{"id":"3783dd79-6bf4-411a-a3f8-36fabc5a517a","arxiv_id":"1908.02049","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A locally rigid semi-Hopf V-category is Hopf and Frobenius if and only if it has non-singular left and right integral families, equivalently if its integral spaces are the monoidal unit.","lead":"This paper generalizes the classical Larson-Sweedler theorem from Hopf algebras to Hopf categories, showing that a locally rigid semi-Hopf category is Hopf and Frobenius exactly when it has non-singular integrals. The result unifies several known generalizations for weak, multiplier, and Turaev Hopf algebras.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Main theorem depends on outsourced fundamental theorem of Hopf modules (Thm 2.24), whose proof is only sketched; if it has hidden hypotheses the central equivalence collapses.","rationale":"The reader's weakest assumption identifies the same load-bearing point: Theorem 2.24 is imported with only a proof sketch and is used directly to prove invertibility of the antipode and the Frobenius equivalences in Theorem 4.13. My reading of the main proof chain confirms that this is the most critical unverified input. Without Theorem 2.24, Lemma 2.25 fails, and Proposition 4.5, Proposition 4.9, and the key implications of Theorem 4.17 lose their support. The cancellation issue in Corollary 4.19 is secondary and downstream: it concerns the claim of subsuming the classical theorem, and it can likely be repaired by a finite-projectivity argument, whereas failure of Theorem 2.24 would invalidate the main equivalence itself. Therefore the appropriate verdict remains CONDITIONAL, unchanged from the reader's assessment: the theorem is plausible and well-structured, but acceptance should wait for a complete proof of the imported fundamental theorem.","tokens_in":48580,"tokens_out":6069,"duration_ms":61319,"concrete_test":"Independently reconstruct the proof of Theorem 2.24 from [BCV16, Theorem 10.2] and check whether the triangle identities for −⊗A ⊣ (−)^coA require any assumption not stated in the paper. Then test the equivalence on a concrete Hopf V-category: take V=Vect_k and A the Hopf category associated to a finite groupoid G (Example 2.14), with M=H*_1 from Example 2.22(3). Compute the coinvariant object (H*_1)^coA and the counit β: (H*_1)^coA ⊗ A → H*_1, and verify explicitly that β is an isomorphism. If it is not, Lemma 2.25 is unsupported and Theorem 4.17 needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central equivalence Theorem 4.17 rests on Theorem 2.24, the fundamental theorem of Hopf modules for Hopf V-categories, imported from [BCV16] with only a proof sketch. The theorem asserts that for a semi-Hopf V-category A, the adjunction −⊗A ⊣ (−)^coA is an equivalence if and only if A is Hopf. This is used essentially in Lemma 2.25 to prove invertibility of the antipode of any locally rigid Hopf V-category, which in turn is needed for Proposition 4.5, Proposition 4.9, Theorem 4.13, and the implications (i)⇒(ii) and (iv)⇒(i) of Theorem 4.17. The paper does not supply a complete proof, and it explicitly omits the parallel 'full' fundamental theorem for opmodules in Section 2.3 as 'not required'. If Theorem 2.24 requires hypotheses beyond 'V has equalizers' (for example coequalizers, finite object sets, or a coinvariants hypothesis), the proof of Lemma 2.25 collapses and with it the main theorem. This is a genuine load-bearing gap rather than a stylistic issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a many-object generalization of the classical Larson-Sweedler theorem. Working with Hopf V-categories, i.e. categories enriched over comonoids in a braided monoidal category V, the authors introduce integral families and integral spaces, prove several characterizations of Frobenius V-categories, and establish Theorem 4.17: for a locally rigid semi-Hopf V-category, being Hopf with a non-singular right integral family is equivalent to having both non-singular left and right integral families, to being Hopf and Frobenius, to being Hopf with trivial left integral spaces, and to the dual statements for the dual opcategory. The paper also derives a version over rings with all projective