{"id":"ca45901e-a201-459c-b544-7289bb156909","arxiv_id":"2504.17589","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives the correct MacWilliams formula for the nu-function of ternary-code lattices and refutes Solé's 1995 conjecture with counterexamples for k=3 to 10.","lead":"This paper proves a new MacWilliams-type identity for lattices built from ternary codes and shows that a 1995 conjecture by Solé is false in general. The result corrects the lattice analogue of a classical coding theory theorem and matters for counting lattice points in pyramids.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Theorem 4 derivation is sound, but the universal 'except binary' classification is unsupported by the zero-code counterexamples.","rationale":"I checked the algebra of Theorem 4, including Lemma 3.3, the parameter relation, and the final coefficient identity. With the intended reading of the weight-enumerator argument as tanh α / (2 − tanh α), the derivation is internally consistent: Lemma 3.3 follows from the three residue-class nu-series, the MacWilliams step correctly transforms W_{C⊥}, and the bracketed factor reduces to ((1+tanh(β/2))/(1−tanh(β/2)))^n ((1−tanh^3(α/2))/(1+tanh^3(α/2)))^n after using |C| det(Λ) = 3^n. The zero-code examples genuinely violate Solé's universal identity, so the paper does contain a valid counterexample to the 1995 conjecture. The remaining issue is the scope of the 'except binary' claim. The reader's weakest_assumption identifies precisely this gap: the examples are not exhaustive, and no theorem proves that every lattice satisfying (1.7) must be a binary Construction A lattice. That is the same load-bearing concern I would raise. Because this overclaim is separable from the paper's core derivation and can be fixed by tightening the conclusion to 'the conjecture fails for the family A_k({0}), k ≥ 3, and the binary A_2(C) family remains the only known supporting case,' the appropriate verdict is unchanged from the reader's CONDITIONAL: the paper should be accepted only after the classification claim is either proved or softened. I do not see a fatal internal inconsistency in the main theorem, nor in the m-tuple or complete-weight-enumerator identities, so no stronger verdict adjustment is warranted.","tokens_in":12012,"tokens_out":19450,"duration_ms":178503,"concrete_test":"Enumerate all ternary linear codes C ⊂ Z_3^n for n = 1, 2, 3, construct Λ = A_3(C), and test identity (1.7) with high-precision arithmetic over a grid of α values (for example α = 0.1, 0.2, ..., 5.0), computing both sides by truncated series with rigorous error bounds. If any A_3(C) that is not equal (up to integer scaling or orthogonal isomorphism) to an A_2(C') lattice satisfies the identity, the 'except binary' classification is false. Alternatively, solve the one-dimensional identity symbolically for cZ; this determines which scaling factors c satisfy the conjecture and directly tests the claimed binary-only classification in the n = 1 case already used in Table 1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's refutation of Solé's conjecture is valid as an existential refutation: Table 1 gives one-dimensional lattices Λ = A_k({0}) = kZ for k = 3,...,10 that fail identity (1.7), and a single such counterexample already disproves the conjecture as stated for arbitrary lattices. The load-bearing weakness is instead the stronger classification claim in the abstract and Section 1.2, namely that the conjecture 'never holds in the general case, except for lattices associated with a binary code.' The evidence for this exclusivity is only the zero-code family, which does not rule out nonzero codes over Z_k for k ≥ 3, higher-dimensional A_k(C), or lattices not of Construction A form. No argument establishes the contrapositive: if (1.7) holds for a lattice Λ, then Λ must be A_2(C) for some binary code C. The paper also does not clarify how to compare 'associated with a binary code' across alphabets; for example, A_3(Z_3^n) = Z^n equals A_2(Z_2^n), so the classification must be a statement about the resulting lattice, not merely about the alphabet used to construct it. Absent such a proof, the universal 'except binary' sentence overclaims. Removing or weakening that sentence leaves Theorem 4 and the refutation intact, so this is an addressable gap rather than a flaw in the main derivation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper