{"id":"9cb56504-c0c8-4190-bcf2-bf6a28862783","arxiv_id":"2507.03178","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper defines a generalized theta series for lattices and uses it, together with numerical calculations, to claim counterexamples to the secrecy gain conjecture for isodual lattices.","lead":"The paper introduces a new lattice invariant, the generalized theta series, and reports counterexamples to an open conjecture about the secrecy gain of isodual lattices. The counterexamples rest on numerical computations detailed in the authors' earlier work, and the definition of the new invariant is not fully specified.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Definition 6 is under-specified: the summation index and the radius parameter m in Eq. (1) are never quantified, so the generalized theta series is not a well-defined lattice invariant and the printed coefficients cannot be reproduced.","rationale":"The reader's weakest assumption correctly identifies Definition 6 as the load-bearing point. My analysis confirms that the problem is not merely notational: Eq. (1) as written contains an undefined index m in the radius bound, and the explanatory text makes m serve double duty as both the radius parameter and the term counter. Consequently, the series is not yet a function of z, and the claimed invariant properties and applications rest on an object that has not been specified. The counterexample to the secrecy gain conjecture is a separate claim; it could in principle be checked directly by computing the theta series of the Construction-A lattice in Example 5, but the paper outsources that computation to [13] without giving enough data to verify the claimed Delta(1) > 1 or the location of the maximum. Both problems point in the same direction: the paper's strongest advertised results are not verifiable from the manuscript alone. I therefore agree with the REJECT verdict and see no reason to soften it without a repaired definition and independently checkable computations.","tokens_in":9423,"tokens_out":10199,"duration_ms":134002,"concrete_test":"Run the following check on the hexagonal lattice A2: fix r=2 and enumerate all unordered pairs of linearly independent vectors lying in B(2*lambda_1^(m)) for each m in {1,...,12}. For each m, compute the multiset of exponents det(A A^T) for all such pairs. If the coefficient of q^(3/4) (or any other exponent) differs as m varies, then Eq. (1) does not define a unique series without an additional convention. Then ask the authors to restate Definition 6 with an explicit summation index (e.g., a sum over m together with a first-occurrence rule) and recompute the A2 coefficients printed in Example 2 from that restated definition. If the coefficients cannot be regenerated, the invariant is not well-posed.","verdict_should_be":"REJECT","load_bearing_attack":"The central new object, the generalized theta series, is not actually defined. In Eq. (1), the sum ranges over sets of r linearly independent vectors in the ball B(r*lambda_1^(m)), but there is no summation over m and no other quantification of m. The surrounding text says lambda_1^(m) is the length of the m-th shortest vector and then explains that the m-th term of the series is the m-th smallest sublattice volume inside that same ball. This makes m simultaneously a radius index and a term index, which is circular: the set of vectors inside the ball determines the list of volumes, while the radius is chosen from that same list. Without an explicit 'first occurrence' or diagonal convention, the same vector set contributes to multiple candidate m values as the radius grows, and no rule is given for choosing the defining one. The displayed coefficients in Examples 2 and 4 therefore cannot be independently recovered from the manuscript. This is not a stylistic gap: the paper's advertised applications, stable-lattice detection and lattice hearing, both depend on the generalized theta series being a genuine isometry invariant. The authors themselves flag in Example 4 that several coefficients were obtained numerically and are not guaranteed by Lemma 1, which further indicates that the higher terms are not pinned down by the definition. The counterexample to the secrecy gain conjecture in Example 5 does not directly use the generalized theta series, but it is supported only by a numerical value outsourced to the authors' prior paper [13], with no error bound or released computation. Thus the main novel contribution is unverifiable as written, and the counterexample, even if true, does not repair the invariant.