{"id":"46b2c358-035d-41f5-b615-28c77dcb5499","arxiv_id":"2607.14920","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A three-point continuous-variable quantum MacWilliams identity is constructed, and shown to collapse exactly to the two-point bound for GKP lattice and certifiable bosonic code sectors.","lead":"Researchers built a three-point version of the continuous-variable quantum MacWilliams identity, a tool for bounding the dimension of bosonic quantum error-correcting codes. They prove that this three-point machinery never improves on the existing two-point bound, and identify the code projector as the cause.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Normalization gap in Theorem 4.7: the equality construction drops the (2π)^N factor from Eq. (37), so the claimed exact collapse to the two-point LP optimum is not established as written.","rationale":"The reader's weakest assumption focuses on the asymmetric sandwich integrand and the fiberwise u-collapse; my independent read finds the same theorem is vulnerable, but through a more concrete internal gap: the equality construction in Theorem 4.7 does not enforce the normalization condition (ii) without introducing a (2π)^N factor. The F-layer transform Eq. (37) fixes ∫g = 2^{2N}(2π)^N, not 2^{2N}; the proof drops the (2π)^N. Consequently the proposed extension does not realize objective g(0) at the CE optimum unless one redefines K_Burch_2 or absorbs a normalization constant, and the stated exact equality K_CV_lat = K_Burch_2 is not justified. This is a proof-normalization issue rather than a refutation: it may be repairable by tracking the convention consistently, which is why I keep the reader's CONDITIONAL disposition rather than moving to REJECT. The CP-collapse certificates in §E.4 are a genuine positive feature but do not bear on this theorem. The concrete test isolates the issue with closed-form quantities from §5.1, so it can settle the matter without heavy numerics.","tokens_in":35247,"tokens_out":14098,"duration_ms":120020,"concrete_test":"Recompute the equality direction of Theorem 4.7 using an explicit one-mode (N=1) CE-type auxiliary, e.g. the Burchards–Levenshtein adapter fLev from §5.1. Take g(w) = A·fLev(w/2) with A chosen so that \\int g = 2^{2} (2π) (i.e. 2^{2N}(2π)^N at N=1), form f(u,w)=χ(||u||)g(w)cos(ω(w,u)/4) with χ(0)=1, and evaluate \\tilde f(0,0) from Eq. (37) together with f(0,0). If \\tilde f(0,0)=1 forces f(0,0) = 2^{2} (2π) g(0)/\\int g — namely (2π) K_CE(2, d) rather than K_CE(2,d) — then the claimed equality needs a corrected normalization or a redefined K_Burch_2. This is a closed-form check, no solver required.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 4.7 is the central claim: K_CV_lat(N,d) = K_Burch_2(N,d). The proof has a normalization gap in the equality direction. Condition (ii) requires \\tilde f(0,0)=1. By the F-layer definition Eq. (37), for g(w)=f(0,w) this gives \\hat g(0) := ∫ g(w) dw = 2^{2N}(2π)^N, not 2^{2N} as the proof states. The proposed extension f(u,w)=χ(||u||)g(w)cos(ω(w,u)/4) must therefore be rescaled by 2^{2N}(2π)^N to be admissible, changing the objective from g(0) to 2^{2N}(2π)^N g(0). For a CE-optimal g normalized by \\hat g(0)=1, this equals 2^{2N}(2π)^N K_CE(2N,2d) = (2π)^N K_CE(2N,d), not K_CE(2N,d). Equivalently, if one insists on \\hat g(0)=2^{2N}(2π)^N, the LP lower bound becomes (2π)^N K_CE(2N,d). The proof then asserts K_CE(2N,d)=K_Burch_2(N,d) without specifying which Fourier normalization K_Burch_2 uses; if K_Burch_2 absorbs (2π)^N, that identification is inconsistent with the stated definition of K_CE. Since the theorem is an exact equality, this missing/ambiguous factor is load-bearing. The reader's concern about the asymmetric dual integrand is related, but the decisive soft spot is this normalization bookkeeping in the collapse proof itself.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs a three-point continuous-variable (CV) quantum MacWilliams identity, deriving a closed-form integral kernel on the four-parameter Hermitian Gram configuration space, and uses the identity to formulate auxiliary-function bounds on the dimension K of bosonic codes. The main results are two collapse theorems: for GKP lattice codes the three-point optimum equals the Burchards two-point LP optimum (Theorem 4.7), and for general bosonic codes a completely-positive variant collapses to the two-point bound on every Laguerre sector certified (Theorem 6.3). The paper attributes both collapses to the code projector, which orients the bound correctly but makes the MacWilliams transform only fiberwise positive-definite. It also reports a classical dual-lattice packing bound that does improve on the two-point bound and carefully lists open problems.","tokens_in":1633,"tokens_out":2125,"duration_ms":137976,"significance":"If the collapse theorems are correct, the paper is a substantial negative result: unlike the classical Cohn-de Laat-Salmon three-point bound and the discrete-variable SDP hierarchies, the CV auxiliary-function three-point method cannot improve on the two-point LP bound. The paper has