)Q_{p-t}(x) for all 1≤t≤p, s≥1. The derivation in (52)–(56) eliminates Q_{p-1} and therefore requires ψ^{p-1}_{p-t}(x)≠0. The paper treats vanishing cases in §3.4, but the resolution is not a three-term recurrence: when Q_p(x)=Q_{p-t}(x)=0 while Q_{p+s}(x)≠0, no finite M,N can exist, and (67) introduces a bias term b(x) with arbitrary l1,l2. This is not a removable singularity. The example in Remark 3 — Legendre, p=5, s=3, t=2, x=0 — is exactly such a case: L3(0)=L5(0)=0, L8(0)≠0. Since Theorem 4 as stated has no exclusion clause, it is literally false at such points. The subsequent assertion in (66) that Q_{p+s}(x)≠0 is not proved and is not needed for the counterexample. The core algebraic identity is valid on the dense set where the denominators are nonzero and where Q_p and Q_{p-t} do not both vanish, so the paper can be repaired by qualifying the theorem; but the abstract claim that any two members suffice for every other member is too strong.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a generalization of the classical three-term recurrence relation for orthogonal polynomial sequences defined by Favard's theorem. For any Favard-defined family {Q_p}, the author derives expressions for Q_{p+s} and Q_{p-t} in terms of adjacent members Q_p, Q_{p-1}, then eliminates Q_{p-1} to obtain a three-term recurrence involving Q_{p+s}, Q_p, Q_{p-t} with coefficient functions M(x), N(x). The paper also treats degree-decrease and end-to-middle directions, gives explicit two-degree-step recurrences for Hermite, Gegenbauer, and Legendre polynomials, and reports numerical precision comparisons against the standard recurrences.","tokens_in":21455,"tokens_out":16354,"duration_ms":142258,"significance":"If the main theorem is stated with the correct qualifications, the paper provides a useful algebraic technique for generating degree-skipping recurrences with explicit rational coefficients for any Favard-defined family. The explicit recurrences for s=t=2 in Section 5 are concrete and checkable, and the numerical experiments support the practical viability of the proposed formulas. However, the central claim as stated in Theorem 4 and the abstract is overreaching: the rigid three-term form does not hold at points where the elimination denominators vanish and Q_p, Q_{p-t} both vanish while Q_{p+s} does not. The paper itself acknowledges this in Remark 3 of Section 4.3 and attempts to patch it in Section 3.4 using bias terms and auxiliary inputs, which are not part of the stated theorem. The algebraic core is sound on the dense set where the coefficients are defined, but the universal formulation must be corrected.","major_comments":[{"comment":"Theorem 4 claims that every Favard-defined family satisfies Q_{p+s}=M(x)Q_p(x)+N(x)Q_{p-t}(x) for all 1≤t≤p, s≥1, with no exclusions. The derivation via (54) requires Ψ^{(p-1)}_{p+s}(x)≠0 and ψ^{(p-1)}_{p-t}(x)≠0. When ψ^{(p-1)}_{p-t}(x)=0 and Q_p(x)=Q_{p-t}(x)=0 but Q_{p+s}(x)≠0, no finite M,N satisfying (55) can exist. The paper's own §3.4, specifically (66)-(67), replaces the three-term relation with a bias expansion Q_{p+s}=l1 Q_p + l2 Q_{p-t} + b(x) with arbitrary l1,l2, which is not a three-term recurrence. The third remark in §4.3 explicitly gives the Legendre example p=5,s=3,t=2,x=0, where L3(0)=L5(0)=0 and L8(0)≠0, and states that no rigid three-term relation exists. Thus Theorem 4 is literally false as stated. The theorem must be restricted to points where the denominators are nonzero, or reformulated as a rational-function identity with exceptional cases handled separately.","section":"Theorem 4, Eqs. (55)-(56); §3.4; §4.3 third remark"},{"comment":"The coefficient recursion in Theorem 3 contains an index error. The first entry of the second row is printed as ψ^{(p)}_{p-t-1}(x), but the correct entry is ψ^{(p)}_{p-t+1}(x). Specializing the printed formula to t=3 gives ψ^{(p)}_{p-3}(x)=ψ^{(p)}_{p-4}(x) T_{p-2}(x), which contradicts the directly derived expression (42), ψ^{(p)}_{p-3}(x)=-D_{p-1}(x)T_{p-2}(x). Since (46) is the general recipe for computing coefficients for arbitrary t, this must be corrected.","section":"Eq. (46)"},{"comment":"In the special case ψ^{(p-1)}_{p-t}(x)=0, the proposed resolution computes Q_{p-1}(x) using Q_0(x) or Q_1(x) as extra inputs (61)/(63), and in the subcase Q_{p-t}(x)=0 it introduces a bias term with arbitrary coefficients (67). Consequently, the recurrence no longer computes Q_{p+s} strictly from Q_p and Q_{p-t}, contrary to the claim in the abstract and Section 1.2 that 'any two members in the sequence' suffice for computing the other member. The theorem and abstract need to state these exceptional cases explicitly as limitations.","section":"§3.4, Eqs. (61)-(67); abstract"}],"minor_comments":[{"comment":"The proof of Lemma 2 refers to 'contradicts (30)' in two places; the reference should be to (50), the determinant of the matrix V for the degree-decrease case.","section":"Lemma 2 proof"},{"comment":"The initial conditions for Gegenbauer polynomials contain a notational inconsistency: 'Gλ1(x)' should be G^{(λ)}_1(x), matching the notation used elsewhere.","section":"§5.2"},{"comment":"Reference [14] has a typo: 'A introduction to orthogonal polynomials' should be 'An introduction to orthogonal polynomials'.","section":"References"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague —\n\nThe short version: this is a legitimate, mostly correct transfer-matrix derivation of arbitrary-step three-term recurrences for Favard-defined orthogonal polynomials, with explicit two-step formulas that check out. The main theorem needs a degeneracy qualifier, and there are a few small slips, but the core is sound.