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a lattice analog of the generalized Hamming weight of linear codes, called the generalized theta series, together with the r-th generalized Euclidean norm (r-th minimum sublattice volume). The authors claim this series is a new lattice invariant with applications to stable-lattice detection and to the lattice isomorphism problem. They also present examples of isodual (and hence formally unimodular) lattices for which the theta series ratio exceeds 1 at the symmetry point tau=1, and claim this disproves the secrecy-gain conjecture for isodual lattices [14, Conj. 1] and its formal-unimodular generalization [12, Conj. 37].","tokens_in":9725,"tokens_out":9351,"duration_ms":105635,"significance":"If the generalized theta series were a well-defined isometry invariant, it would be a genuinely useful new tool, connecting the determinant and the theta series and potentially distinguishing non-isometric lattices that share the same ordinary theta series. The paper is commendably transparent that some displayed coefficients are numerical and not covered by Lemma 1, and it correctly credits [22] for the determinantal minima. The counterexample to the secrecy-gain conjecture for isodual lattices, if verified, would be a significant negative result for a conjecture that has been open for a decade. However, the central definition is not well posed as written, and the main counterexample is not self-contained, so the advertised claims are not presently established.","major_comments":[{"comment":"The summation index m is not quantified in (1): the radius B(r*lambda_1^(m)) contains a free parameter m, but no summation over m or convention for choosing m is given, while the accompanying text identifies m with the position of a term in the ordered list of sublattice volumes. Since the radius is chosen from the same ordering of volumes it is supposed to define, the series is circular as written. Consequently the coefficients displayed in Examples 2 and 4 cannot be independently recovered from Definition 6, and the authors' own caveat in Example 4 that the boldfaced coefficients are only numerical and not guaranteed by Lemma 1 confirms that the higher terms are not pinned down. Because isometry invariance is asserted only for the (undefined) series, the applications in Section V-A to stable-lattice detection and to lattice hearing are not supported until Definition 6 is replaced by a well-posed definition with an explicit enumeration rule or diagonal convention.","section":"Definition 6, Eq. (1)"},{"comment":"The discussion of Conjecture 1 is logically misplaced. Conjecture 1 is quantified over stable lattices, but both Lambda_{A4}(C3) and Lambda_{A4}(C4) are shown to be non-stable in the same examples. The sentence in Example 5 that Delta_Lambda(1)>1 'demonstrates that Conjecture 1 is not true for isodual lattices' is therefore misleading, and Example 6's statement that Delta_Lambda(tau)>1 'reveals that Conjecture 1 does not hold for this isodual lattice' is not a violation of the conjecture as stated. The authors should either remove these remarks or clearly state that they concern a different, non-stable class of lattices.","section":"Section V-B, Examples 5 and 6"},{"comment":"The central counterexample to Conjecture 2, and hence to [12, Conj. 37], is not self-contained. The key numerical fact Delta_{Lambda_{A4}(C3)}(1) approx 1.0026 is only referenced to [13], and the isoduality of the construction is asserted via [24, Lemma 2.4] and [25, p. 378] without reproducing the relevant details. Since the paper's main advertised result is the invalidation of the secrecy-gain conjecture, the manuscript should provide a checkable computation: for instance, the exact theta series coefficients used to plot Fig. 3, or a rigorous inequality Delta_{Lambda_{A4}(C3)}(1) > 1 derived from a finite initial segment with a bound on the tail.","section":"Example 5"},{"comment":"The sentence 'Given that formally unimodular lattices are also isodual' states the implication in the wrong direction; the true implication is that every isodual lattice is formally unimodular, and formally unimodular lattices need not be isodual. The extension of the counterexample to [12, Conj. 37] should be justified by the correct inclusion isodual subset formally unimodular, as is done later in Example 5, not by the stated inclusion.","section":"Section II-B, Preliminaries"}],"minor_comments":[{"comment":"There is an extra space before the period in 'generalized theta series .'