real strengths: the kernel derivation is detailed and cross-checked by marginalizations and numerical checks; the CP-collapse certificates for N=1, M=2,...,8 are exact rational LDLT certificates rather than solver outputs; the comparison with Burchards is made without fitted parameters; and the scope of the claims--especially the open full trace-class cone--is stated explicitly. These features make the paper a useful contribution to the CV coding theory literature, provided the normalization gap in the proof of Theorem 4.7 is repaired.","major_comments":[{"comment":"The proof of the collapse has a normalization error that is load-bearing. From Definition 4.1(ii), ef(0,0)=1, and Eq. (37) gives ef(0,0) = 2^{-2N}(2pi)^{-N} times the integral over w of f(0,w). Hence for g(w)=f(0,w), condition (ii) fixes the integral of g to be 2^{2N}(2pi)^N. The proof nevertheless states that bg(0)=2^{2N}(2pi)^N ef(0,0) times (2pi)^{-N} = 2^{2N}; the extra (2pi)^{-N} is not present in Eq. (37). With the correct value bg(0)=2^{2N}(2pi)^N, the lower-bound direction yields g(0) >= (2pi)^N K_CE(2N,d) for the standard normalization of the Cohn-Elkies LP optimum, not g(0) >= K_CE(2N,d). Similarly, the equality-direction extension f(u,w)=chi(||u||) g(w) cos(omega(w,u)/4) must be rescaled by 2^{2N}(2pi)^N to satisfy ef(0,0)=1, which multiplies the objective f(0,0) by the same factor. Thus Eq. (46) is not established as written. The authors should either correct Eq. (37) and all","section":"Section 4.4, Theorem 4.7 (Eqs. (37), (45)-(46))"},{"comment":"The theorem identifies K_CV_lat(N,d) with K_Burch_2(N,d), but K_Burch_2 is only defined by reference to [Bur25, Thm. 1]. The Cohn-Elkies LP optimum used in the proof is defined with the normalization hat h(0) = integral of h(x) dx. Without an explicit statement of the normalization in Burchards' two-point LP optimum--in particular whether it absorbs any factor of (2pi)^N--the asserted equality K_CE(2N,d) = K_Burch_2(N,d) cannot be checked. This is not a cosmetic issue: the missing factor in the preceding comment changes the result by (2pi)^N unless K_Burch_2 is defined with that same factor. The manuscript should give the precise definition of K_Burch_2 in the conventions of Section 2.1 and prove or cite the identity relating it to the Cohn-Elkies LP optimum with the chosen Fourier normalization.","section":"Section 4.4, Eq. (46) and Remark 4.8"}],"minor_comments":[{"comment":"The numerical gain factors 0.871 and 0.564 should be recomputed against the stated definition of j_N as the first positive zero of J_N. For example, with j_4 approximately 7.588, the expression (4pi)^4 * 4! * 2^4 / j_4^8 is about 0.087, not 0.871. If j_N denotes a different zero or a different Bessel order, that needs to be stated explicitly. This does not affect the collapse theorem but undermines the saturation examples as currently written.","section":"Sections 5.1-5.2, Eqs. (48)-(51) and Table 1"},{"comment":"The theorem's scope depends on the asymmetric dual integrand Eq. (15) and the resulting fiberwise transform. Remark 4.8 argues this is the canonical extension, but the theorem statement itself should make the dependence explicit, for example by saying 'for the class of auxiliary functions defined by the fiberwise MacWilliams transform Eq. (37).' This would help readers who do not accept the canonicity argument.","section":"Definition 4.1 and Theorem 4.7"},{"comment":"The quantities K times d^{2N} are dimensionful in the Burchards convention. The conversion in Section 2.6 is useful, but Table 1 and Eqs. (50)-(51) should state that all numerical values are in that convention, and ideally the table should include the conversion factors so the numerical claims can be checked independently.","section":"Section 2.6 and Table 1"}],"recommendation":"major_revision","confidential_remarks":"The central collapse theorem is a strong and interesting claim, and the paper contains rigorous local proofs and exact certificates. However, the normalization gap in Theorem 4.7 is a genuine load-bearing issue, and the relationship between K_Burch_2 and the Cohn-Elkies optimum is under-specified. Since the paper's headline result depends on this equality, I cannot recommend acceptance until the normalization is fixed and the definition of K_Burch_2 is made precise. The numerical inconsistency in Section 5 should also be corrected. I do not see this as a reject: the issue appears local and fixable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short take: this paper constructs a three-point continuous-variable quantum MacWilliams identity and then shows, on the natural admissible class, that the three-point SDP bound does not beat the two-point LP bound for GKP lattice codes or on the certifiable CP sectors. That is a real negative result, novel and well-motivated, and the structural explanation—code projector orients the bound but kills the full positivity that powers the classical improvement—is plausible.