{"id":"8cb359ef-a812-47fe-a616-d282a2ad0990","arxiv_id":"2608.00601","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Explicit symplectic Barnes-Wall GKP codes achieve rate (1/2) log2 N with constant distance and an exact O(N log^2 N) bounded-distance decoder.","lead":"This paper constructs a family of continuous-variable quantum error-correcting codes based on Barnes-Wall lattices, achieving an encoding rate that grows logarithmically with the number of modes and a deterministic O(N log^2 N) decoder. It illustrates a clear rate-distance tradeoff in GKP codes and offers a template for adapting classical lattice decoders to quantum error correction.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4's decoder-inheritance proof ignores the 1/c scaling between the symplectic dual and the primal Barnes-Wall lattice, so the claimed identical decoding radius is wrong; the effective radius is cρ.","rationale":"The central construction is algebraically plausible, and the rate–distance parameters are internally consistent. The reader's stated weakest assumption—that the real-to-complex lattice identification requires G_m to commute with Ω—is somewhat misdirected: the complexification step in Appendix A only needs the column lattice of G_m to be closed under the symplectic form, which follows from the recursion and Lemma 1 by induction, even though the paper does not spell out that proof. The more concrete, load-bearing defect is in the decoder-inheritance proof: the transformation in Theorem 4 is not distance-preserving because of the c-scaling, and the effective decoding radius is cρ rather than ρ. This is an internal inconsistency in a proof underpinning the paper's headline O(N log² N) bounded-distance decoder. The paper's main text asserts ρ ∼ 1/2, but that requires the external Barnes-Wall decoder to achieve the optimal packing radius, which is not cited or verified. This does not invalidate the entire construction—it is an addressable gap—but it does keep the result conditional on a corrected and completed decoder-inheritance analysis.","tokens_in":8586,"tokens_out":48868,"duration_ms":376050,"concrete_test":"Re-derive the norm relations in (S10)–(S14) for a concrete case, say m = 4 (N = 8). Choose a target τ at Euclidean distance ε = 1/2 from Λ⊥_SBW and compute the input to D_BW, namely (1/c)Q†Φ(ΩTτ). Verify that its distance to Λ_BW is ε/c = 2^(3/2)/2 ≈ 2.83, not ε. If this scaling is confirmed, Theorem 4's 'same radius' statement is false. Then inspect the Micciancio–Nicolosi decoder [15] for its proven BDD radius; if it is less than 2^((m−1)/2)/2, recompute the effective radius c·ρ and check whether the paper's claim ρ ∼ 1/2 holds.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 4 (Supplemental Appendix B) contains a norm-scaling error. The chain (S10)–(S14) reduces CVP on the symplectic dual Λ⊥_SBW to CVP on the primal Barnes-Wall lattice Λ_BW, but the step from (S13) to (S14) uses Λ*_BW = cQΛ_BW, so the distance from the target to cQΛ_BW equals c times the distance from (1/c)Q†Φ(ΩTτ) to Λ_BW. Since c = 2^(−(m−1)/2) < 1 for m > 1, the maps in (S13)–(S14) scale distances by c and are not strictly distance-preserving as the proof claims. Consequently, if D_BW has radius ρ, the composite decoder succeeds only for errors of norm < cρ, not ρ. For the advertised BDD radius Δ/2 = 1/2 (main text, after Eq. (S8)), the primal decoder must have radius ρ = 1/(2c) = 2^((m−1)/2)/2, i.e., the optimal packing radius of Λ_BW. The paper never states or proves that the cited decoder [15] achieves this optimal radius, and the proof's 'same decoding radius' claim is internally inconsistent with Δ² = 1: an error of norm ρ (exponentially large) cannot be uniquely decoded, so the claimed invariant radius would exceed the packing radius. This leaves the effective BDD radius of the GKP decoder unproven, which is central to the paper's comparison with heuristic decoders whose radii shrink with N.