modules free (Corollary 4.19) and discusses applications to Hopf algebras in braided categories, groupoid algebras, Turaev's Hopf group-coalgebras, and weak (multiplier) Hopf algebras. The proof is structured and heavily diagrammatic, with an integral theory modeled on the one-object case but adapted to the many-object setting.","tokens_in":48795,"tokens_out":10514,"duration_ms":115870,"significance":"If the result is correct, it is a substantial generalisation: it unifies the classical Larson-Sweedler theorem, its braided-monoidal one-object version, and several related results for weak and multiplier Hopf algebras, and it gives a categorical explanation of the integral/Frobenius correspondence. The paper also provides genuinely useful characterizations of Frobenius V-categories in terms of Casimir families, dual modules, trace maps, and Frobenius functors, and it contains explicit formulas for the antipode and Frobenius isomorphisms constructed from integrals. The examples in Section 5, especially those involving Turaev Hopf group-coalgebras and groupoid algebras, add concrete value. The main limitation is that the central theorem inherits a substantial dependency on the fundamental theorem of Hopf modules imported from earlier work with only a proof sketch; the reader cannot fully verify the load-bearing input from the manuscript alone.","major_comments":[{"comment":"Theorem 2.24, the fundamental theorem of Hopf modules for Hopf V-categories, is stated with only a proof sketch: the unit and counit are named, but the verification that they form an adjoint equivalence, and the proof of the 'if and only if' claim, are not given. This theorem is used essentially in Lemma 2.25 (invertibility of the antipode), in Theorem 4.13, and in Corollary 4.19. Since the statement as written assumes only that V has equalizers, whereas the paper elsewhere needs limits or completeness, the precise hypotheses of the imported theorem should be stated and either a complete proof supplied or a precise pointer to the proof in [BCV16] must be given. This is a load-bearing dependence, not a stylistic matter.","section":"§2.2 (Theorem 2.24)"},{"comment":"In the proof of (ii)⇒(iii), the argument passes from a right H*_{y,x}-module isomorphism H*_{y,x} ≅ H_{y,x} to the conclusion that H*_{y,x} is a Frobenius monoid via the one-object case of Proposition 3.12. This step also needs the identification of the dual of H*_{y,x} with H_{y,x}, which follows from local rigidity but is not explicitly stated or proved. Please add the missing identification so that the application of Proposition 3.12 is transparent.","section":"§4.3 (Theorem 4.13, (ii)⇒(iii))"},{"comment":"The proof of (viii)⇒(iv) concludes from A*_{x,x} ≅ ∫ℓ A,x ⊗ A_{x,x} and the equality of dimensions that ∫ℓ A,x is free of rank one. This cancellation step is valid only after one observes that ∫ℓ A,x is a direct summand of the free module A*_{x,x} (or otherwise justifies projectivity), so that the ring hypothesis applies. The argument is easily repairable, but as written it is a compressed step in a key corollary and should be spelled out.","section":"§4.4 (Corollary 4.19, (viii)⇒(iv))"},{"comment":"Theorem 4.17 is stated for a locally rigid semi-Hopf V-category, but the notions of non-singular integral (Definition 4.4) and integral space (Definition 4.7) require V to be braided monoidal closed and to have limits. These standing hypotheses should be incorporated into the theorem statement, or at least explicitly recalled there, so that the statement is self-contained.","section":"§4 (Theorem 4.17 statement)"}],"minor_comments":[{"comment":"The proof of Theorem 4.16(iv) contains apparent notation errors: in the displayed verification, composites such as qxx◦qxx and pxx◦pxx are ill-typed; the intended maps appear to be the right inverses qx and px. Please correct these composites.","section":"§4.3 (Theorem 4.16(iv), displayed computation)"},{"comment":"The notation H* is used in Theorem 4.13(iv), Proposition 4.10, and elsewhere without consistently indicating whether it means the dual V-graph, the dual V-opcategory H*,op, or the associated semi-Hopf V-opcategory. Please fix the notation, for example by writing H*,op whenever an