{"id":"7cd0a972-f7c1-4cd7-9588-3ffbb9195395","arxiv_id":"2507.07807","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors prove that Gr(Ch×Dv) is naturally equivalent to C⊗D, settling Gaitsgory and Rozenblyum's final conjecture, and establish a companion-based universal property of the squares functor.","lead":"This paper proves the last open conjecture from Gaitsgory and Rozenblyum's foundational work on (∞,2)-categories, relating the Gray tensor product to the squares construction. It also shows that the squares functor is the universal way to add 'companions' to a two-dimensional category, a question posed in a recent thesis.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem B rests on Lemma 4.4's epimorphism claims, imported from the authors' unpublished [Lou25]/[Lou24]; this is the true black-box dependency, not the Čech-nerve effectivity the reader flags.","rationale":"I read the paper in good faith. The overall strategy is coherent: define Gr and Sq via a realization-nerve adjunction, then prove Gr(Ch×Dv)≃C⊗D by density and pushout computations. If every cited lemma is true, Theorem B follows. I found no internal contradiction or obvious diagram error in the main chain of Section 4. The reader's conditional verdict is therefore appropriate. However, the reader's stated weakest assumption identifies [Lou25, Theorem 3.4.1] and [Rui25a, Theorems 4.13/4.15] as the crucial black boxes. Those theorems support Theorem A (the universal property of squares), not Theorem B: Section 4 never invokes Theorem 3.10 or Theorem 3.14. The proof of Theorem 4.1 relies instead on a sequence of lemmas whose linchpin is Lemma 4.4, an unproved epimorphism assertion about the canonical comparison between Gray and cartesian products of globular sums. Lemma 4.4 is used in Construction 4.5, Lemma 4.6, and Lemma 4.8, which are indispensable for the final colimit computation. Its proof is delegated to [Lou25, Lemma 1.6.12] and [Lou24, Proposition 2.2.1.50], both from the first author's unpublished preprints. This is a genuine load-bearing dependency. It does not force a stronger verdict than CONDITIONAL, because the dependency is isolated and testable: one can compute the minimal case explicitly. If the test passes, the concern is resolved; if it fails, the main theorem is unsupported.","tokens_in":18363,"tokens_out":30499,"duration_ms":294718,"concrete_test":"Verify Lemma 4.4 for the minimal nontrivial case a=[1;1], b=[1] by an explicit generator computation in 2Gaunt without citing [Lou24, Prop 2.2.1.50]: show that [1;1]⊗[1]→[1;1]×[1] and [1;1]×[1]→[1] are epimorphisms. If this minimal computation fails, Construction 4.5's uniqueness arguments and Lemma 4.8's outer-square argument break; if it succeeds, repeat the check for all globular sums to confirm the imported lemma is correctly applied.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's weakest-assumption list is partly off-target for the central claim. Theorem 3.10 (and [Lou25, Thm 3.4.1]) is used to prove Theorem A, and Theorem 3.14/[Rui25a, Thms 4.13/4.15] is not invoked in Section 4. The proof of Theorem 4.1 instead rests on Lemma 4.4: for globular sums a,b, the canonical maps a⊗b→a×b and a×b→b are epimorphisms in (∞,2)Cat. This lemma is used in Construction 4.5 for the uniqueness of the dashed factorization, in Lemma 4.6 via Lemma 4.3, and in Lemma 4.8 to pass from the outer pushout to the right pushout; Lemma 4.8 then feeds Lemma 4.9 and the final span calculation. Lemma 4.4 is not proved here; it cites [Lou25, Lemma 1.6.12] and [Lou24, Prop 2.2.1.50]. If either of those fails, or if their application to globular sums is invalid, the pushout chain supporting Gr(Ch×Dv)≃C⊗D collapses. This is the true load-bearing black box for the paper's central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops the theory of the squares functor Sq on (∞,2)-categories and its left adjoint Gr, in order to resolve the last open Gaitsgory–Rozenblyum conjecture. The two main advertised results are: Theorem A, a universal property of Sq(C) as the free double ∞-category obtained by adding companions to the vertical inclusion Cv, and Theorem B, the natural equivalence Map(Ch×Dv, Sq(E)) ≃ Map(C⊗D, E), equivalently Gr(Ch×Dv) ≃ C⊗D. The proof of Theorem B proceeds by density arguments reducing to globular sums, a series of pushout computations, and the uniqueness of endomorphisms of the Gray tensor product. The paper also contains a status table for all eight conjectures from [GR17].","tokens_in":18677,"tokens_out":8027,"duration_ms":84712,"significance":"If correct, the paper settles a long-standing conjecture in the (∞,2)-categorical foundations of derived algebraic geometry, and it provides a clean statement of the universal property of the squares construction. The strategy is well organized, and the reduction to generators of the form [n;m] together with the uniqueness argument via Proposition 2.18 are valuable contributions. However, the proof is not self-contained: several load-bearing lemmas are imported from unpublished preprints by the same authors, notably the epimorphism claims of Lemma 4.4. The paper therefore presents a convincing architecture, but the verification burden is not fully met within the manuscript as it stands.","major_comments":[{"comment":"The epimorphism claims for the canonical maps a⊗b→a×b and a×b→b are the central pivot of the proof of Theorem 4.1: they are used to obtain uniqueness of the dashed factorization in Construction 4.5, to apply Lemma 4.3 in Lemma 4.6, and to pass from the outer pushout to the right pushout in Lemma 4.8. Yet Lemma 4.4 is proved only by citing [Lou25, Lemma 1.6.12] and [Lou24, Proposition 2.2.1.50], both unpublished preprints by the first author. Please state the precise results being cited and show that they apply to the combinations of globular sums used here, or give a direct proof of these epimorphism assertions. Without this, the chain supporting Gr(Ch×Dv)≃C⊗D is not established in the paper.","section":"Section 4, Lemma 4.4"},{"comment":"The proof of Lemma 4.8 ends with the assertion that the square τ0[m]×[k] → τ0[m]; [m]×[k] → [m] is a pushout in ∞Cat, justified only by the phrase 'as can be directly verified by reducing to m=k=1'. A reduction from arbitrary m,k to m=k=1 requires an argument showing that the square is preserved under the operations that build general m,k, and that argument is not supplied. Since Lemma 4.8 is used in Lemma 4.9 and in the final colimit computation of Theorem 4.1, please provide the missing verification or a proof that the reduction is legitimate.","section":"Section 4, Lemma 4.8"},{"comment":"Theorem 3.10, which is the foundation for Theorem A, is proved by a direct appeal to [Lou25, Theorem 3.4.1], and the intermediate Lemma 3.13 is dismissed with 'one may readily verify' while Lemma 3.15 invokes [Rui25a, Theorem 4.13]. Since Theorem A is one of the two advertised theorems of the paper, these inputs should either be proved in the manuscript or stated explicitly as assumptions with precise statements of the cited results. As written, the reader cannot independently check the main theorem of Section 3.","section":"Section 3, Theorem 3.10"},{"comment":"The paper's claim that the Gray tensor product used here coincides with the one in [GR17] proceeds through Proposition 2.22, whose proof relies on [Lou25, Theorem 1.4.14, Lemma 1.6.12, Proposition 1.4.22] and [Lou24, Lemma 2.1.1.5], and on Lemma 2.20, which is only verified by citing [Lou24, Lemma 2.1.1.5] and a 'readily verified' computation in PShSet(Θ2). This identification is load-bearing for the advertised resolution of the Gaitsgory–Rozenblyum conjecture, because the conjecture is about the Gray tensor product of [GR17]. Please make these dependencies explicit and supply the missing verification of Lemma 2.20 or a precise reference for it.","section":"Section 2, Proposition 2.22 and Lemma 2.20"}],"minor_comments":[{"comment":"The phrase 'all conjectures are now resolved' overstates what is proved in this paper, since several of the eight conjectures are resolved only by cited works; consider saying that with the present paper, proofs of all eight conjectures now exist in the literature.","section":"Abstract and §1.4"},{"comment":"There is a typo in 'cateogries' near the discussion of Grandis–Paré's theorem.","section":"§1.4"},{"comment":"The description of the image in Theorem A leaves implicit that the companion of a vertical arrow is unique up to contractible choice; Remark 3.7 supplies this, but stating it in the theorem would make the statement easier to read.","section":"Theorem A"},{"comment":"The notation MapDbl∞Cat([0,0],Q), MapDbl∞Cat([1]v,Q)′, and MapDbl∞Cat(Sq([1]),Q)′ is introduced only inside the proof diagram; the two middle columns are not named, which makes the diagram harder to follow.","section":"Lemma 3.15"},{"comment":"The proof uses a density argument claiming full faithfulness of certain restriction functors