.","section":"Abstract"},{"comment":"The notation for the scaled lattice in Example 3 is visually unclear; please specify clearly which generator matrix is used for the equivalent lattice and how the scaling factor alpha enters Proposition 1.","section":"Example 3"},{"comment":"The caption 'a = -8, n = 5' is cryptic; the parameters a and n should be explained so that the geometric meaning of the figure is self-contained.","section":"Fig. 2 caption"},{"comment":"The notation Lambda_{A4}(C3) is confusing because A4 may be read as the lattice A_4; consider writing Lambda_{A_4}(C3) or Lambda_{Z_4}(C3).","section":"Notation throughout Section V"},{"comment":"References [5] and [19] are the same work by Regev and Stephens-Davidowitz; please unify the citation or explain why both versions are needed.","section":"References [5] and [19]"}],"recommendation":"major_revision","confidential_remarks":"The primary blocker is the ill-posed Definition 6: until the summation convention for m is fixed, the new invariant cannot be used as the basis for the applications. The false implication about formally unimodular lattices and the misplaced comments about Conjecture 1 suggest that Section V-B needs a careful rewrite. I would also ask the editor to confirm that the numerical counterexample in Example 5 is fully substantiated in [13], since the present manuscript does not provide enough information to verify the main claimed disproof independently."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper has a genuinely interesting idea—a lattice analog of the generalized Hamming weight—but as written the central object, the generalized theta series, is not well-defined. Definition 6 (Eq. 1) sums over r-tuples of independent vectors in a ball B(rλ_1^(m)), but m is never quantified. The text says λ_1^(m) is the m-th shortest vector length and that the m-th term of the series corresponds to the m-th smallest sublattice volume inside that same ball. That is circular: the radius is chosen from the same ordering it is supposed to define. Without an explicit first-occurrence or diagonal convention, the same sublattice will be counted for multiple m as the ball grows, and the coefficients in Examples 2 and 4 cannot be independently recovered. The authors even note in Example 4 that some coefficients were obtained numerically and are not guaranteed by Lemma 1—which is a red flag that the definition is not pinning down the series.\n\nWhat is genuinely new: the proposal to organize sublattice volumes into a generating function, and the bridge it builds between coding-theoretic generalized weights, the r-DSP, and stable-lattice detection. The paper is also honest about the relation between the r-th generalized Euclidean norm and the existing determinantal minima of Dadush, and the counterexample to the secrecy gain conjecture would be a significant result if it holds up.\n\nThe counterexamples are separate from the new invariant—they use ordinary theta series ratios. But they rely on numerical values outsourced to a self-cited prior paper [13], with no error bounds and no released computation. A plot and a single value Δ(1)≈1.0026 are not a proof of a global minimum claim. That is a real soft spot, though secondary to the definition problem.\n\nWho should read this: lattice theorists and coding theorists interested in wiretap lattices and the lattice isomorphism problem. The idea is worth pursuing, but the paper in its current form is not usable: the main invariant is unverifiable. I would send it to a serious referee rather than desk-reject, because the idea is new and the conjecture counterexample is important. But I would not accept it without a major revision that makes Definition 6 rigorous and provides verifiable computations.","headline":"A promising but under-specified new lattice invariant; the counterexample is potentially important but not proven.","tokens_in":10239,"tokens_out":6088,"would_cite":false,"duration_ms":66268,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper introduces a sublattice-volume theta series and uses it to disprove the secrecy gain conjecture for isodual lattices.","keywords":["generalized theta series","lattice invariant","generalized Euclidean norm","isodual lattices","secrecy gain conjecture","stable lattices","lattice isomorphism problem","densest sublattice problem"],"falsifier":"An independent enumeration of the sublattices of Λ_{A_4}(C_3) could settle the disproof: if the second generalized $\\theta$ series does not contain the term $144 q^{3/4}$ with the stated multiplicity, or if $\\Delta_{\\Lambda_{A_4}(C_3)}(1)$ computes to at most 1, the counterexample to the secrecy gain conjecture collapses.","tokens_in":9237,"feed_emoji":"📐","tokens_out":9448,"duration_ms":95564,"temperature":0.7,"pith_summary":"This