has two main parts. The first extends MacWilliams theory in a coding-theoretic setting: Theorem 2 gives an m-tuple support weight enumerator identity for linear codes over Z_k, and Theorem 3 gives a complete weight enumerator MacWilliams identity for linear codes over the ring Z_k[xi]. The second part addresses Solé's 1995 conjecture (1.7) on nu-functions of lattices. Theorem 4 proves a new identity for lattices obtained by Construction A from ternary codes: for Lambda = A_3(C), the paper derives 3^n nu_{Lambda^*}(tanh^3(beta/2)) = det(Lambda)[((1+tanh(beta/2))(1-tanh^3(alpha/2)))/((1-tanh(beta/2))(1+tanh^3(alpha/2)))]^n nu_Lambda(tanh(alpha/2)), with the parameter relation e^{-2beta} = 3 tanh(alpha)/(8 - 5 tanh(alpha)). The paper then uses the one-dimensional zero-code lattices A_k({0}) = kZ for k = 3, ..., 10 as numerical counterexamples to (1.7) and claims that the conjecture never holds except for lattices associated with binary codes.","tokens_in":12255,"tokens_out":16110,"duration_ms":140932,"significance":"The proof of Theorem 4 is elementary, self-contained, and, as far as I checked, algebraically sound; it gives a genuine new analog of the MacWilliams identity for ternary Construction A lattices, with a parameter relation different from Solé's conjectured one. The m-tuple and complete weight enumerator identities are useful extensions of known results and are proved carefully. If the numerical counterexamples can be made rigorous, they disprove Solé's conjecture for arbitrary lattices. However, the paper's stronger 'except binary' classification claim is not established by the evidence presented, so the paper as it stands does not deliver a complete resolution of Solé's problem.","major_comments":[{"comment":"The claim that Solé's conjecture (1.7) 'never holds in the general case, except for lattices associated with a binary code' is not supported by the evidence in the manuscript. Section 4 provides counterexamples only for the one-dimensional zero codes A_k({0}) = kZ for k = 3, ..., 10. These examples show that (1.7) fails for some lattices; they do not imply that every lattice satisfying (1.7) is A_2(C) for a binary code C. To support the 'except binary' statement, the authors would need a proof of the contrapositive for all Construction A codes over Z_k with k >= 3 and for lattices outside Construction A, or at least a precise characterization of the equality cases. As written, the assertion overclaims; weakening it to 'there exist lattices for which (1.7) fails' would leave Theorem 4 intact.","section":"Abstract; §1.2; §4"},{"comment":"The numerical counterexamples are not rigorously established. The proof of the k=5 counterexample states that inequality (4.2) can be demonstrated by choosing alpha = 1 and beta ≈ 0.136, but it does not give the numerical evaluation or an error bound. Table 1 lists decimal values for k = 3, ..., 10, and in the k = 3 row the two sides differ by 0.1000 at the displayed precision; without certified interval arithmetic or an exact algebraic comparison, the table is numerical evidence rather than a proof. Since the paper's negative answer to Solé's conjecture rests on these examples, the authors should either provide a rigorous error analysis or explicitly label these as numerical counterexamples and weaken the corresponding claims.","section":"§4; Table 1"},{"comment":"The phrase 'lattices associated with a binary code' is never defined. In particular, A_3({0}) = 3Z is homothetic to A_2({0}) = 2Z, so if 'associated with a binary code' means 'up to homothety or scaling', then the k=3 row of Table 1 would not be a counterexample to the 'except binary' claim; if it means exactly Construction A_2(C), that should be stated. The classification claim is not testable until this equivalence relation is specified.","section":"§1.2"}],"minor_comments":[{"comment":"The terms 'nu-function' and 'nut function' are used interchangeably; please standardize the terminology.","section":"§1.2; §3"},{"comment":"In Lemma 2.1, the displayed formula uses the symbol a both as the summation index and as the test element, which is confusing; the intended identity should be written with separate variables.","section":"§1.1; Lemma 2.1"},{"comment":"The sentence 'For any modulo k ≥ 3, it is possible to assign a value of one' is not grammatical and should be rewritten to state clearly that the authors set beta = 1 and alpha ≈ 0.136.","section":"§4"},{"comment":"Table 1 should explicitly state that n = 1, C = {0}, alpha ≈ 0.136, beta = 1, and which parameter relation is used; the text confusingly mentions e^{-2alpha} = tanh(beta), while the conjecture uses e^{-2beta} = tanh(alpha). The relation is symmetric for these values, but the exposition should be precise.","section":"§4; Table 1"},{"comment":"Reference [29] is unpublished; since the binary case of the conjecture is attributed to it, the authors should clarify its status or provide a proof sketch in the paper.