{"id":"31ee4f11-d4cf-44ce-a5ea-0822fb773e5a","arxiv_id":"2607.13783","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Matroid correspondences define functors between matroid poset categories and package many standard matroid operations as instances of one intersection-and-delete construction.","lead":"Matroid correspondences are a new bridge construction: a fixed matroid on a disjoint union turns every matroid on one side into a matroid on the other, functorially, and reproduces deletion, contraction, truncation, intersection, union, and pullback. The paper extends this to polymatroids and connects it to Lorentzian linear operators.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6.4(4)'s contraction bases coincide with its deletion bases, so the stated polymatroid contraction correspondence is false as written; e.g., it produces (M\\e)^* instead of (M/e)^*.","rationale":"I read the paper's central construction as sound: Definitions 1.1 and 6.1 are elementary, Proposition 2.3's functoriality and Theorem 7.3's support computation are credible, and the zero-image mismatch is openly disclosed. However, Theorem 6.4(4) is a positive claim in the abstract's list of standard functors, and it is internally inconsistent with the matroid case. The printed condition γ_i=α_i duplicates deletion and cannot implement contraction. This is not a matter of contesting a theorem's proof; it is a direct failure of the statement under the paper's own definitions. The omitted proof in Theorem 6.4 converted what might have been a fixable typo into an undetected false assertion. I therefore keep the reader's CONDITIONAL verdict but shift the burden: the paper should not be accepted until Theorem 6.4(4) is corrected (likely γ_i=0) and proved, or contraction is removed from the list. The zero-image issue identified by the reader remains a scope limitation, not my primary concern.","tokens_in":14941,"tokens_out":24007,"duration_ms":226689,"concrete_test":"With α=(1,1,1,1), take the matroid M on {1,2,3,4} with bases {1,2} and {3,4}, e=1. Compute C_{/1}^*(M) using the base set in Theorem 6.4(4) and compare with (M/1)^*, whose unique basis on {2*,3*,4*} is {3*,4*}; the printed formula gives (M\\1)^* with unique basis {2*}. This single calculation settles whether the contraction statement is false as stated; if it fails, replace γ_i=α_i by γ_i=0 and re-prove the theorem.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The reader's zero-image caveat is real but self-acknowledged; the more serious problem is unacknowledged and internal. In Theorem 6.4(4), the 'contraction' correspondence C_{/e} is given bases {(γ, α−γ) : 0≤γ≤α, γ_i=α_i}. This is the same condition as deletion (3). Contracting e* in C_id would require a basis of C_id to contain all α_i copies of e*, i.e. α_i−γ_i=α_i, hence γ_i=0, and then to remove those copies; γ_i=α_i describes deleting e* (or contracting the input copy e), not contracting e*. For α=(1,...,1), the printed C_{/e} and C_{ze} have identical bases, so C_{/e}^*(M) = (M\\e)^*, not (M/e)^*. Concretely, for M on {1,2,3,4} with bases {1,2} and {3,4} and e=1, (M\\1)^* has unique basis {2*}, while (M/1)^* has unique basis {3*,4*}; the theorem's formula gives the former. The proof of Theorem 6.4 is omitted, so the error is not caught. Because Theorem 6.4 is a central component of the claimed polymatroid generalization, this is a load-bearing correctness issue; if the condition is a typo for γ_i=0, the theorem is likely repairable, but as written it is false.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces matroid/polymatroid correspondences as categorical constructions: a matroid C on E1∪E2 defines a functor C^* from Mat(E1) to Mat(E2) by (M⊕B_{E2})∩C then deletion of E1. It proves functoriality (Prop. 2.3), composition of correspondences (Prop. 2.6), and preservation of representability/algebraicity (Props. 2.5, 5.6, 6.3). It realizes deletion, contraction, free extension, truncation, intersection, union, pullback, and constant maps as correspondences (Thm. 3.1, Cor. 3.2, Thm. 3.3), and extends the construction to caged polymatroids (Sec. 6), including compatibility with multisymmetric lifts (Prop. 6.2) and a relation to supports of linear operators with Lorentzian symbols (Thm. 7.3). The paper is concise and mostly elementary, with the main conceptual contribution being a combinatorial analogue of algebraic correspondences.","tokens_in":15354,"tokens_out":60096,"duration_ms":480055,"significance":"If the main results are correct, the framework gives a uniform way to encode many standard matroid operations as functors between poset categories, with a meaningful dictionary to Lorentzian polynomials and linear operators. The functoriality proof and the composition formula are self-contained and the preservation statements are useful. The paper is honest about the zero-image limitation in the Lorentzian correspondence, and the connection in Section 7 is a natural formal dictionary. However, the polymatroid generalization currently contains a concrete false statement in the contraction correspondence, so the significance is contingent on correcting that part and supplying the missing verification.","major_comments":[{"comment":"The paper acknowledges that zero-image operators are not captured and instead produce a polymatroid of larger rank. This is a real limitation of the claimed analogy, and it is good that it is stated, but the discussion in the introduction could mention that Theorem 7.3 only applies in the nonvanishing case and that the larger-rank