paper introduces a lattice invariant called the generalized theta series, which counts the squared volumes of r-dimensional sublattices of a lattice instead of just the lengths of its vectors. The authors use it to identify stable lattices, to separate non-isomorphic lattices that have identical ordinary theta series, and to construct explicit isodual lattices whose theta series ratio exceeds one or whose symmetry point is a maximum rather than the global minimum. Those examples refute the secrecy gain conjecture for isodual lattices and for formally unimodular lattices, which had asserted that the ratio relative to the integer lattice is minimized at the symmetry point τ = 1. If the computations hold, the paper replaces a universality claim with concrete exceptions and gives a new invariant for studying lattice structure.","feed_headline":"Sublattice theta series breaks secrecy gain conjecture","feed_subtitle":"Counting sublattice volumes exposes isodual lattices whose secrecy ratio exceeds 1 and peaks at τ=1.","key_machinery":"The key object is the r-th generalized $\\theta$ series, a lattice analogue of the generalized Hamming weight enumerator. It is defined as $\\Theta_\\Lambda^{(r)}(z)=\\sum_{\\{a_i\\}_{i=1}^r\\subseteq\\Lambda\\cap B(r\\lambda_1^{(m)}),\\,\\mathrm{rank}(A)=r} q^{\\det(AA^T)}$ with $q=e^{i\\pi z}$ and $A^T=[a_1^T,\\ldots,a_r^T]$, so each term contributes the squared volume of the generated sublattice and each unordered set is counted once. The argument is carried by the leading exponent of each series, the r-th generalized Euclidean norm $\\nu_r(\\Lambda)$, the minimum squared volume of an r-dimensional sublattice, which connects the invariant to the r-dimensional densest sublattice problem and lets the authors compute enough terms to pin down the ratio $\\Delta_\\Lambda(\\tau)$ for their counterexamples.","core_discovery":"The central discovery is that the secrecy gain conjecture fails for isodual lattices and hence for formally unimodular lattices. The paper exhibits two six-dimensional Construction A lattices over Z4, obtained from isodual codes, and computes enough of their first and second generalized $\\theta$ series to evaluate the $\\theta$ series ratio ΔΛ(τ) = ΘΛ(iτ)/Θ_{Z^n}(iτ). For Λ_{A_4}(C_3), the ratio at τ = 1 is about 1.0026, strictly greater than one, and τ = 1 is the symmetry point where the ratio reaches its maximum; for Λ_{A_4}(C_4), the ratio still exceeds one for some τ even though its minimum sits at τ = 1. These volume-one isodual lattices therefore invalidate the conjecture that the global minimum of the ratio is always at τ = 1, as stated in the isodual and formally unimodular versions.","pith_inferences":["Editorial inference: if the generalized theta series is as sensitive as the examples suggest, it gives a natural fingerprint for the lattice isomorphism problem, though the paper itself notes the computations are too costly to threaten cryptographic schemes.","Editorial inference: the role of the index m in the ball radius rλ_1^{(m)} is not fully pinned down in the paper; a direct check of whether the series is independent of the chosen ordering of shortest vectors would decide whether coefficients beyond the leading term are canonical.","Editorial inference: the counterexamples suggest that the secrecy gain conjecture might survive only for restricted subfamilies such as stable or extremal lattices; a systematic scan of all isodual lattices in small dimension would show how often τ = 1 is actually the minimum."],"forward_implications":["The invariant refines audibility: two lattices with the same ordinary theta series can be told apart when their generalized theta series differ, which the paper demonstrates for a pair of Construction A lattices from codes.","Computing the leading term of each generalized theta series solves the r-dimensional densest sublattice problem, giving a computational handle on sublattice volumes.","The generalized Euclidean norm hierarchy gives a necessary condition for stability, since a stable lattice must have all r-dimensional minimum sublattice volumes at least one.","The isodual and formally unimodular counterexamples mean the secrecy gain claim cannot be used as a universal design guarantee for wiretap lattice codes from these families, and any such guarantee must be argued case by case or under extra hypotheses.","For the exhibited lattices, the symmetry point of the theta series ratio is not always its minimum, so the weak secrecy gain at τ = 1 does not automatically bound the eavesdropper's error probability in