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper constructs an explicit infinite family of multimode GKP codes from a symplectic realization of the Barnes-Wall lattice. The generator recursion G_{m+1} = [[G_m, 0], [G_m, R_m G_m]] with R_m = I + Omega is claimed to satisfy symplectic integrality (Theorem 1), to encode k_m = (m-1)2^{m-2} logical qubits into N = 2^{m-1} modes, and to have constant squared distance Delta^2 = 1 via an identification with the Gaussian-integer Barnes-Wall lattice (Appendix A). The principal algorithmic claim is an O(N log^2 N) deterministic bounded-distance decoder inherited from the Barnes-Wall decoder (Theorem 4, Appendix B), which yields logarithmic rate at constant BDD radius. The paper also derives a Minkowski-type rate-distance bound and compares the construction with NTRU/M-SIS and surface-GKP codes.","tokens_in":8880,"tokens_out":27244,"duration_ms":256243,"significance":"If the decoder-radius claim is repaired, this is a valuable explicit data point: the code family is concrete, has no fitted parameters, and makes the rate-distance tradeoff transparent. Theorem 1 and the rate calculation in Corollary 1 are sound and easy to verify, and the use of established Barnes-Wall facts is appropriate. The claimed deterministic O(N log^2 N) decoder with a nontrivial radius would be a genuine contribution to multimode GKP decoding, and the comparison with heuristic or shrinking-radius decoders is useful. The main weakness is not the overall strategy but a specific scaling error in the decoder-inheritance proof, plus one omitted commutation argument in the lattice identification.","major_comments":[{"comment":"The proof of Theorem 4 contains a norm-scaling error. For a = Phi(Omega^T tau), the step from (S13) to (S14) is min_{x in Lambda_BW} ||a - cQ x|| = c min_{x in Lambda_BW} ||(1/c)Q^dagger a - x||, so the map feeding D_BW scales error vectors by 1/c, not by 1. The assertions immediately after (S14) that (S10)-(S14) are strictly distance-preserving and that D_SBW_perp has the same decoding radius rho as D_BW are therefore false; the inherited radius is c rho. Since c = 2^{-(m-1)/2} < 1 for m > 1, the advertised radius Delta/2 = 1/2 (main text after Eq. (S8)) requires D_BW to have radius 1/(2c) = 2^{(m-3)/2}, which is exactly the packing radius of the primal Barnes-Wall lattice. The manuscript neither states nor proves that the decoder of Ref. [15] achieves this optimal radius. The theorem statement and proof should be corrected, and the required radius of the primal decoder should be made explicit.","section":"Supplemental Material, Appendix B, Eqs. (S10)-(S14)"},{"comment":"The claim that 'the real recursion for G_m translates to a complexified matrix tilde G_m over Z[i]' implicitly assumes that Phi G_m Phi^{-1} is C-linear, which is equivalent to G_m Omega = Omega G_m. This commutation is not proved anywhere, yet the lattice identification in Theorem 3 and the distance computation in Remark 1 rest on it. The commutation is in fact true by induction from Eq. (2) because R_m = I + Omega commutes with Omega and the induction hypothesis gives [G_m, Omega] = 0, but a short proof should be supplied before Theorem 3.","section":"Supplemental Material, Appendix A, paragraph before Eq. (S3)"}],"minor_comments":[{"comment":"The determinant of R_m is stated as 2^{2m-1}; for the 2^m x 2^m matrix R_m = I + Omega, the determinant is 2^{2^{m-1}}. Please correct the exponent or clarify the notation.","section":"Main text, after Eq. (2)"},{"comment":"The statement D_m = det(G_m) should read D_m = |det(G_m)|, or should note that all determinants in this construction are positive.","section":"Corollary 1"},{"comment":"In the composition defining D_SBW_perp, the matrix Q is the isoduality unitary from Lambda*_BW = cQ Lambda_BW, but this is not restated at Eq. (S9); adding this clarification would help readability.","section":"Appendix B, Eq. (S9)"},{"comment":"The vertical axis 'Normalized BDD radius' is not defined; please specify the normalization, for example by Delta/2. Also, Table I lists the actual distance of SBW-GKP as O(1), whereas the text says it is exactly 1; use '1' for precision.","section":"Figure 3(b) and Table I"},{"comment":"Since Theorem 4's corrected statement requires a specific decoding-radius guarantee for the primal Barnes-Wall decoder, the citation