opcategory is meant.","section":"§4.2 and §4.3 (notation H* vs. H*,op)"},{"comment":"Proposition 4.9 asserts a natural isomorphism between the diagrams (41) and (42) but does not fully verify that the proposed morphisms commute with all legs of the limit diagrams. Adding the explicit compatibility check would make the isomorphism of integral spaces rigorous and easier to follow.","section":"§2.3 and §4.2 (Proposition 4.9)"},{"comment":"The phrase 'same dimension' in Lemma 4.18 and Corollary 4.19 should be defined precisely for modules over the base ring k; the paper's convention that finite dimensional means finitely generated projective of finite rank is recalled, but the dimension equalities are used in contexts where nonzero hom-objects are involved, and a short clarification would prevent ambiguity.","section":"§4.4 (Lemma 4.18 and Corollary 4.19)"},{"comment":"Several displayed commutative diagrams are visually dense and some labels (e.g. in the proof of Theorem 4.13 and Remark 4.14) are nearly illegible in the text version. Reproducing the diagrams with clearer placement of tensor factors and labels would improve verifiability.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is clearly valuable and the central claim is plausible, but its main theorem rests on the fundamental theorem of Hopf modules, which is imported from the authors' own earlier work with only a sketch. I would advise the editor to ensure that [BCV16] is publicly available and that the imported theorem exactly covers the hypotheses used here; otherwise the main result is conditional on an unverified input. The remaining issues (the cancellation in Corollary 4.19, notational inconsistencies, and ill-typed composites in Theorem 4.16) are local and repairable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a genuine many-object Larson-Sweedler theorem, and the proof structure looks right. I would send it to a serious referee. The biggest caveat is exactly the one the stress-test flags: the main chain uses the fundamental theorem of Hopf modules (Thm 2.24) as a black box from [BCV16], with only a sketch and a pointer to Böhm-Lack. That is a published result and the pointer is reassuring, but the equivalence in Theorem 4.17 cannot be checked without trusting it, so a referee should verify the hypotheses in [BCV16]. The paper would be stronger with a full proof or at least a precise statement of the hypotheses.\n\nWhat is new: the multi-object theorem itself, the treatment of integral spaces as limits rather than equalizers, and the Frobenius characterizations in Section 3 are real contributions. The one-object folklore case is honestly labelled. The applications to Turaev and weak multiplier Hopf algebras look illustrative rather than forced. The self-citations are appropriate; the paper builds directly on [BCV16] and [BFVV17].\n\nWhere the soft spots are: Corollary 4.19 has a cancellation step (A*xx ≅ ∫ℓ A,x ⊗ Axx implies ∫ℓ A,x has rank one). It is probably fine, but the authors should say that ∫ℓ A,x is finitely generated projective; otherwise the rank argument is implicit. Minor. Also, non-singularity is defined through split epimorphisms rather than isomorphisms; this is fine for fields but deserves a remark in the concrete cases. The proof of Proposition 4.9 is more of a diagram sketch than a full verification, but I do not see a substantive gap.\n\nThe stress-test note overstates the 'outsourced' problem: importing a published theorem from a previous paper is standard practice. But it is fair to say the main result inherits any hidden hypotheses in that theorem. The paper explicitly omits the converse version of the opmodule fundamental theorem, saying it is not needed; that is accurate as far as I can tell.