\n\nWhat's new: the general formula Q_{p+s}=M Q_p+N Q_{p-t}, with M,N given as ratios of the coefficients that express Q_{p+s} and Q_{p-t} in the basis {Q_p,Q_{p-1}}, is a clean algebraic consequence of the classical recurrence. I checked the s=t=2 Hermite, Gegenbauer, and Legendre formulas against the standard recurrences; they are correct. The degree-decrease and end-to-middle variants are natural but useful. The numerical comparison is honest: errors are around 1e-10%, which is what you'd expect from a stable recurrence.\n\nThe paper is also honest in one important way: it does not fit parameters. The coefficients are derived from the Favard data. The special-case discussion in §3.4 is a genuine attempt to handle vanishing denominators, and Remark 3 explicitly identifies the Legendre p=5,s=3,t=2,x=0 case where no rigid three-term relation exists.\n\nSoft spots, in proportion:\n\n1. Theorem 4 as stated is false at points where Q_p(x)=Q_{p-t}(x)=0 but Q_{p+s}(x)≠0. The paper's own Remark 3 gives exactly such a point. The §3.4 patch introduces a bias term with arbitrary coefficients, so the 'three-term' and 'completely irrelevant to the other members' claims in the abstract and introduction are overstrong. This is fixable: state the theorem on the dense set where the denominators and both Q_p and Q_{p-t} do not vanish, and treat the exceptional locus separately.\n\n2. Equation (46) has an index typo in the second row (it should involve ψ_{p-t}^{(p)}, not ψ_{p-t-1}^{(p)}). Lemma 2's proof cites Eq. (30) where it should cite Eq. (50). Both are minor.\n\n3. The paper claims applicability to 'any class' of Favard-defined orthogonal polynomials, but the explicit examples are all classical families. The general recipe is fine, but the presentation would be stronger with a nontrivial non-classical example.\n\nWho this is for: anyone doing skip evaluation of orthogonal polynomials, parallel computation, or studying the algebraic structure of recurrence relations. It is not a breakthrough, but it is a solid, citable method.\n\nRecommendation: send it to peer review. The core derivation is sound, the explicit formulas are new, and the overclaim is repairable with a modest revision.","headline":"A correct and useful transfer-matrix recipe for arbitrary-step three-term recurrences, but the universal claim needs a nondegeneracy qualifier and a few slips before it is a reliable reference.","tokens_in":21887,"tokens_out":2341,"would_cite":true,"duration_ms":23449,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["33C45","42C05","65D20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper's central claim: any orthogonal polynomial sequence generated by Favard's theorem satisfies a three-term recurrence with arbitrary degree step, so any member can be computed directly from two non-adjacent members through explicit","keywords":["Favard's theorem","orthogonal polynomials","three-term recurrence","arbitrary degree step","degree-skip recurrence","Hermite polynomials","Gegenbauer polynomials","Legendre polynomials"],"falsifier":"For Legendre polynomials at x=0, take p=5, s=3, t=2: L8(0)=35/128≠0 while L5(0)=L3(0)=0; any relation Q8=M Q5+N Q3 would give 0, so no rigid three-term relation of the asserted form can hold there. This is a concrete case where the unqualified statement fails.","tokens_in":20949,"feed_emoji":"🪜","tokens_out":9053,"duration_ms":91848,"temperature":0.7,"pith_summary":"For orthogonal polynomial families generated by Favard's recurrence, the paper derives a three-term recurrence relation with arbitrary degree step: Q_{p+s}(x) = M(x)Q_p(x) + N(x)Q_{p-t}(x). The upshot is that any member of such a family can be computed directly from any two members, skipping intermediate degrees. The coefficients M and N are explicit rational expressions in the Favard coefficients, and eliminating Q_{p-1} is the key step. The paper also gives decreasing-degree and end-to-middle versions, and explicit two-step recurrences for Hermite, Gegenbauer, and Legendre polynomials, with numerical comparisons showing accuracy close to the standard one-step recurrences. A reader should care because this turns a sequential recurrence into a jump recurrence, a structural fact about every Favard-defined orthogonal family.","feed_headline":"Jump recurrences skip degrees for any orthogonal polynomial family","feed_subtitle":"From any two members of a Favard orthogonal family, the third is reachable directly—no step-by-step climb needed.","key_machinery":"The central object is the transfer-matrix product W generated by the Favard step, whose