to Ref. [15] should explicitly state which radius is available (ideally the full packing radius).","section":"References, [15]"}],"recommendation":"major_revision","confidential_remarks":"The construction itself appears sound, and the main problem is a fixable scaling error in the decoder-inheritance proof rather than a flaw in the code algebra. The authors should be asked to verify the decoding radius of the cited Barnes-Wall decoder and to add the missing commutation proof in Appendix A. If the required optimal radius is not available in Ref. [15], the advertised constant BDD radius claim would need substantial revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. The construction is real: an explicit symplectic recursive generator that satisfies the GKP stabilizer condition and yields a Barnes-Wall lattice with rate (1/2) log_2 N. That combination is new, as far as I can tell, and the distance calculation via scaled-isoduality is clean. Theorems 1 and 3 look correct to me; the lattice equivalence via the complexification map is legitimate, and the commutation worry in the report doesn't land—Φ is a homomorphism, so the recursion passes to the complex domain directly.\n\nThe soft spot is Theorem 4. The chain (S10)–(S14) is not distance-preserving. Step (S13)–(S14) pulls the factor c out of the norm, so the composite decoder inherits radius cρ, not ρ. Since c = 2^{-(m-1)/2}, that is exponentially smaller than ρ, and the 'same radius' claim is wrong as written. The good news is the fix is clear: if D_BW decodes up to the packing radius of Λ_BW (which is 2^{(m-1)/2}/2), then cρ = 1/2, exactly the advertised Δ/2. So the constant-distance story survives, provided the cited decoder [15] actually has that optimal radius. The paper does not state or prove that. The author needs to pin down the radius of [15] and rewrite the inheritance proof with the scaling made explicit. This is a repairable bug, not a fatal one, but it is load-bearing: the paper's whole selling point is exact bounded-distance decoding, and the proof as it stands doesn't establish the radius.\n\nOne more thing: the paper's comparison with surface-GKP and random lattices is okay but a bit schematic. The constant distance and the 1/(8N) noise-variance requirement are spelled out fairly, so no complaint there.\n\nVerdict: worth refereeing seriously. I would not accept it in this form. The main theorems are explicit and checkable, and the decoder issue is localized. With a corrected Appendix B, this is a solid contribution. I'd bring it to the reading group and I'd want to see a revised version.","headline":"The construction is clever and mostly sound, but the decoder-inheritance proof has a real scaling error that must be fixed before the radius claim is trusted.","tokens_in":9443,"tokens_out":8142,"would_cite":false,"duration_ms":65715,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11H71","81P70"],"pacs":[],"model":"deepseek-v4-flash","headline":"A recursive symplectic construction turns Barnes-Wall lattices into GKP codes that pack a half-log rate of qubits per mode and decode deterministically in near-linear time.","keywords":["GKP codes","Barnes-Wall lattice","symplectic integrality","bounded-distance decoding","multimode quantum error correction","continuous-variable quantum codes","logarithmic encoding rate"],"falsifier":"For a small level, say $m=3$ or $m=4$, directly compute $G_m$ from the recursion and check whether $G_m^T\\Omega G_m$ is integer-valued and antisymmetric, and whether the lattice generated by $G_m$ equals the Gaussian-integer Barnes-Wall lattice up to a unimodular factor; a single counterexample would overturn the distance and decoder claims. Alternatively, enumerate the shortest nonzero vector of the symplectic dual for $N=8$ and verify that its squared norm is exactly $1$ in units of $2\\pi$.","tokens_in":8359,"feed_emoji":"⚛️","tokens_out":7797,"duration_ms":69196,"temperature":0.7,"pith_summary":"This paper constructs an explicit infinite family of multimode Gottesman-Kitaev-Preskill (GKP) quantum error-correcting codes. The encoding rate grows as $\\frac{1}{2}\\log_2 N$ logical qubits