phenomenon is not studied further.","section":"Introduction, zero-image caveat"}],"minor_comments":[{"comment":"There is a typo in the cage in item (4): 'α_N' should presumably be 'α_n'.","section":"Theorem 6.4(4)"},{"comment":"The paper uses 'matroid intersection' in the sense of the dual of matroid union (as in [GHM25, Definition 5.10]), not the common-independent-set intersection, which is not generally a matroid. This should be stated explicitly at first use to avoid confusion.","section":"Definition 1.1 / Section 4"},{"comment":"The step identifying Supp(sym(T)^*(f y^β)) with the cap product should be expanded to mention nonnegativity of coefficients and the effect of setting x=0; as written it is a gap in the proof.","section":"Section 7, proof of Theorem 7.3"}],"recommendation":"major_revision","confidential_remarks":"The contraction error in Theorem 6.4(4) is a clear false statement, but it appears to be a typo (γ_i=α_i should be γ_i=0) and is likely repairable. The more serious issue is the omitted proof of Theorem 6.4, which allowed this error to pass. I would ask for a full proof or at least detailed derivations for the polymatroid basis sets before acceptance. The rest of the paper's core construction and examples seem consistent with the correct matroid-intersection convention."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper's central definition of matroid correspondence is a good idea and the functoriality works. The matroid half is basically correct. But Theorem 6.4(4) is wrong as stated: its contracted polymatroid has the same bases as the deletion case, so C*_{/e}(P) comes out as (P\\e)* rather than (P/e)*. The fix is to change the condition γ_i=α_i to γ_i=0. Since the proof of the theorem is omitted, the typo is not caught; that is the piece that needs attention.\n\nWhat is new: the notion of matroid correspondence is genuinely new relative to the cited literature, and showing that deletion, contraction, free extension, truncation, intersection, union, and pullback all arise as instances is more than a repackaging. Proposition 2.3 is elementary and correct, and the composition formula in Proposition 2.6 checks out. The preservation results for representability and algebraicity are plausible and, where they rely on outside theorems (Piff–Welsh, GHM's algebraic matroid intersection), those are legitimate citations.\n\nSoft spots: the polymatroid section is less polished. Theorem 6.4 is stated without proof, and the contraction typo is a concrete consequence of that omission. The Section 7 bridge to linear operators is honest about the zero-image mismatch, but it is also close to a support-level restatement of the symbol identity; it does not add much beyond translation. The paper would benefit from a direct proof or a reference for Theorem 6.4, and from a note that the polymatroid contraction/deletion dichotomy mirrors the matroid case once γ_i=0 versus γ_i=α_i is used.\n\nThe central framework still holds up. This is a useful unifying perspective for matroid operations, not a revolutionary one, and the polymatroid part is repairable. I would send it to a competent referee and ask for the correction, then be happy to see it in print.\n\nFor the reading group: I would bring it up as an example of how a simple categorical idea can organize many known operations, but I would warn people about the typo in advance.","headline":"Fresh organizing framework for matroid/polymatroid operations; the matroid half is solid, but the polymatroid contraction theorem (6.4(4)) has a typo that swaps contraction for deletion.","tokens_in":15821,"tokens_out":3907,"would_cite":true,"duration_ms":36870,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B35","52B40"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper introduces matroid correspondences—a single functorial construction that produces deletion, contraction, free extension, truncation, intersection, union, and pullback—and shows they preserve representability and match Lorentzian o","keywords":["matroid correspondences","polymatroids","matroid quotients","Lorentzian polynomials","algebraic correspondences","representability","algebraicity","multisymmetric lift"],"falsifier":"Run the basis computation for the zero linear operator: the correspondence $C^*(P)$ will be nonempty while the operator's image is empty, exhibiting the rank discrepancy the paper itself flags. Separately, try to find a matroid quotient $M \\twoheadrightarrow N$ and a matroid $C$ such that $C^*(M)$ is not a quotient of $C^*(N)$; Proposition 2.3 rules this out, so any such example would refute the functoriality claim.","tokens_in":14864,"feed_emoji":"🔗","tokens_out":5664,"duration_ms":57078,"temperature":0.7,"texified_at":"2026-08-05T21:21:45.245533+00:00","pith_summary":"This paper introduces a matroid correspondence: a matroid $C$ on the disjoint union of two ground sets determines a functor $C^*$ that sends each matroid $M$ on the first ground set to the matroid $((M \\oplus B) \\cap C) \\setminus E_1$ on the second. The authors show that this single construction realizes deletion, contraction, free extension, truncation, intersection, union, pullback, and constant maps as special cases, and that it preserves representability and algebraicity when $C$ does. The same mechanism works for polymatroids, commuting with multisymmetric lifts. For a linear operator with Lorentzian symbol and nonzero image, the support of the operator's output equals the polymatroid correspondence of the input's support. The picture gives a uniform