\n\nWho this is for: Hopf algebra people and enriched category theorists. The paper unifies a lot of specialized results and gives a clean framework. It deserves referee time.","headline":"A substantial many-object generalization of Larson-Sweedler that mostly delivers, with a real but manageable dependency on prior work and a small gap in one corollary.","tokens_in":49323,"tokens_out":6099,"would_cite":true,"duration_ms":66833,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["18D20","16T05","18M05"],"pacs":[],"model":"deepseek-v4-flash","headline":"A many-object generalization of the Larson-Sweedler theorem is proved for Hopf V-categories.","keywords":["Larson-Sweedler theorem","Hopf V-categories","Frobenius V-categories","integral theory","Hopf monoids","many-object Hopf algebras","non-singular integrals","Turaev Hopf group algebras"],"falsifier":"Find a locally rigid semi-Hopf V-category over $\\mathbf{Vect}_k$ that has non-singular left and right integral families but whose antipode is not invertible, or whose left integral space is not isomorphic to $k$; either example would directly contradict Theorem 4.16(iv) and Theorem 4.17. Equivalently, exhibit a Hopf V-category whose regular Hopf module violates the coinvariant reconstruction of the fundamental theorem.","tokens_in":48399,"feed_emoji":"🧮","tokens_out":8038,"duration_ms":93731,"temperature":0.7,"pith_summary":"This paper proves a many-object version of the classical theorem that a finite-dimensional bialgebra is a Hopf algebra exactly when it admits a non-singular integral. The setting is a Hopf V-category, a category enriched over comonoids in a braided monoidal category V, with a Hopf algebra as the one-object case; the paper develops integral families and integral spaces for these categories and shows that when every hom-object is dualizable (local rigidity), 'Hopf plus a non-singular right integral' is equivalent to 'Hopf plus Frobenius' and to 'integral spaces isomorphic to the monoidal unit.' If correct, one uniform theorem covers classical Hopf algebras, Hopf monoids in braided categories, Turaev Hopf group algebras, and packed forms producing weak and multiplier Hopf algebras.","feed_headline":"Larson-Sweedler test now works for many-object Hopf algebras","feed_subtitle":"Non-singular integrals, antipodes, and Frobenius structures are shown equivalent for categories of comonoids.","key_machinery":"The load-bearing objects are Hopf V-categories: categories enriched over the monoidal category of comonoids in a braided monoidal category V, so each hom-object $A_{x,y}$ is a comonoid with local comultiplication and counit, and an antipode $s_{xy}: A_{x,y} \\to A_{y,x}$. Integral families are morphisms from the unit graph to the regular module, and integral spaces are defined as limits, not equalizers; non-singularity means the induced maps $p_{xx}$ and $q_{xx}$ on internal hom objects are split epimorphisms. The proof chain uses the fundamental theorem of Hopf modules (imported from earlier work), a dual fundamental theorem for Hopf opmodules proved here, and new characterizations of Frobenius V-categories by Casimir families, module isomorphisms $A \\cong A^{*,\\mathrm{op}}$, non-degenerate trace pairings, and Frobenius adjunctions.","core_discovery":"The central result, Theorem 4.17, states that for a locally rigid semi-Hopf V-category A the following are equivalent: (i) A is Hopf and has a non-singular right integral family; (ii) A has both a non-singular right and a non-singular left integral family; (iii) A is Hopf and Frobenius; (iv) A is Hopf and the left integral spaces $\\int^\\ell A_x$ are isomorphic to the monoidal unit $I$ for all $x$; plus the dual statements for the dual semi-Hopf V-opcategory and the left-right interchanged versions. Over k-modules where projective modules are free, the list is further equivalent to A being simply Hopf (Corollary 4.19), which is exactly the classical Larson-Sweedler statement in the one-object case.","pith_inferences":["The theorem suggests that 'many-object quantum groups' could be defined by integral existence rather than by antipodes, in the same way the classical theorem motivated locally compact quantum groups; this is a natural programmatic next step, not stated in the paper.","Because non-singularity is tested only on diagonal hom-objects $p_{xx}$ and $q_{xx}$ yet Theorem 4.16 shows all $p_{xy}$ and $q_{xy}$ become isomorphisms, there is a local-to-global principle worth testing: checking endo-hom components alone may certify integral non-singularity everywhere.","Replacing the object-indexing set $X \\times X$ by an arbitrary groupoid could unify Hopf G-algebras with Hopf categories and