per mode, while a deterministic bounded-distance decoder runs in $O(N\\log^2 N)$ time. This is achieved by a recursive matrix generator that is at once a valid symplectic stabilizer and an exact realization of the Barnes-Wall lattice. The price is a constant code distance, $\\Delta^2=1$ in units of $2\\pi$, reflecting a deliberate rate-distance tradeoff. A sympathetic reader would care because high-rate multimode GKP codes usually lack efficient, deterministic decoding; this construction provides one.","feed_headline":"Log-rate GKP codes decode in near-linear time","feed_subtitle":"A recursive symplectic lattice packs up to (1/2)log N qubits per mode and decodes deterministically in O(N log^2 N) time.","key_machinery":"The load-bearing object is the recursive block generator $G_{m+1}$ together with the scaled symplectic rotation $R_m = I + \\Omega$, where $\\Omega$ is the interleaved symplectic form. The identity $R_m^T R_m = 2I$ makes $R_m/\\sqrt{2}$ simultaneously orthogonal and symplectic, which keeps the overlap matrix $G_m^T\\Omega G_m$ integral and antisymmetric. The second key mechanism is the coordinate map $\\Phi(q_j,p_j)=q_j+ip_j$, which converts real symplectic multiplication by $\\Omega$ into complex multiplication by $-i$; under this map $R_m$ becomes the Gaussian scalar $1-i$, so the real recursion becomes the Gaussian-integer Barnes-Wall recursion. The scaled-isoduality of the Barnes-Wall lattice then fixes the dual distance, and a composite decoder built from the primal bounded-distance decoder plus six isometric transformations gives the $O(N\\log^2 N)$ decoding guarantee.","core_discovery":"The central claim is that the recursion $G_{m+1} = \\begin{pmatrix} G_m & 0 \\\\ G_m & R_m G_m \\end{pmatrix}$ with $R_m = I + \\Omega$ generates, for every $m$, a valid GKP stabilizer lattice with symplectic overlap matrix $K_m = G_m^T\\Omega G_m$ that is integer-valued and antisymmetric. The same recursion, after a change of coordinates that turns $\\Omega$ into multiplication by $-i$, reproduces the Gaussian-integer Barnes-Wall lattice up to a unimodular transformation. As a result, the code encodes $k_m = (m-1)2^{m-2}$ logical qubits into $N = 2^{m-1}$ modes, giving rate $R = \\frac{1}{2}\\log_2 N$. Using the scaled-isoduality of the Barnes-Wall lattice, the symplectic dual distance evaluates exactly to $\\Delta^2 = 1$, and the known $O(N\\log^2 N)$ bounded-distance decoder for the primal lattice carries over through a chain of isometries to the symplectic dual. The paper positions this as the first explicit infinite GKP family combining logarithmic rate with a deterministic near-linear-time decoder, at the cost of distance that does not scale with $N$.","pith_inferences":["Beyond the paper: if the symplectic-commutation gap flagged in the supplement is closed, a natural next test is Monte Carlo benchmarking against Gaussian and burst noise; the constant-distance tradeoff suggests concatenation with a classical outer code as the practical route to fault tolerance.","Beyond the paper: the same recursive pattern might be adapted to other scaled-isodual lattices with fast decoders, potentially yielding GKP families with the same logarithmic rate but a different distance-versus-$N$ curve.","Beyond the paper: one could numerically test small instances ($m=3,4$) by exact enumeration of the symplectic dual's shortest vector, which would directly verify the $\\Delta^2=1$ claim without relying on the unimodular equivalence proof."],"forward_implications":["For $N\\ge 8$ modes, the encoding rate exceeds one logical qubit per physical mode; for instance, the $m=4$ level encodes 12 logical qubits into 8 modes.","The deterministic decoder succeeds whenever the total displacement error has norm below the constant decoding radius $\\rho\\sim\\Delta/2=1/2$, which for i.i.d. Gaussian noise requires single-mode variance $\\sigma^2\\lesssim 1/(8N)$.","The construction abandons geometric locality: a global symplectic scrambling circuit scatters localized burst errors into diffuse syndrome patterns that the global decoder can correct, in contrast to surface-GKP codes where adjacent-mode bursts form