{"id":"dc0c338a-45ef-4eac-a119-0401c5657888","arxiv_id":"1908.04137","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A new reduction and signed Hermite normal form let every representation of a torsion-free arithmetic matroid be computed up to equivalence, yielding a sharpened upper bound and counterexamples to two shellability conjectures.","lead":"This paper gives an algorithm that lists all ways to build a toric arrangement from a given combinatorial skeleton, the arithmetic matroid, up to a natural equivalence. It also uses the algorithm to find examples that disprove two conjectures about the posets attached to these arrangements.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Enumeration completeness depends on external uniqueness results and Theorem 4.6; a brute-force cross-check of the implementation on small matroids would settle the main risk.","rationale":"The reader's weakest assumption correctly identifies the same load-bearing point: the completeness of the enumeration in Section 5.2 depends on Theorem 4.6 and on the cited uniqueness results for surjective representations, neither of which is fully proved inside this paper. I checked the internal logic of Theorem 5.4 and Corollary 5.5 assuming those ingredients; the determinant typo in Theorem 5.4 is minor and fixable, and Corollary 5.5's counting argument is sound once the diagonal entries are read as det(H_i)=m({1..i})/det(A_i). The paper is mathematically plausible and the examples are explicit, but because the central claim is an algorithmic completeness claim, the lack of a reproducible cross-check of the implementation leaves a genuine epistemic gap. I therefore keep the reader's CONDITIONAL verdict unchanged rather than upgrading to ACCEPT. No ad hominem concerns; the critique is about verifyability and external dependencies, not about the authors' integrity.","tokens_in":17332,"tokens_out":30834,"duration_ms":305543,"concrete_test":"Write a small verification script for the Arithmat implementation: (1) generate all rank-2 and rank-3 torsion-free arithmetic matroids with ground set size at most 6, (2) for each matroid, run the Section 5.2 algorithm to obtain the claimed list of essential representations, (3) independently enumerate all integer matrices whose entries are bounded by m(E) times the largest basis multiplicity, filter by the multiplicity conditions with Smith normal form, and quotient by the equivalence action (GL(r,Z) and column sign changes), and (4) compare the two lists. If the algorithm omits any representation or produces an invalid one, then the reduction/uniqueness chain in Theorem 5.4 and Section 5.1 has a concrete failure; if the lists match for all tested matroids, the central enumeration claim gains strong independent support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that every essential representation of a torsion-free arithmetic matroid M is equivalent to HA for a fixed representation A of the reduction and some H in Hermite normal form with det(H)=m(E)—rests on two externally cited pillars. First, Theorem 4.6 asserts that a representable M has a representable reduction Mbar; the proof is sketched but relies on Corollary 3.6, which in turn depends on [DM13, Remark 3.1] and on the strong gcd property. Second, Section 5.1 constructs the unique surjective representation using [Pag17] for the absolute values of the entries and [Len17b, Lemma 6] for the signs. If either cited result were false, or if the reduction construction failed to capture some representation of M, then the list HA could omit valid essential representations. The proof of Theorem 5.4 also contains an apparent typo in the determinant relation—the line 'm(E) = det(H)·m(E) = det(H)' is not meaningful as printed—which should read roughly m(E)=det(H')=det(H) using mbar(E)=1. Independently, the two counterexamples in Section 8 (non-Cohen-Macaulay posets) rely on Sage/Arithmat computations that are not pinned by a commit or a reproduction script, so the disproof of the conjectures is not independently checkable from the manuscript alone. None of these points is demonstrated to be a mathematical error, but they are exactly the load-bearing points that would need to hold for the central claim to be fully secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies torsion-free arithmetic matroids and their representations as integer matrices (equivalently, central toric arrangements). It introduces the strong gcd property, a new operation called reduction, and a signed Hermite normal form. The main result (Theorem 5.4) states that every essential representation of a torsion-free arithmetic matroid M is equivalent to HA, where A is the unique essential representation of its reduction Mbar and H is a matrix in Hermite normal form with det(H)=m(E). This yields Corollary 5.5: a rank-r torsion-free arithmetic matroid has at most m(E)^(r-1) essential representations up to equivalence, together with an algorithm to list them. The paper also presents a polynomial-time signed Hermite normal form algorithm, a decomposition criterion, and, as applications, two computational counterexamples to conjectures that the poset of layers and the arithmetic independence poset of toric