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The main derivation in Theorem 4 appears credible and is worth publishing once the overreaching 'except binary' claim is removed or properly supported. I would advise the editor to require the authors to revise the abstract and §1.2 accordingly, and to either make the numerical counterexamples rigorous or present them explicitly as numerical evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper has one genuinely new result—the ternary nu-function MacWilliams analogue with its parameter relation—and a valid refutation of Solé's 1995 conjecture. The rest is bread-and-butter finite Fourier transform. The Theorem 4 derivation is clean: lemmas 3.1–3.4 are correct, the parameter relation checks out, and the symmetry property is nice.\n\nThe counterexamples are simple: Λ = A_k({0}) = kZ for k = 3..10, one-dimensional. That is enough to disprove the conjecture in full generality, since the conjecture claimed to hold for arbitrary lattices. The table has no error bounds, but the inequalities are not close, and the exact expression for ν_Z shows the gap is real. Fine as a counterexample.\n\nWhere it overreaches: the abstract and Section 1.2 claim the conjecture 'never holds in the general case, except for lattices associated with a binary code.' That exclusivity is not proven. The zero-code family is a single slice; it says nothing about nonzero codes over Z_k, higher-dimensional constructions, or lattices not of Construction A form. Also, 'associated with a binary code' is ambiguous across alphabets: A_3(Z_3^n)=Z^n=A_2(Z_2^n), so the classification must be a statement about the lattice, not the alphabet. The reader's assessment on this is right. Removing the 'except binary' sentence leaves Theorem 4 and the refutation intact; the gap is addressable.\n\nTheorems 2 and 3 are routine: m-tuple support enumerator via Poisson summation over a matrix ring, and complete weight enumerator over Z_k[ξ] as a finite Fourier transform. They are correct but not new in any deep sense; the paper does not cite Kaplan for m-tuple weight enumerators, which is a small gap.\n\nCitation pattern is fine. The self-citation [29] for the binary case is not load-bearing here.\n\nWho this is for: coding theorists and people working on the code-lattice dictionary, especially anyone using Solé's conjecture. It deserves a serious referee: the central theorem is correct and the counterexamples settle the conjecture's general form. I would recommend conditional acceptance after the authors either prove the 'except binary' classification or weaken it to a statement about the provided counterexamples.","headline":"Ternary nu-function identity is a genuine new result and the counterexamples refute Solé's conjecture as stated, but the universal 'except binary' claim overreaches.","tokens_in":12786,"tokens_out":1477,"would_cite":true,"duration_ms":13399,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94B05","11H06"],"pacs":[],"model":"deepseek-v4-flash","headline":"Lattices from ternary codes obey a new dual-counting identity, and the 1995 conjecture fails there.","keywords":["MacWilliams identity","weight enumerator","codes over Z_k","nu-function","lattices","ternary codes","Construction A","1995 lattice conjecture"],"falsifier":"Take the nonzero ternary code $C=\\{00,11,22\\}\\subset\\mathbb{Z}_3^2$, so that $\\Lambda=A_3(C)=\\{(a,b)\\in\\mathbb{Z}^2: a\\equiv b\\pmod{3}\\}$ has index 3 and is not a binary Construction A lattice. Numerically evaluate both sides of the original 1995 identity (1.7) at $\\alpha=1$, with $\\beta\\approx 0.136$ determined by $e^{-2\\beta}=\\tanh(\\alpha)$: if the two sides agree, the paper's 'except binary' claim is false; if they differ, the failure extends beyond zero codes to a genuinely ternary lattice.","tokens_in":11793,"feed_emoji":"🔢","tokens_out":19113,"duration_ms":158577,"temperature":0.7,"pith_summary":"The