\n\nWhat is genuinely good: the identity and its closed-form kernel are new; the two collapse theorems are proved in detail; and the paper is unusually honest about its boundaries. It says explicitly that the CP collapse is only for the first eight Laguerre levels at N=1, that the full trace-class cone is open, and that the certificate margins decay geometrically. The numerical checks and exact rational LDLT certificates on the CP side are real evidence, not hand-waving.\n\nThe soft spot is in Theorem 4.7, which is the headline. Eq. (37) defines the transform with a 2^{-2N}(2π)^{-N} prefactor, so the normalization ef(0,0)=1 forces ∫g = 2^{2N}(2π)^N, not 2^{2N} as the proof states. That missing (2π)^N is load-bearing: it turns the lower bound into (2π)^N times the Cohn–Elkies optimum unless K_Burch_2 uses a convention that absorbs this factor. The proof then identifies K_CE(2N,d) with K_Burch_2(N,d) without spelling out which Fourier normalization applies. As written, the equality is not established; it may become an inequality with a factor, which would still be a no-go but a weaker one. This is central, not cosmetic, and a referee must check the conventions.\n\nA softer second issue: the collapse depends on the asymmetric sandwich dual integrand, which makes the transform fiberwise in u. The paper argues that this is the unique BCH-closed extension, and that is reasonable within the projector-pairing framework, but the negative result is tied to that modeling premise. A different three-point identity built on a full symplectic Fourier transform might not collapse. The paper acknowledges this in Remark 4.8, so it is not hidden.\n\nVerdict: send it to peer review. The novelty is clear, the technical core is substantial, and the normalization gap looks repairable rather than fatal. A careful referee should pin down the Fourier convention in the collapse proof and confirm whether the equality survives or becomes a factor-off inequality.","headline":"A genuinely new CV three-point MacWilliams identity and a plausible no-go result, but the proof of the main collapse theorem has a normalization gap that needs fixing before the equality claim holds.","tokens_in":36154,"tokens_out":4565,"would_cite":true,"duration_ms":39779,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs a three-point continuous-variable quantum MacWilliams identity and proves that, for GKP lattice codes, the three-point bound exactly equals the two-point LP optimum, while the general bosonic completely-positive cone c","keywords":["continuous-variable quantum error correction","GKP codes","quantum MacWilliams identity","three-point bound","linear programming bound","semidefinite programming","sphere packing","symplectic lattices"],"falsifier":"Exhibit any admissible auxiliary f for the lattice bound at some (N,d) with f(0,0) strictly below the two-point LP optimum while satisfying adjoint-positivity, normalization, and the phase-aware sign condition on the lattice triangle; that would refute Theorem 4.7. For the general-bosonic collapse, find a radial Choi form J ⪰ 0 on the ninth Laguerre level at one mode (or any level beyond the certified eight) satisfying the phase-sign condition with J ≠ 0; its existence would show the CP collapse is a finite-truncation artifact.","tokens_in":35100,"feed_emoji":"⚛️","tokens_out":8520,"duration_ms":78498,"temperature":0.7,"pith_summary":"The paper builds the three-point version of the continuous-variable quantum MacWilliams identity and then shows that, unlike in classical sphere packing, the three-point method cannot strengthen the two-point bound. For GKP lattice codes, the three-point optimum exactly equals the two-point linear-programming optimum: no admissible auxiliary, however it depends on the new edge coordinate, improves the bound. For general bosonic codes, the natural completely-positive reformulation collapses to two-point on every finite Laguerre sector the author certifies, with the full trace-class cone left open. The paper identifies a single structural cause: the code projector orients the bound correctly through an asymmetric K²/K¹ prefactor, but that same asymmetry makes the transform fiberwise and removes the full positivity that powers the classical three-point gain.","feed_headline":"Three-point bound collapses to two-point for GKP lattice codes","feed_subtitle":"The new three-point machinery exactly reproduces the two-point LP optimum for GKP lattice codes.","key_machinery":"The load-bearing structure is the asymmetric two-displacement sandwich F_B^(3)(v₁,v₂) = tr(D(v₁)O₁ D(v₂)† O₂†). A Baker–Campbell–Hausdorff expansion shows the resulting transform is fiberwise in the edge coordinate u = v₂ − v₁, while the center-of-mass coordinate w = v₁ + v₂ is symplectic-Fourier paired; the kernel carries an edge-norm shell constraint, an on-shell phase, and a Bessel factor of order N−2. This fiberwise structure makes the normalization and objective of the bound local to the u = 0 slice, which is why the admissible class reduces to the two-point LP problem. The configuration space is the Hermitian 2×2 Gram matrix H(v₁,v₂) with invariants (r₁², r₂², α, ω₁₂), where ω₁₂ is the","core_discovery":"The central claim is that the natural three-point extension of the