combinatorial analogue of algebraic correspondences, with the caveat that zero-image operators produce a correspondence of larger-than-expected rank.","texify_model":"deepseek-v4-flash","texify_usage":{"total_tokens":4509,"prompt_tokens":724,"completion_tokens":3785,"prompt_tokens_details":{"cached_tokens":0},"prompt_cache_hit_tokens":0,"prompt_cache_miss_tokens":724,"completion_tokens_details":{"reasoning_tokens":3119}},"feed_headline":"One matroid rule yields deletion, contraction, and truncation","feed_subtitle":"A correspondence functor mirrors algebraic correspondences and matches Lorentzian-operator supports when the image is nonzero.","key_machinery":"The matroid correspondence $C^*$ (Definition 1.1): direct-sum the input matroid with a boolean matroid, intersect with the correspondence matroid $C$, then delete the input ground set. The poset-category structure—where morphisms are matroid quotients, meaning every flat of the target is a flat of the source—and the composition formula (Proposition 2.6) carry the functoriality and make the construction a combinatorial analogue of the product of algebraic correspondences.","core_discovery":"The central claim is that matroid correspondences, defined as $C^*(M) = ((M \\oplus B_{E_2}) \\cap C) \\setminus E_1$ for a matroid $C$ on $E_1 \\sqcup E_2$, are functors between poset categories of matroids whose morphisms are matroid quotients. This one construction subsumes the standard matroid operations of deletion, contraction, free extension, truncation, intersection, union, pullback, and constant maps, and it preserves representability over infinite fields and algebraicity over any field. In the polymatroid setting, the correspondence commutes with multisymmetric lift and, when a linear operator with Lorentzian symbol has nonzero image, the support of the operator's output is exactly the polymatroid correspondence of the","pith_inferences":["The zero-image caveat suggests that a variant with an explicit rank parameter—perhaps a 'relative' correspondence—could remove the rank discrepancy and match algebraic correspondences even when the operator vanishes.","The composition formula for correspondences parallels the product of algebraic correspondences, hinting that intersection-theoretic identities (for example, projection formulas) may have matroid analogues worth making explicit.","Because the construction relies only on matroid intersection and deletion, it likely extends to valuated matroids or flag matroids, giving analogous functors in tropical or higher-rank settings.","The delta-matroid remark points to a natural next step: a delta-matroid correspondence defined via delta-matroid union and contraction could realize operations that reverse the quotient direction, such as duality."],"forward_implications":["Standard matroid operations—deletion, contraction, free extension, truncation, intersection, union, pullback, and constant maps—are all instances of one functorial construction, so results proved for correspondences apply to all of them at once.","If the correspondence matroid is algebraic or representable over a field, then every matroid in its image is algebraic or representable (representability requires an infinite field).","Polymatroid correspondences commute with multisymmetric lifts, making them compatible with polarization of volume polynomials.","For a linear operator with Lorentzian symbol and nonzero image, the support of the output equals the polymatroid correspondence of the support of the input, connecting algebraic operators to purely combinatorial transformations.","Not every rank-preserving, weak-map-preserving functor that fixes uniform matroids is a correspondence—the class of correspondences is strictly smaller, as shown by an explicit functor that cannot arise in this way."],"fun_headline_variants":["Matroid correspondences unify six standard operations","One matroid functor subsumes all standard ops","Matroid correspondences match Lorentzian operator supports","Matroid correspondence functor: deletion to union in one"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The link to linear operators holds only when the operator's image is nonzero; in the zero-image case the correspondence produces a polymatroid of strictly larger rank than the (empty) output, so the support-matching theorem does not cover that case.","fun_headline_variants_meta":{"raw":{"variants":["Matroid correspondences unify six standard operations","One matroid functor subsumes all standard ops","Matroid correspondences match Lorentzian operator supports","Matroid correspondence functor: deletion to union in one"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00142,"raw_usage":{"total_tokens":5515,"prompt_tokens":634,"completion_tokens":4881,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":378,"completion_tokens_details":{"reasoning_tokens":4818}},"tokens_in":378,"tokens_out":4881,"duration_ms":36440,"temperature":1.0,"reasoning_tokens":4818,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T03:42:33.091280+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the basis computation for the zero linear operator: the correspondence $C^*(P)$ will be nonempty while the operator's image is empty, exhibiting the rank discrepancy the paper itself flags. Separately, try to find a matroid quotient $M \\twoheadrightarrow N$ and a matroid $C$ such that $C^*(M)$ is not a quotient of $C^*(N)$; Proposition 2.3 rules this out, so any such example would refute the functoriality claim.","supporting_citations":[],"review_version":1}