paper generalizes MacWilliams' identity on two fronts and then supplies a long-missing lattice analogue: a dual-counting identity for the nu-function, which enumerates lattice points by $L^1$ norm. Its main result, Theorem 4, gives a closed-form identity for the nu-function of any lattice built by Construction A from a ternary code, with a parameter relation $e^{-2\\beta}=3\\tanh(\\alpha)/(8-5\\tanh(\\alpha))$ instead of the relation conjectured in 1995. The proof partitions the ternary lattice into residue classes modulo 3, identifies the dual lattice as one-third of the dual-code lattice, and converts the lattice sum into an ordinary code weight-enumerator duality. The authors also tabulate numerical failures of the 1995 conjecture for the zero-code lattices $A_k(\\{0\\})$ with $k=3$ to $10$, and conclude that, combined with their earlier binary-code result, the original conjecture holds only for lattices from binary codes.","feed_headline":"Ternary-code lattices break a 1995 conjecture","feed_subtitle":"A corrected dual-counting formula replaces the original guess, which survives only for binary-code lattices.","key_machinery":"The nu-function $\\nu_\\Lambda(z)=\\sum_{x\\in\\Lambda}z^{|x|_1}$, which counts lattice points by $L^1$ norm, is the object under study; the carrying mechanism is Construction A, the lattice $A_k(C)=\\{x\\in\\mathbb{Z}^n: x\\equiv c\\pmod{k}\\text{ for some }c\\in C\\}$. Four lemmas do the work: $\\nu_{3L}(z)=\\nu_L(z^3)$; $A_3(C)^*=(1/3)A_3(C^\\perp)$; $\\nu_{A_3(C)}(\\tanh(\\alpha/2))=\\left(\\frac{1+\\tanh^3(\\alpha/2)}{1-\\tanh^3(\\alpha/2)}\\right)^n W_C\\left(\\frac{\\tanh\\alpha}{2-\\tanh\\alpha}\\right)$, with $W_C$ the Hamming weight enumerator; and $|C|\\det(A_3(C))=3^n$. The parameter relation $e^{-2\\beta}=3\\tanh(\\alpha)/(8-5\\tanh(\\alpha))$ is load-bearing: it is chosen so that the ternary MacWilliams substitution maps the weight-enumerator argument correctly, and it is symmetric, so the identity holds in both dual directions.","core_discovery":"The central claim is an exact $L^1$-sphere counting identity for lattices from ternary codes. For $\\Lambda=A_3(C)$, Theorem 4 states $$3^n \\nu_{\\Lambda^*}(\\$tanh^{3}$(\\$\\beta$/2))=\\det(\\Lambda)\\left[\\frac{(1+\\tanh(\\$\\beta$/2))(1-\\$tanh^{3}$(\\$\\alpha$/2))}{(1-\\tanh(\\$\\beta$/2))(1+\\$tanh^{3}$(\\$\\alpha$/2))}\\right]^n \\nu_\\Lambda(\\tanh(\\$\\alpha$/2)),$$ where $e^{-2\\beta}=3\\tanh(\\alpha)/(8-5\\tanh(\\alpha))$, a relation that is symmetric between $\\alpha$ and $\\beta$. The proof rests on three structural facts: multiplying a lattice by 3 changes the nu-function by $z\\mapsto z^3$; the dual of $A_3(C)$ is $(1/3)A_3(C^\\perp)$; and for residue classes modulo 3, the nu-function of $A_3(C)$ factors into a hyperbolic-tangent factor times the Hamming weight enumerator of $C$. The final step applies the MacWilliams identity for ternary codes, so the lattice identity is exactly as strong as the classical code identity. Against the 1995 conjecture, the paper reports numerical disagreement for the zero-code lattices $A_k(\\{0\\})$, $k=3$ to $10$, and asserts that the original conjecture survives only in the binary-code case.","pith_inferences":["The categorical 'never except binary' conclusion is stronger than the tabulated evidence: the displayed counterexamples are all zero codes $A_k(\\{0\\})$ for $k=3$ to $10$, and no argument shows those examples cover nonzero ternary codes or all moduli $k$. A cautious reader should treat the refutation as established for this family, and the broader statement as a supported conjecture.","Because $A_k(\\{0\\})$ is just the one-dimensional arithmetic progression $k\\mathbb{Z}$, the numerical failures in the table test the parameter relation, not the code structure; testing nonzero codes in higher dimensions (such as the length-2 ternary repetition code) would show whether the obstruction is genuinely about the alphabet or only about the modulus.","The symmetric rational relation between $\\tanh\\alpha$ and $\\tanh\\beta$ suggests that each modulus $k$ may carry its own MacWilliams-style nu identity, with the binary case as the only one coinciding with the 1995 form; deriving such a family would be a natural next step."],"forward_implications":["For every ternary linear code $C$, the dual lattice $A_3(C)^*$ has an exact $L^1$-sphere enumerator expressed through $C$'s weight