{"id":"381d46fe-8ec3-4e63-a1ec-38fbcfa6fb5d","arxiv_id":"1908.03420","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Stringent skew hyperfields are perfect: over them every vector of a matroid is orthogonal to every covector, and weak matroids coincide with strong matroids.","lead":"This paper proves that every matroid defined over a stringent skew hyperfield has vectors orthogonal to its covectors, so weak and strong matroid notions coincide for such hyperfields. It also gives vector-style axioms for these matroids, generalizing the rules for oriented and valuated matroids.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 30, the internal bridge from residue orthogonality to H-orthogonality, is stated without proof; Theorem 33 and Corollary 34 depend on it.","rationale":"The paper's central claim is that stringent skew hyperfields are perfect. The proof route is: classify H as R ⋊ Γ (Theorem 27), build the residue matroid M0 (Lemma 31), then transfer vector/covector orthogonality between M and M0 (Lemma 32 and Theorem 33). The transfer in both directions uses Lemma 30. The reader's weakest assumption was the external Bowler–Su classification; I agree that classification is load-bearing, but the more immediately checkable weakness is the unproved Lemma 30. A false Lemma 30 would not merely shrink the scope of the theorem; it would break the reduction for all three residue types. I can find no internal contradiction elsewhere, and the vector-axiom work in Section 4.3 is extensive and plausible. The right verdict remains conditional: accept if Lemma 30 is supplied and the classification reference is sound; otherwise the proof is incomplete. Since the reader already returned CONDITIONAL, my read does not move the verdict.","tokens_in":26261,"tokens_out":13899,"duration_ms":155957,"concrete_test":"Work in H = Q((t))/G with residue R = Q, where G = 1 + tQ[[t]]. Take X = (1,1) and Y = (1, -1+t). Verify the lemma's converse explicitly: top supports meet, and the residue sum 1·1 + 1·(-1) = 0 gives X^Ò ⊥ Y^Ò; then exhibit g_1,g_2 ∈ G such that g_1 + (-1+t)g_2 = 0 (e.g. g_1 = 1-t, g_2 = 1), showing X ⊥ Y. Generalize this lifting argument to k ≥ 2 top positions and to the sign-hyperfield residue; if a case with 0 in the residue sum but no admissible choice of g_e exists, Lemma 30 is false and Theorem 33 collapses.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 30 is the bridge that converts residue orthogonality into H-orthogonality, and it is stated without proof. The converse direction — if X^Ò and Y^Ò are orthogonal and their top supports intersect, then X ⊥ Y — is not a formal consequence of the Γ_max case in Lemma 13. When the residue R is a skew field, orthogonality of the top parts means the residue sum is exactly zero; to obtain 0 in the H-hypersum one must choose unit-group factors g_e with ψ(g_e) = 1 so that the lifted sum vanishes, and then absorb all lower-valuation terms. This lifting is exactly the content of the stable-sums condition in the Bowler–Su description, but the paper never supplies the argument. The proof of Theorem 33 invokes Lemma 30 in the final step to conclude V ⊥ U from V^Ò ⊥ U^Ò, so a gap here is inherited by Corollary 34 and Theorem 1. The Bowler–Su classification is a second external dependency, but Lemma 30 is an internal missing proof that can be checked without leaving the paper.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates matroids over stringent skew hyperfields, a class of hyperfields that includes the Krasner hyperfield, the sign hyperfield, skew fields, and the hyperfields arising from valued fields via Krasner's construction. The two central results are Theorem 1, which states that for any stringent skew hyperfield H and any left H-matroid M, every vector of M is orthogonal to every covector of M (so H is perfect), and Theorem 44, which gives vector axioms characterizing exactly the sets of vectors of left H-matroids over such H. The proofs proceed by using the Bowler--Su classification of stringent skew hyperfields to reduce to three cases according to the residue hyperfield (Krasner, sign, or skew field), constructing a residue matroid M0 for each H-matroid M, and then importing perfection and vector-axiom results from the known cases. The paper also proves that vectors behave well under taking minors for stringent hyperfields, generalizing results of Anderson and of Murota--Tamura.","tokens_in":26376,"tokens_out":4464,"duration_ms":51409,"significance":"If the main theorems hold, the paper unifies and extends a substantial body of work: it generalizes Dress--Wenzel's perfection results for fuzzy rings and the Baker--Bowler treatment of hyperfields, it provides a common framework for oriented matroids, valuated matroids, and matroids over skew fields, and it supplies vector axioms that specialize to the Murota--Tamura axioms for valuated matroids and to the standard oriented-matroid vector axioms. The applications sketched in Section 5, particularly to real tropical singularities and to algebraic matroids via derivations, indicate that the framework has real explanatory power. The manuscript is generally careful and detailed, with full proofs for most internally developed results, and the construction of the residue matroid is nontrivial and interesting. However, the present version has a load-bearing gap: Lemma 30, which bridges residue orthogonality to orthogonality over H, is