arrangements are shellable/Cohen-Macaulay.","tokens_in":17615,"tokens_out":8425,"duration_ms":80098,"significance":"If correct, the paper gives an effective classification of representations of torsion-free arithmetic matroids, improving the earlier bound m(E)^r of [Pag17] to m(E)^(r-1) and providing a concrete algorithm with a Sage implementation. The signed Hermite normal form is a useful new algorithmic tool. The two counterexamples in Section 8 disprove widely discussed conjectures and are therefore significant for the theory of toric arrangements. The paper contains detailed combinatorial proofs in Sections 3-6 and the main results are crisp and potentially influential. However, the central enumeration theorem rests on two external uniqueness results and on the reduction theorem, and the computational disproofs are not independently reproducible from the manuscript, so the significance is conditional on those points being secured.","major_comments":[{"comment":"The determinant chain in the proof of Theorem 5.4 contains the line 'm(E) = det(H)·m(E) = det(H)', which is not correct as printed; it would force det(H)=1 or m(E)=0. The intended statement should be m(E) = det(H') = det(H), using mbar(E)=1. Since det(H)=m(E) is exactly the conclusion of the theorem, this needs to be fixed.","section":"Theorem 5.4, proof"},{"comment":"The proof that the reduction Mbar is representable is only sketched. It relies on Corollary 3.6 (which itself depends on [DM13, Remark 3.1]) and on the identification of Mbar with the arithmetic matroid M' of the reduced vectors. The step where gcd{m(B)|B basis} = |T|·|K/G| and the subsequent conclusion m'(X)=mbar(X) for all X need to be written out in detail, because Theorem 4.6 is the bridge that makes the reduction from the general case to the surjective case valid.","section":"Theorem 4.6"},{"comment":"The uniqueness of the surjective representation is imported wholesale from [Pag17] (absolute values of entries) and [Len17b, Lemma 6] (signs of entries). These results are not stated or proved in the paper, yet the completeness of the enumeration in Theorem 5.4 and the bound in Corollary 5.5 depend on them. Please either state the exact propositions and give proofs (even condensed), or clearly mark this as a black-box dependence.","section":"Section 5.1"},{"comment":"The two disproofs of the shellability conjectures depend on computer calculations (the 13 representations and the homology groups (0,Z5,Z48) and (0,Z5,Z73)) that are not reproducible from the manuscript. No script, commit identifier of the Arithmat repository, or detailed verification is provided. Since this is the paper's advertised application, a reproducible computation (e.g., a small script and the exact matrices) should be included or referenced unambiguously.","section":"Section 8"}],"minor_comments":[{"comment":"There is a typo: 'matorid' should be 'matroid'.","section":"Lemma 3.4"},{"comment":"The reduction is denoted with a bar in the text, but the displayed definition writes M = (E, rk, m) without bars, which makes the notation confusing.","section":"Definition 4.1"},{"comment":"The product formula for the number of Hermite normal form matrices with given diagonal entries is printed in a garbled way; the intended expression \\prod_i d_i^{r-i} should be typeset clearly.","section":"Corollary 5.5, proof"},{"comment":"The homology notation (0,Z5,Z48) should specify whether this is reduced homology and in which degrees.","section":"Section 8"},{"comment":"The running time is stated as O(r^{\\theta-1} n^2 (r^2+n)), but the displayed bound includes subsumed terms; please state explicitly the assumption that integer operations are O(1) and simplify the expression.","section":"Proposition 6.9"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid contribution to arithmetic matroids and toric arrangements, but it is not fully self-contained in its main enumeration theorem and the computational counterexamples are not reproducible. I recommend major revision rather than rejection: the mathematical core appears sound, and the issues are fixable within the manuscript's scope. I would also note that several results are cited from the authors' own previous work; this is appropriate in context, but the referees should verify that those external results are published or otherwise verifiable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Pagaria and Paolini give the first effective enumeration of essential representations of torsion-free arithmetic matroids. The key moves are a reduction operation that passes from a general representable arithmetic matroid to a surjective torsion-free one, and a signed Hermite normal form that makes equivalence classes of representations canonical. The bound m(E)^(r-1) improves Pagaria's earlier m(E)^r, and Section 8 delivers two concrete counterexamples to the conjectures that layer posets and arithmetic independence posets of toric arrangements are shellable. That is real content, and the paper is mostly clean.