CV quantum MacWilliams framework collapses to the two-point bound. The paper constructs three-point weight distributions A₃ and B₃ on the four-parameter configuration space of Hermitian 2×2 Gram matrices, whose fourth parameter is the symplectic invariant ω₁₂ carrying the GKP quantization condition and a ±1 phase, and derives a closed-form integral kernel. For GKP lattice codes, Theorem 4.7 shows the three-point optimum equals the two-point LP optimum exactly; the edge coordinate added by the three-point construction has no effect on the optimal value. For general bosonic codes, the natural factored-form kernel-positive cone","pith_inferences":["If the lattice collapse is generic, the auxiliary-function route to CV three-point bounds is structurally exhausted; the paper's own list of successors — a noncommutative moment/SOS hierarchy, copositive cones, conditional twisted enumerators, and CV shadow enumerators — is where any genuine three-point improvement for bosonic codes would have to live.","A direct test of the boundary is available: extend the N = 1 completely-positive certificate beyond the eight certified radial Laguerre levels using exact rational LDLT certificates. The paper reports that the certificate margin decays geometrically and that a sparse rank-12 net flips to spurious feasibility, so the full trace-class collapse is plausible but not assured.","The same fiberwise-collapse mechanism may recur in other projector-based auxiliary MacWilliams bounds whenever the weight enumerator pair carries asymmetric prefactors; the paper does not claim this, but its quantum–classical contrast is stated generically enough to invite the extrapolation."],"forward_implications":["For GKP lattice codes satisfying the stated distance conditions, no admissible three-point auxiliary can improve on the two-point LP optimum; the E8 and Leech magic functions saturate that optimum rather than beat it.","The three-point identity reproduces the known two-point Levenshtein-type, E8, and Leech bounds exactly on the overlap range, so the three-point apparatus adds no new numerical regime to the two-point theory.","For general bosonic codes, on the completely-positive cone, the three-point term f₃(0,0) ≥ 0 together with the phase-sign condition forces the Choi form to zero on every certified Laguerre rank, reducing the bound to two-point there; the full trace-class cone is not settled.","A genuine three-point improvement survives only for well-conditioned GKP lattices through a classical sphere-packing bound on the symplectic dual lattice, which is roughly 5% below the two-point bound at 2N = 4.","The quantum–classical contrast is structural: the projector asymmetry (K² vs K¹) both orients the bound as an upper bound and removes the full positive-definiteness that powers the classical three-point improvement, and within this auxiliary-function class the two effects cannot coexist."],"fun_headline_variants":["Three-point CV MacWilliams identity collapses to two-point for GKP","GKP lattice codes: three-point optimum equals two-point bound exactly","Three-point bound vanishes for general bosonic codes","Symplectic invariant kills three-point improvement in CV codes","Three-point quantum identity can't beat two-point for GKP"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The three-point dual distribution is defined by the asymmetric sandwich tr(D(v₁)O₁ D(v₂)† O₂†); this choice makes the MacWilliams transform fiberwise in the edge coordinate, so the objective, normalization, and constraints collapse onto the u = 0 slice, and if a different dual integrand produced a full 4N-dimensional symplectic Fourier transform, the equality with the two-point LP optimum would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Three-point CV MacWilliams identity collapses to two-point for GKP","GKP lattice codes: three-point optimum equals two-point bound exactly","Three-point bound vanishes for general bosonic codes","Symplectic invariant kills three-point improvement in CV codes","Three-point quantum identity can't beat two-point for GKP"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000168,"raw_usage":{"total_tokens":1107,"prompt_tokens":766,"completion_tokens":341,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":510,"completion_tokens_details":{"reasoning_tokens":257}},"tokens_in":510,"tokens_out":341,"duration_ms":3841,"temperature":1.0,"reasoning_tokens":257,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T00:40:49.248768+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit any admissible auxiliary f for the lattice bound at some (N,d) with f(0,0) strictly below the two-point LP optimum while satisfying adjoint-positivity, normalization, and the phase-aware sign condition on the lattice triangle; that would refute Theorem 4.7. For the general-bosonic collapse, find a radial Choi form J ⪰ 0 on the ninth Laguerre level at one mode (or any level beyond the certified eight) satisfying the phase-sign condition with J ≠ 0; its existence would show the CP collapse is a finite-truncation artifact.","supporting_citations":[],"review_version":1}