stated without proof, and the main theorem depends on it. For this reason I cannot recommend acceptance in the current form.","major_comments":[{"comment":"Lemma 30 is stated without proof, but its converse direction is load-bearing for the paper's central claim. The lemma claims that for X,Y ∈ H^E with X^Ò ∩ Y^Ò ≠ ∅, residue orthogonality X^Ò ⟂ Y^Ò implies X ⟂ Y over H. This is not a formal consequence of the analogous Lemma 13 for Γ_max, because when the residue R is a skew field, orthogonality of the top parts only says that the residue sum is zero; to obtain 0 in the H-hypersum one must lift the residue terms with unit-group factors satisfying ψ(g_e)=1 and then absorb all lower-valuation contributions. That lifting argument is exactly where the stable-sums condition in the Bowler--Su description enters, but the paper never supplies it. The gap is inherited by Theorem 33 (the final step concludes V ⟂ U from V^Ò ⟂ U^Ò), by Corollary 34, and by Theorem 1. Please provide a complete proof of Lemma 30, or move it to an appendix with a full argument; the current one-line statement is not sufficient for such a central step.","section":"§4.2, Lemma 30"},{"comment":"The main theorems depend essentially on the classification Theorem 27, imported from [BS20], which asserts that every stringent skew hyperfield has the form R ⋊_{U,ψ} Γ with R equal to the Krasner hyperfield, the sign hyperfield, or a skew field. This is an external result, and at the time of writing [BS20] is a preprint by one of the present authors and Ting Su. Since every reduction in Section 4, including the construction of the residue matroid and the three-case perfection argument, relies on this classification, the paper's main theorem is conditional on it. This is not a circularity, but it is a completeness concern: the reader cannot verify Theorem 1 from the present manuscript alone. The authors should either include a proof of the classification, state explicitly that the main theorem is a consequence of [BS20], or cite a published version if one now exists.","section":"§4.1, Theorem 27"},{"comment":"The proof of Lemma 32 contains the sentence 'it follows that V M Y by Lemma 32', which cites the very lemma being proved; the intended reference is presumably Lemma 30. This is not merely a typo: it underscores that the missing proof of Lemma 30 is being used as a black box. Similarly, in the proof of Theorem 33, after the rescaling step the text says 'U^Ò ∈ UpMρ q', which should almost certainly be 'U^Ò ∈ UpM0q'; the notation Mρ is not defined at that point and the covector rescalings have not been tracked carefully. These points should be corrected along with the proof of Lemma 30.","section":"§4.2, Lemma 32 and Theorem 33"}],"minor_comments":[{"comment":"In the proof of Theorem 15, the phrase 'we apply (D2) for C and D' appears to be a typo for '(M2)'.","section":"§2.4, Theorem 15 proof"},{"comment":"The vector axioms in Theorem 44 refer to 'VpMq' inside the statement of the axioms before the matroid M has been constructed; these occurrences should be 'V' throughout, as is already done in the introductory Theorem 2.","section":"§4.3, Theorem 44 statement"},{"comment":"Lemma 13 is stated for Γ_max with a one-line proof relying on the order structure; it would help the reader to spell out the notation X^Ò · Y^Ò and the case distinction in the converse, since this lemma is the model for the unproved Lemma 30.","section":"§3.1, Lemma 13"},{"comment":"The name 'Jürgens' is garbled in the text and in the reference list; it should be spelled consistently, and the citation [J18] should give the full author name.","section":"§5.2"}],"recommendation":"major_revision","confidential_remarks":"The main concern is the unproved Lemma 30, which is load-bearing for Theorem 33 and Corollary 34. I believe the gap is likely fixable: the lifting argument sketched in the skeptic's note is plausible and the paper already contains the necessary stable-sums machinery. The second concern is the dependence on the classification [BS20], which is a preprint by one of the authors; the editor may wish to verify the current publication status of that paper. The vector-axiom part and the minor-minor behaviour results are largely self-contained and well argued. If the authors supply a complete proof of Lemma 30 and clarify the status of Theorem 27, the paper would be a strong contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the quick take: this is a real advance, and the main theorem is probably true, but there's one internal gap and one heavy external dependency that a referee needs to push on.\n\nThe paper proves that over any stringent skew hyperfield, every vector of a weak matroid is orthogonal to every covector, hence weak matroids are strong. That unifies earlier perfection results for doubly distributive hyperfields, oriented matroids, valuated matroids, and skew fields, and it extends them to noncommutative hyperfields. The vector axioms in Theorem 44 are a genuine generalization of the Murota–Tamura and oriented matroid axioms, and Theorem 40 shows that vector sets behave well under minors, which was not known for this class. The residue matroid construction, especially the noncommutative case in Lemma 31, is substantial and looks plausible. The paper is honest about the one place it cannot get a clean single-circuit elimination axiom when the residue is a skew field.