enumerator, so those point counts can be obtained without summing over the dual lattice.","Because the $\\alpha\\leftrightarrow\\beta$ relation is symmetric, Theorem 4 behaves like a true duality: applied twice, it returns the original lattice with the transformed parameter.","The generalized $m$-tuple identity over $\\mathbb{Z}_k$ extends the earlier single-code identity to products of $m$ codes, and the complete weight-enumerator identity over $\\mathbb{Z}_k[\\xi]$ covers Galois rings as a special case.","If the negative conclusion is accepted, any future MacWilliams-type lattice identity for nonbinary alphabets must carry code-dependent parameter relations rather than the single form guessed in 1995."],"supporting_citations":[{"why":"It is the source of the 1995 conjecture and supplies the exact identity (1.7) that Theorem 4 replaces and Section 4 tests numerically.","marker":"[22]"},{"why":"It is the authors' prior result showing the original conjecture holds for lattices associated with a binary code, and it defines the one surviving family.","marker":"[29]"},{"why":"It is the earlier single-code MacWilliams identity over Z_k that Theorem 2 generalizes to m-tuple support enumerators.","marker":"[9]"},{"why":"It is the Galois-ring MacWilliams identity that Theorem 3 extends to complete weight enumerators over Z_k[ξ].","marker":"[24]"},{"why":"It is the original weight-enumerator theorem used in the final step of Theorem 4 to relate the dual code's weight enumerator to C's.","marker":"[13]"}],"fun_headline_variants":["1995 lattice conjecture fails for ternary codes","Ternary lattices break Solé's 1995 conjecture","New formula for nu-function on ternary lattices","Lattice MacWilliams identity corrected beyond binary","Solé conjecture only holds for binary-code lattices"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's exclusive conclusion—that the 1995 conjecture holds only for lattices from binary codes—rests on finitely many zero-code examples $A_k(\\{0\\})$ for $k=3$ to $10$, with no demonstrated argument that those examples rule out all nonzero ternary codes or all other Construction A lattices; if even one such lattice satisfied the original relation, the exclusive claim would collapse.","fun_headline_variants_meta":{"raw":{"variants":["1995 lattice conjecture fails for ternary codes","Ternary lattices break Solé's 1995 conjecture","New formula for nu-function on ternary lattices","Lattice MacWilliams identity corrected beyond binary","Solé conjecture only holds for binary-code lattices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000214,"raw_usage":{"total_tokens":1447,"prompt_tokens":986,"completion_tokens":461,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":389}},"tokens_in":602,"tokens_out":461,"duration_ms":4566,"temperature":1.0,"reasoning_tokens":389,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:37:04.971696+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the nonzero ternary code $C=\\{00,11,22\\}\\subset\\mathbb{Z}_3^2$, so that $\\Lambda=A_3(C)=\\{(a,b)\\in\\mathbb{Z}^2: a\\equiv b\\pmod{3}\\}$ has index 3 and is not a binary Construction A lattice. Numerically evaluate both sides of the original 1995 identity (1.7) at $\\alpha=1$, with $\\beta\\approx 0.136$ determined by $e^{-2\\beta}=\\tanh(\\alpha)$: if the two sides agree, the paper's 'except binary' claim is false; if they differ, the failure extends beyond zero codes to a genuinely ternary lattice.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It is the source of the 1995 conjecture and supplies the exact identity (1.7) that Theorem 4 replaces and Section 4 tests numerically."},{"cited_title":"Zheng, F","cited_arxiv_id":null,"evidence_quote":"It is the authors' prior result showing the original conjecture holds for lattices associated with a binary code, and it defines the one surviving family."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It is the earlier single-code MacWilliams identity over Z_k that Theorem 2 generalizes to m-tuple support enumerators."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It is the Galois-ring MacWilliams identity that Theorem 3 extends to complete weight enumerators over Z_k[ξ]."},{"cited_title":"MacWilliams","cited_arxiv_id":null,"evidence_quote":"It is the original weight-enumerator theorem used in the final step of Theorem 4 to relate the dual code's weight enumerator to C's."}],"review_version":1}