the intended way."],"supporting_citations":[{"why":"Defines the generalized Hamming weight for linear codes, the analogue that motivates the new lattice invariant.","marker":"[1]"},{"why":"Defines the r-dimensional densest sublattice problem and supplies Algorithm 1 used to compute the generalized Euclidean norms in the examples.","marker":"[15]"},{"why":"The isodual secrecy gain conjecture that the paper's Example 5 counterexample refutes.","marker":"[14]"},{"why":"The formally unimodular generalization of the conjecture, also invalidated by the paper's counterexample.","marker":"[12]"},{"why":"Provides the earlier numerical secrecy gain calculation cited for Δ_{Λ_{A_4}(C_3)}(1) ≈ 1.0026.","marker":"[13]"},{"why":"Introduced the original secrecy gain conjecture for unimodular lattices that the isodual extension builds on.","marker":"[7]"},{"why":"Establishes that the code C_3 is an isodual bordered double circulant code, a needed step in showing the lattice is isodual.","marker":"[24]"},{"why":"Supplies the fact that construction from an isodual code over Z4 yields an isodual lattice.","marker":"[25]"}],"fun_headline_variants":["New lattice invariant disproves secrecy gain conjecture","Isodual lattices break secrecy gain conjecture","Secrecy gain conjecture fails for isodual lattices","Generalized theta series refutes isodual secrecy gain","Counterexample in isodual lattices ends secrecy gain"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the generalized theta series is a well-defined lattice invariant, meaning each term counts an unambiguous set of r-dimensional sublattices inside the stated ball and the result does not depend on arbitrary choices of vectors or the ordering of shortest vectors.","fun_headline_variants_meta":{"raw":{"variants":["New lattice invariant disproves secrecy gain conjecture","Isodual lattices break secrecy gain conjecture","Secrecy gain conjecture fails for isodual lattices","Generalized theta series refutes isodual secrecy gain","Counterexample in isodual lattices ends secrecy gain"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00067,"raw_usage":{"total_tokens":2988,"prompt_tokens":814,"completion_tokens":2174,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":430,"completion_tokens_details":{"reasoning_tokens":2097}},"tokens_in":430,"tokens_out":2174,"duration_ms":16802,"temperature":1.0,"reasoning_tokens":2097,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:17:48.194285+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An independent enumeration of the sublattices of Λ_{A_4}(C_3) could settle the disproof: if the second generalized $\\theta$ series does not contain the term $144 q^{3/4}$ with the stated multiplicity, or if $\\Delta_{\\Lambda_{A_4}(C_3)}(1)$ computes to at most 1, the counterexample to the secrecy gain conjecture collapses.","supporting_citations":[{"cited_title":"Generalized Hamming weights for linear codes,","cited_arxiv_id":null,"evidence_quote":"Defines the generalized Hamming weight for linear codes, the analogue that motivates the new lattice invariant."},{"cited_title":"Algorithms for the densest sub- lattice problem,","cited_arxiv_id":null,"evidence_quote":"Defines the r-dimensional densest sublattice problem and supplies Algorithm 1 used to compute the generalized Euclidean norms in the examples."},{"cited_title":"Lattice codes for the wiretap Gaussian channel: Construction and analysis,","cited_arxiv_id":null,"evidence_quote":"The isodual secrecy gain conjecture that the paper's Example 5 counterexample refutes."},{"cited_title":"Formally unimodular pack- ings for the Gaussian wiretap channel,","cited_arxiv_id":null,"evidence_quote":"The formally unimodular generalization of the conjecture, also invalidated by the paper's counterexample."},{"cited_title":"Secrecy gain of formally unimodular lattices from codes over the integers modulo 4,","cited_arxiv_id":null,"evidence_quote":"Provides the earlier numerical secrecy gain calculation cited for Δ_{Λ_{A_4}(C_3)}(1) ≈ 1.0026."},{"cited_title":"Unimodular lattices for the Gaussian wiretap channel,","cited_arxiv_id":null,"evidence_quote":"Introduced the original secrecy gain conjecture for unimodular lattices that the isodual extension builds on."},{"cited_title":"Isodual codes over Z2k and isodual lattices,","cited_arxiv_id":null,"evidence_quote":"Establishes that the code C_3 is an isodual bordered double circulant code, a needed step in showing the lattice is isodual."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the fact that construction from an isodual code over Z4 yields an isodual lattice."}],"review_version":1}