\n\nNow the soft spots, in proportion. Lemma 30 is the bridge that turns residue orthogonality into H-orthogonality, and it is stated without proof. The forward direction is easy; the converse, when the top supports intersect, needs an argument that uses the stable-sums condition from the Bowler–Su classification. This is not a trivial consequence of Lemma 13, and since Theorem 33 and Corollary 34 depend on it, a referee should insist on a proof. It may be fillable without changing the architecture, but it is a real missing piece.\n\nThe second dependency is the classification of stringent skew hyperfields [BS20], also by one of the present authors. Theorem 27 is imported wholesale. That is not a flaw per se—the classification is a separate paper—but the present paper's three-case proof collapses without it, so the referee should ask for a precise statement and, ideally, enough of the argument to see why the three residues are exhaustive.\n\nOverall, I do not see a load-bearing error. The reductions are careful, the vector axioms are checked in both directions, and the limitations are acknowledged. The main theorem is significant and the paper deserves serious peer review. I'd ask for a proof of Lemma 30 and a clearer statement of the external classification before acceptance. It is a good paper for anyone working on matroid theory, tropical geometry, or hyperfields; I would bring it to a reading group and cite it.","headline":"A genuine unification of perfection results for stringent skew hyperfields, but with an unproved bridging lemma and a heavy external classification that a referee should push on.","tokens_in":26960,"tokens_out":3899,"would_cite":true,"duration_ms":40492,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Over every stringent skew hyperfield, weak matroids are strong: all vectors and covectors are orthogonal.","keywords":["stringent hyperfields","matroids over hyperfields","valuated matroids","oriented matroids","vector axioms","residue matroids","orthogonality of circuits and cocircuits","skew hyperfields"],"falsifier":"Exhibit a left H-matroid over a stringent skew hyperfield with a vector V and a covector U such that $0\\notin\\sum_{e\\in E} V_e U_e$; Theorem 1 asserts no such pair exists. A finite search over small stringent hyperfields and small ground sets, checking orthogonality of all circuits against all cocircuits, would either find such a pair or confirm the claim in the tested range.","tokens_in":25974,"feed_emoji":"🧮","tokens_out":12682,"duration_ms":122978,"temperature":0.7,"pith_summary":"The paper establishes that stringent skew hyperfields—hyperfields in which addition is multivalued only when the two summands are negatives—form a class over which the weak and strong definitions of matroid coincide. The key theorem says that for any left H-matroid over such an H, every vector is orthogonal to every covector; from this the authors derive that H is perfect, hence weak matroids over H are strong matroids. The paper also gives a vector-axiom characterization: a subset of $H^E$ is the set of vectors of some left H-matroid exactly when it satisfies closure under scaling, a residue-dependent composition, and an elimination rule. This matters because valued fields, via the standard quotient construction, produce stringent hyperfields, so the framework simultaneously covers classical, oriented, valuated, and certain algebraic matroids while preserving the good minor and vector behavior that fails for general hyperfields.","feed_headline":"Stringent hyperfields make weak matroids strong","feed_subtitle":"All vectors and covectors meet orthogonally, unifying oriented, valuated, and classical matroids.","key_machinery":"The load-bearing machinery is the structural classification of stringent skew hyperfields as semidirect products $H=R\\rtimes_{U,\\psi}\\Gamma$: a residue hyperfield R equal to the Krasner hyperfield, the sign hyperfield, or a skew field, a group U mapped by $\\psi$ onto an ordered value group $\\Gamma$, with stable sums. From a matroid M over H, the paper builds the residue matroid $M_0$ over R whose circuits and cocircuits are the minimal supports $X^{\\uparrow}$ of the original circuits and cocircuits, where $X^{\\uparrow}$ records the entries of maximal size. The central transfer lemma says orthogonality passes down to these supports and, when two supports meet, back up again; this lets the authors lift the known perfection of the three residue types to all of H. Around this core sit a composition operation $\\circ$ on H, used in the vector axiom (V2)$^1$, and minor theorems showing that vectors of contractions and deletions are exactly restrictions of vectors, which carry the inductions.","core_discovery":"Over a stringent skew hyperfield H, the paper proves Theorem 1: if M is a left H-matroid on a finite ground set E, V is a vector of M, and U is a covector of M, then V is orthogonal to U, meaning $0 \\in \\sum_{e\\in E} V_e U_e$. Since the paper defines a hyperfield to be perfect exactly when this orthogonality holds for all matroids over it, this says every stringent skew hyperfield is perfect, and then Theorem 7 gives the equivalence of weak and strong matroids: every weak H-matroid has strong duality. The second main result, Theorem 44, characterizes vector sets: a set $\\mathcal{V}\\subseteq H^E$ is the vector set of a left H-matroid if and only if it satisfies (V0) $0\\in\\mathcal{V}$, (V1) closure under