\n\nThe reduction (Theorem 4.6) and the surjective uniqueness algorithm are the load-bearing pieces. The former is proved internally but uses Corollary 3.6, which relies on [DM13, Remark 3.1] and the strong gcd property; the latter imports [Pag17] for absolute values and [Len17b, Lemma 6] for signs. These are credible citations, not circulars, and the paper says clearly what it is importing. I do not see a gap where a reader would need the target theorem to be true.\n\nThe main soft spots are two. First, in the proof of Theorem 5.4 the line 'm(E) = det(H)·m(E) = det(H)' is not meaningful as printed; it should be something like m(E) = det(H') = det(H), and this is easy to fix but should be fixed before publication. Second, the Section 8 counterexamples are computed with the Arithmat package, and the manuscript does not pin a commit or provide a reproduction script for the homology computations. Since the disproofs are the kind of finite computation that should be independently checkable, that is a real but minor reproducibility gap. I would not call it a mathematical flaw; the examples are explicit and small enough to be rechecked.\n\nWho is this for: arithmetic matroid people and toric arrangement people, and anyone using Sage for matroid computations. It deserves a serious referee. My recommendation: send it out, ask for the typo fix and a reproducibility statement for Section 8.","headline":"A solid, genuinely useful enumeration algorithm for torsion-free arithmetic matroids, with a corrected typo and a reproducibility note needed before publication.","tokens_in":18133,"tokens_out":1610,"would_cite":true,"duration_ms":16902,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B35","52C35","05E45","06A07"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that every torsion-free arithmetic matroid of rank r has at most $m(E)^{r-1}$ essential representations up to equivalence, and gives an algorithm that constructs them all.","keywords":["arithmetic matroids","torsion-free","representability","toric arrangements","signed Hermite normal form","reduction","poset of layers","Cohen-Macaulay"],"falsifier":"A concrete falsifier is to search all rank-3 torsion-free arithmetic matroids with $m(E)=2$: finding one with more than $2^2=4$ essential representations up to equivalence would refute the main bound, and a brute-force search for any representable matroid whose reduction is not representable would refute Theorem 4.6.","tokens_in":17115,"feed_emoji":"🧮","tokens_out":15196,"duration_ms":146047,"temperature":0.7,"pith_summary":"Torsion-free arithmetic matroids package the combinatorics of finite lists of integer vectors: a matroid together with integer multiplicities that record the sizes of torsion subgroups in quotients by the listed vectors. This paper establishes that, up to equivalence, a rank-r torsion-free arithmetic matroid has at most $m(E)^{r-1}$ essential representations, meaning representations in a lattice of rank r, and gives an explicit algorithm that lists them all. The key move is a reduction that collapses any such matroid to a surjective one, where the whole ground set spans the lattice and a single forced representation exists; all original representations are then obtained by multiplying that reduced representation by integer matrices in Hermite normal form of determinant $m(E)$. As an application, the authors compute all representations of a small example and find toric arrangements whose layer and independence posets are not Cohen-Macaulay, disproving two conjectures.","feed_headline":"At most $m(E)^{r-1}$ representations of a torsion-free matroid","feed_subtitle":"A reduction and a signed Hermite normal form make all representations computable and settle two poset conjectures.","key_machinery":"The two load-bearing constructions are the reduction $\\overline{M}$ of a quasi-arithmetic matroid, defined by $$\\overline{m}(X)=\\frac{\\gcd\\{m(B)\\mid B\\text{ basis and }|X\\cap B|=\\operatorname{rk}(X)\\}}{\\gcd\\{m(B)\\mid B\\text{ basis}\\}},$$ with rank unchanged, and the signed Hermite normal form: the lexicographically smallest Hermite normal form obtainable from a matrix by changing column signs. The reduction extracts the torsion-free, surjective core of a representable matroid; the signed Hermite normal form canonicalizes matrices up to left multiplication by $\\mathrm{GL}(r,\\mathbb{Z})$ and column sign changes, via a polynomial-time algorithm whose inner loop uses the fact that the orbit of each entry has size at most 4.","core_discovery":"The paper's central claim is that reduction funnels representation theory: if $A$ is an essential representation of the reduced matroid $\\overline{M}$, then every essential representation of $M$ is equivalent to $HA$ for some integer matrix $H$ in Hermite normal form with $\\det(H)=m(E)$. Consequently a torsion-free arithmetic matroid of rank $r$ has at most $m(E)^{r-1}$ essential representations up to equivalence. The reduction operation, which keeps the rank and replaces subset multiplicities by ratios of gcds over bases, preserves representability, and the surjective reduced matroid has a unique essential representation, so multiplying by the admissible $H$'s exhausts all possibilities.","pith_inferences":["Because the bound depends only on the single integer $m(E)$ and the rank, not on the finer collection of basis multiplicities, one could test whether a sharper bound holds in terms of