yield one common Larson-Sweedler theorem; the paper names this as a direction for future work.","The identification of locally rigid Calabi-Yau categories with symmetric Frobenius categories (Corollary 3.22) suggests integral data could serve as a categorical trace in TQFT constructions, beyond anything the paper proves."],"forward_implications":["In the one-object case, the theorem yields a Larson-Sweedler statement for Hopf monoids in any braided monoidal category, including monoidal Hom-Hopf algebras, graded Hopf algebras, and Yetter-Drinfel'd Hopf algebras.","For a k-linear Hopf category with finitely many objects and finite-dimensional hom-objects, the packed form is a Frobenius weak Hopf algebra; with infinitely many objects it gives a Frobenius weak multiplier Hopf algebra.","Finite-dimensional Turaev Hopf G-algebras become Frobenius categories, and the paper spells out explicit cocomposition and trace maps for the associated Hopf category and for groupoid algebras.","Over rings where every finitely generated projective module is free, the conditions collapse: A is Hopf if and only if it has a right non-singular left integral family and the non-zero hom-objects have equal dimensions (Corollary 4.19).","Every Frobenius Hopf V-category is locally Frobenius: each hom-object $H_{x,y}$ is itself a Frobenius algebra in V, so the local comonoid and local monoid structures coexist on the same objects."],"supporting_citations":[{"why":"Supplies the definition of Hopf V-categories and the fundamental theorem of Hopf modules (Theorem 2.24) on which the main chain of implications rests.","marker":"[BCV16]"},{"why":"Introduces Frobenius V-categories and provides the starting characterizations that Section 3 reproves and extends.","marker":"[BFVV17]"},{"why":"States the classical Larson-Sweedler theorem that this paper generalizes to the many-object setting.","marker":"[LS69]"},{"why":"Refines the classical result by showing non-singular integrals make a Hopf algebra Frobenius, the one-object ancestor of Theorem 4.17.","marker":"[Par71]"},{"why":"Supplies the method, via Hopf modules, for proving invertibility of the antipode in the locally rigid setting, used in Lemma 2.25.","marker":"[Tak99]"},{"why":"Gives the enriched category theory background, including composition and change of base, underlying the V-category formalism.","marker":"[Kel05]"}],"fun_headline_variants":["Larson-Sweedler theorem now covers many-object Hopf categories","Non-singular integrals characterize Hopf V-categories","Hopf V-categories: integrals force Frobenius structure","Many-object Hopf algebras: integral criterion proven"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is an imported theorem, used with only a proof sketch: for every Hopf module over a Hopf V-category, the coinvariants tensored back with the category reconstruct the original module ($M^{coA} \\otimes A \\cong M$). If that reconstruction fails, the proofs of antipode invertibility and of the Frobenius equivalence collapse.","fun_headline_variants_meta":{"raw":{"variants":["Larson-Sweedler theorem now covers many-object Hopf categories","Non-singular integrals characterize Hopf V-categories","Hopf V-categories: integrals force Frobenius structure","Many-object Hopf algebras: integral criterion proven"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00053,"raw_usage":{"total_tokens":2496,"prompt_tokens":830,"completion_tokens":1666,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":446,"completion_tokens_details":{"reasoning_tokens":1598}},"tokens_in":446,"tokens_out":1666,"duration_ms":12702,"temperature":1.0,"reasoning_tokens":1598,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:54:52.082083+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a locally rigid semi-Hopf V-category over $\\mathbf{Vect}_k$ that has non-singular left and right integral families but whose antipode is not invertible, or whose left integral space is not isomorphic to $k$; either example would directly contradict Theorem 4.16(iv) and Theorem 4.17. Equivalently, exhibit a Hopf V-category whose regular Hopf module violates the coinvariant reconstruction of the fundamental theorem.","supporting_citations":[],"review_version":1}