scalar multiplication, (V2)$^1$ closure under the composition operation whenever $V\\circ W=V\\cup W$, and (V3) the elimination rule that from $V_e=-W_e\\neq 0$ one can find $Z\\in\\mathcal{V}$ with $Z_e=0$ and $Z\\in V\\oplus W$; the circuits are then exactly the minimal nonzero vectors.","pith_inferences":["The three-case reduction suggests a testable boundary: if 'stringent' is relaxed while keeping the residue classification and support-transfer lemma intact, perfection may still hold; the paper's own examples show the minor and vector behavior fails for general hyperfields, so the boundary is likely near stringent.","The composition operation is associative only when the residue is the Krasner hyperfield and commutative only when it is not the sign hyperfield, so the vector axioms require only a weak, possibly non-associative composition; this hints that other axiom systems for matroids over tracts could be weakened in the same way.","For algebraic matroids in positive characteristic, the paper's field-extension construction packages Frobenius rescalings as rescalings of one matroid; a concrete next step would be to test whether the derivation space discussed in the field-extension section can be read off directly from the residue matroid, giving a local, characteristic-free criterion."],"forward_implications":["For every stringent skew hyperfield, weak matroids and strong matroids coincide: every weak H-matroid has globally orthogonal circuits and cocircuits, so the dual is well-behaved.","Vector sets of H-matroids are cryptomorphically described by the four vector axioms, so membership in a vector set can be checked by closure under scaling, composition, and elimination rather than by quantifying over bases.","Minors of H-matroids behave like minors of oriented and valuated matroids: contracting or deleting an element restricts the vectors exactly as expected, a property that fails for general hyperfields.","The residue matroid construction unifies the classical residue matroids of valuated matroids, oriented initial matroids, and residue linear spaces over skew fields within a single construction.","Because valued skew fields produce stringent hyperfields through the quotient construction, a linear space over a valued field carries its valuation and residue structure in one perfect matroid, so the two layers of data cannot contradict each other."],"supporting_citations":[{"why":"Supplies the classification of stringent skew hyperfields as semidirect products with residue the Krasner hyperfield, the sign hyperfield, or a skew field; this is the three-case split that carries the proof.","marker":"[BS20]"},{"why":"Defines perfect fuzzy rings and proves that perfection makes weak matroids strong; the paper adapts this notion of perfection.","marker":"[DW92a]"},{"why":"Introduces valuated matroids and their residue matroids, the construction whose non-commutative generalization appears here.","marker":"[DW92b]"},{"why":"Minty's circuit/cocircuit characterization is used to show that the minimal supports of circuits and cocircuits form a residue matroid.","marker":"[Min66]"},{"why":"Provides the vector axioms for valuated matroids that the paper's Theorem 44 generalizes to stringent skew hyperfields.","marker":"[MT01]"},{"why":"Gives vector axioms for matroids over tracts and shows the general failure of the minor-behavior theorems that stringent hyperfields restore.","marker":"[And19]"},{"why":"Constructs hyperfield matroids for algebraic matroids and residue matroids over skew hyperfields; the paper's residue-matroid lemma generalizes its Lemma 14.","marker":"[Pen18]"}],"fun_headline_variants":["Stringent hyperfields turn weak matroids into strong ones","Weak matroids over stringent hyperfields are always strong","Orthogonality in stringent hyperfields upgrades weak matroids","Stringent hyperfields prove weak equals strong for matroids"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes the imported classification theorem that every stringent skew hyperfield is a semidirect product whose residue is the Krasner hyperfield, the sign hyperfield, or a skew field; if that classification misses a case, the three-case reduction does not cover all stringent skew hyperfields.","fun_headline_variants_meta":{"raw":{"variants":["Stringent hyperfields turn weak matroids into strong ones","Weak matroids over stringent hyperfields are always strong","Orthogonality in stringent hyperfields upgrades weak matroids","Stringent hyperfields prove weak equals strong for matroids"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000727,"raw_usage":{"total_tokens":3253,"prompt_tokens":934,"completion_tokens":2319,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":550,"completion_tokens_details":{"reasoning_tokens":2253}},"tokens_in":550,"tokens_out":2319,"duration_ms":16484,"temperature":1.0,"reasoning_tokens":2253,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:14:41.587382+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a left H-matroid over a stringent skew hyperfield with a vector V and a covector U such that $0\\notin\\sum_{e\\in E} V_e U_e$; Theorem 1 asserts no such pair exists. A finite search over small stringent hyperfields and small ground sets, checking orthogonality of all circuits against all cocircuits, would either find such a pair or confirm the claim in the tested range.","supporting_citations":[],"review_version":1}