fatal logical strings.","The constant distance $\\Delta^2=1$ sits well below the asymptotic Minkowski bound $\\Delta^2_{\\max}\\sim O(\\sqrt{N})$ for this rate, making the code an explicit operating point where logarithmic rate is bought with non-scaling protection.","For small $N$ (up to 64 modes), the code inherits the optimal sphere-packing properties of exceptional Barnes-Wall relatives such as the Gosset and laminated lattices, giving deterministic decoding without the tail risk of random lattice instances."],"supporting_citations":[{"why":"Defines the GKP code and its displacement-stabilizer structure, which the whole construction builds on.","marker":"[1]"},{"why":"Supplies the symplectic integrality condition $M^T\\Omega M=2\\pi K$ that a valid multimode GKP stabilizer lattice must satisfy.","marker":"[7]"},{"why":"Introduces the Barnes-Wall lattice and its recursive generation structure, the object being symplectically realized.","marker":"[14]"},{"why":"Provides the $O(N\\log^2 N)$ bounded-distance decoding algorithm for the primal Barnes-Wall lattice that the GKP decoder inherits.","marker":"[15]"},{"why":"Supplies the scaled-isoduality relation for the Barnes-Wall lattice used to compute the constant code distance $\\Delta^2=1$.","marker":"[21]"},{"why":"Contains the unimodular lattice equivalence and the isometric decoder-inheritance proof on which the main claims rest.","marker":"[22]"}],"fun_headline_variants":["Logarithmic-rate GKP codes with near-linear decoding","Deterministic O(N log²N) decoder for log-rate GKP codes","GKP codes: logarithmic rate, constant distance, near-linear decode","Log-rate GKP codes decode in O(N log²N) time","Symplectic Barnes-Wall GKP: log rate, fast decoding"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument that the real recursive generator produces exactly the classical Barnes-Wall lattice depends on the unproved step that the generator commutes with the symplectic form $\\Omega$, so that the real recursion can be rewritten as the Gaussian-integer recursion; if that equivalence fails, the distance calculation and the decoder inheritance collapse.","fun_headline_variants_meta":{"raw":{"variants":["Logarithmic-rate GKP codes with near-linear decoding","Deterministic O(N log²N) decoder for log-rate GKP codes","GKP codes: logarithmic rate, constant distance, near-linear decode","Log-rate GKP codes decode in O(N log²N) time","Symplectic Barnes-Wall GKP: log rate, fast decoding"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000785,"raw_usage":{"total_tokens":3508,"prompt_tokens":1035,"completion_tokens":2473,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":651,"completion_tokens_details":{"reasoning_tokens":2379}},"tokens_in":651,"tokens_out":2473,"duration_ms":16079,"temperature":1.0,"reasoning_tokens":2379,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:22:54.354911+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a small level, say $m=3$ or $m=4$, directly compute $G_m$ from the recursion and check whether $G_m^T\\Omega G_m$ is integer-valued and antisymmetric, and whether the lattice generated by $G_m$ equals the Gaussian-integer Barnes-Wall lattice up to a unimodular factor; a single counterexample would overturn the distance and decoder claims. Alternatively, enumerate the shortest nonzero vector of the symplectic dual for $N=8$ and verify that its squared norm is exactly $1$ in units of $2\\pi$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the Barnes-Wall lattice and its recursive generation structure, the object being symplectically realized."},{"cited_title":"Micciancio and A","cited_arxiv_id":null,"evidence_quote":"Provides the $O(N\\log^2 N)$ bounded-distance decoding algorithm for the primal Barnes-Wall lattice that the GKP decoder inherits."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the scaled-isoduality relation for the Barnes-Wall lattice used to compute the constant code distance $\\Delta^2=1$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contains the unimodular lattice equivalence and the isometric decoder-inheritance proof on which the main claims rest."}],"review_version":1}