the matroid's lattice of flats; the known small examples would be the first place to look.","If the arithmetic independence poset is indeed an invariant of the arithmetic matroid, as the paper's closing question asks, then it could be studied without choosing a representation; the 13-representation example is consistent with this, and a second example with two different independence posets would settle the question negatively.","The reduction operation is a canonical simplification with the strong gcd property, so it may be useful beyond representability, for instance in computing Tutte-like invariants or in testing whether other matroid invariants are invariant under reduction.","A census-style run of the published algorithm over all rank-3 torsion-free arithmetic matroids with small $m(E)$ could reveal whether the $m(E)^{r-1}$ bound is ever tight and, if so, on which matroids."],"forward_implications":["Any torsion-free arithmetic matroid of rank $r$ has at most $m(E)^{r-1}$ essential representations up to equivalence, a finite bound depending only on the multiplicity of the whole ground set and the rank.","Representability of a torsion-free arithmetic matroid is decidable by the paper's algorithm: failure of the sign-forcing step certifies non-representability, and success produces a candidate that is checked directly against all subset multiplicities.","The signed Hermite normal form solves the equivalence problem for representations in polynomial time, so listing non-equivalent representations is a matter of linear algebra rather than search over group elements.","For the displayed example, the 13 non-equivalent representations of one arithmetic matroid yield exactly 3 non-isomorphic posets of layers, while their arithmetic independence posets are pairwise isomorphic.","The same computations exhibit toric arrangements whose poset of layers and arithmetic independence poset have homology with $\\mathbb{Z}_5$ summands, so these posets are not Cohen-Macaulay over fields of characteristic 5 and are not shellable."],"supporting_citations":[{"why":"Fixes the absolute values of the entries of the unique surjective representation from basis multiplicities, and supplies the earlier bound $m(E)^r$ that the paper improves.","marker":"[Pag17]"},{"why":"Shows uniqueness of representations for weakly multiplicative arithmetic matroids and gives the spanning-forest procedure that determines the signs of the surjective representation.","marker":"[Len17b]"},{"why":"Introduces arithmetic matroids and the gcd property for representable torsion-free matroids, which the paper uses to derive the strong gcd property.","marker":"[DM13]"},{"why":"Provides the orientability-and-strong-gcd criterion for representability of surjective torsion-free matroids used as the algorithm's final check.","marker":"[Pag18]"},{"why":"Supplies the 13-representation example and the observation that different representations of one arithmetic matroid can have non-isomorphic posets of layers.","marker":"[Pag19]"},{"why":"Gives the background result that these posets are homology Cohen-Macaulay in all but finitely many characteristics and frames the shellability conjectures that the examples disprove.","marker":"[DD18]"},{"why":"Provides the algorithm used to compute the posets of layers of toric arrangements in the applications section.","marker":"[Len17a]"}],"fun_headline_variants":["All representations of torsion-free matroids: at most m(E)^(r-1), conjectures refuted","Reduction and Hermite normal form yield all reps of torsion-free matroids","Two conjectures on torsion-free matroid posets disproved via new algorithm","Torsion-free matroids: rep count bounded by m(E)^(r-1), all reps computable"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that no representation is lost when one passes to the reduced matroid, where the whole ground set spans the lattice: if the reduction omitted a representation, or if the forced representation of the reduced matroid were not unique, the list $HA$ would be incomplete.","fun_headline_variants_meta":{"raw":{"variants":["All representations of torsion-free matroids: at most m(E)^(r-1), conjectures refuted","Reduction and Hermite normal form yield all reps of torsion-free matroids","Two conjectures on torsion-free matroid posets disproved via new algorithm","Torsion-free matroids: rep count bounded by m(E)^(r-1), all reps computable"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001131,"raw_usage":{"total_tokens":4595,"prompt_tokens":733,"completion_tokens":3862,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":349,"completion_tokens_details":{"reasoning_tokens":3764}},"tokens_in":349,"tokens_out":3862,"duration_ms":30362,"temperature":1.0,"reasoning_tokens":3764,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:50:56.872546+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete falsifier is to search all rank-3 torsion-free arithmetic matroids with $m(E)=2$: finding one with more than $2^2=4$ essential representations up to equivalence would refute the main bound, and a brute-force search for any representable matroid whose reduction is not representable would refute Theorem 4.6.","supporting_citations":[],"review_version":1}