entries are the bridge coefficients Ψ_{p+s}^{(p)}, Ψ_{p+s}^{(p-1)}, ψ_{p-t}^{(p)}, ψ_{p-t}^{(p-1)} satisfying (26) and (47). The argument pivots on forming the 2×2 matrix K from these four coefficients; since each transfer factor has determinant C_p≠0, K has nonzero determinant precisely when the two bridge relations are independent. Dividing the two representations of Q_{p+s} and Q_{p-t} by their Q_{p-1} coefficients makes Q_{p-1} cancel, yielding M=det(K)/ψ_{p-t}^{(p-1)}, N=Ψ_{p+s}^{(p-1)}/ψ_{p-t}^{(p-1)}. So K and its determinant carry the entire reduction: the existence of the skip-step relation is a","core_discovery":"The central discovery is Theorem 4: whenever {Q_p} obeys Favard's recurrence Q_{p+1}=(A_p x+B_p)Q_p - C_p Q_{p-1}, then for any s≥1 and 1≤t≤p the identity Q_{p+s}(x)=M(x)Q_p(x)+N(x)Q_{p-t}(x) holds, with M and N given by ratios and a determinant built from the two coefficient sequences that express Q_{p+s} and Q_{p-t} in terms of the adjacent pair Q_p,Q_{p-1}. The determinant condition encodes that the two bridge relations are independent, so Q_{p-1} can be eliminated. The paper's intended scope is universal: every orthogonal polynomial sequence by Favard's theorem, not only classical families, has such skip-step relations. It makes the construction explicit for Hermite, Gegenbauer and Legen","pith_inferences":["Beyond the paper, the bridge coefficients are products of Favard transfer matrices, so the determinant and denominator conditions are a rank condition: a rigid jump relation exists at x exactly when the two known members are not simultaneously zero there.","Beyond the paper, parity of symmetric orthogonal families forces such simultaneous zeros at symmetric points, so the paper's non-rigid bias cases are inevitable whenever s and t have different parity.","Beyond the paper, one can test the same construction on Laguerre, Jacobi, or exceptional orthogonal families and check whether the denominators' zeros match common zeros of the bridge polynomials.","Beyond the paper, since products of transfer matrices appear, the jump recurrences are equivalent to exponentiating the Favard transfer matrix, suggesting matrix-power algorithms for high-degree evaluation."],"forward_implications":["The standard one-step recurrence is the special case (s,t)=(1,1); the new relation therefore generalizes the classical three-term recurrence without leaving the Favard class.","Any Favard-defined orthogonal polynomial, classical or not, can be evaluated in arbitrary jumps, so high-degree values can be assembled from two widely separated known members rather than by iterating every degree.","The degree-decrease and end-to-middle formulas allow reconstruction of lower or intermediate members from two outer ones, which is useful for inverse evaluations or for moment and quadrature problems.","The explicit two-step Hermite, Gegenbauer and Legendre recurrences give directly usable formulas for blockwise evaluation; the numerical experiments show relative errors near the standard recurrences, down to roughly 10^{-10}–10^{-12} percent in the tested ranges.","If the clean formula holds at a point, it gives a rational expression for the skipped value in terms of the two known values, so the evaluation cost per target is O(1) once the coefficients are precomputed."],"fun_headline_variants":["Skip ahead or back: new recurrences for any orthogonal polynomials","Favard families leap degrees with new three-term recurrences","Orthogonal polynomials: jump degrees with arbitrary step recurrences","From adjacent pairs to far degrees: universal polynomial recurrences","Skipping steps in orthogonal polynomials: a general recurrence method"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The clean three-term formula is obtained by dividing by two coefficient functions; if either is zero at the evaluation point, the formula is not derived, and the paper's patches restore only a non-rigid relation with a bias term.","fun_headline_variants_meta":{"raw":{"variants":["Skip ahead or back: new recurrences for any orthogonal polynomials","Favard families leap degrees with new three-term recurrences","Orthogonal polynomials: jump degrees with arbitrary step recurrences","From adjacent pairs to far degrees: universal polynomial recurrences","Skipping steps in orthogonal polynomials: a general recurrence method"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00088,"raw_usage":{"total_tokens":3666,"prompt_tokens":793,"completion_tokens":2873,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":537,"completion_tokens_details":{"reasoning_tokens":2789}},"tokens_in":537,"tokens_out":2873,"duration_ms":21934,"temperature":1.0,"reasoning_tokens":2789,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T11:38:10.559119+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For Legendre polynomials at x=0, take p=5, s=3, t=2: L8(0)=35/128≠0 while L5(0)=L3(0)=0; any relation Q8=M Q5+N Q3 would give 0, so no rigid three-term relation of the asserted form can hold there. This is a concrete case where the unqualified statement fails.","supporting_citations":[],"review_version":2}