on cocontinuous bifunctors; the argument is compressed and would benefit from a reference to a standard nerve theorem or a short explanation of why the restriction is fully faithful.","section":"Proposition 2.18"}],"recommendation":"major_revision","confidential_remarks":"The main risk is not circularity: the paper does not assume the conjecture it proves. The risk is that the central proof of Theorem B rests on Lemma 4.4, whose proof is entirely delegated to the authors' unpublished preprints [Lou25] and [Lou24]; the stress-test note correctly identifies this as more important than the other black boxes such as [Lou25, Theorem 3.4.1]. If the editors accept reliance on these unpublished works, the paper may become publishable after the authors spell out the exact cited statements and prove or explicitly import Lemma 4.4 and Lemma 2.20. Otherwise, the manuscript should be required to include proofs of these load-bearing inputs."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe short version: this paper plausibly settles the last open Gaitsgory–Rozenblyum conjecture, and it earns a serious referee. It proves Theorem B, the equivalence Gr(Ch×Dv) ≃ C⊗D, and also proves Theorem A, the universal property of the squares functor that Ruit conjectured in his thesis. The table of the eight conjectures and their statuses is genuinely useful.\n\nWhat I like. The proof strategy is natural: reduce by density to globular sums, then pushout arguments. Proposition 2.18 (no non-trivial endomorphisms of the Gray tensor product) is elegant and useful, because it lets the authors upgrade existence to uniqueness of the comparison. Proposition 2.22, comparing their Gray tensor product with Loubaton's, is also a necessary piece. The paper is careful about which model of Gray tensor product is being used.\n\nThe soft spots are real, but they are soft rather than fatal. The proof is not self-contained. Theorem 3.10 uses [Lou25, Thm 3.4.1] as a black box. Theorem 3.14 uses [Rui25a, Thm 4.13]. The stress-test note is right that the most load-bearing dependency for Theorem 4.1 is Lemma 4.4, the epimorphism claims a⊗b→a×b and a×b→b, which cites [Lou25, Lemma 1.6.12] and [Lou24, Prop 2.2.1.50]. Lemma 4.4 drives the pushout chain in Construction 4.5, Lemma 4.6, Lemma 4.8, Lemma 4.9 and the final span computation. If that lemma fails, Theorem B collapses. The paper does not prove it, and those preprints are not yet refereed. There are also a few 'readily verified' reductions (Lemma 3.13, the m=k=1 reduction in Lemma 4.8) that deserve a little more detail, but those are minor.\n\nThe reader's report flags effectivity as the main black box; I think that is partly off-target. Effectivity is used in the proof of Theorem 3.10, but the central claim of Section 4 rests on Lemma 4.4. Both dependencies deserve scrutiny, but the second is the one to check carefully.\n\nWho this is for: anyone relying on GR's (∞,2)-categorical foundations in derived algebraic geometry, and anyone working on Gray tensor products or double ∞-categories. The paper is not revolutionary in technique—the tools are mostly established—but it closes a genuine gap in an influential framework.\n\nRecommendation: send it to a good referee. The referee should be asked to verify Lemma 4.4 and its cited sources, and to check Theorem 3.10's dependency. If those hold, I'd accept.","headline":"Credible proof of the last Gaitsgory–Rozenblyum conjecture, with a real black-box dependency on the authors' own unpublished preprints.","tokens_in":19184,"tokens_out":2775,"would_cite":true,"duration_ms":28261,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["18N10","18N60"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves the final open Gaitsgory–Rozenblyum conjecture: the squares functor and the Gray tensor product are linked by a natural equivalence.","keywords":["(∞,2)-categories","double ∞-categories","Gray tensor product","squares functor","companions","directed Čech nerve","Gaitsgory–Rozenblyum conjectures","higher category theory"],"falsifier":"The quickest check is to compute both sides for the first non-trivial shapes: take $C=[1]$, $D=[1]$, and $E=[1;1]$, and compare $\\mathrm{Map}(C^h\\times D^v,\\mathrm{Sq}(E))$ with $\\mathrm{Map}(C\\otimes D,E)$; if they are not equivalent, Theorem B is false. The paper's own reduction says it suffices to check globular sums, so a reader could also verify the pushout decompositions of Lemmas 4.6 and 4.8 directly in the category of gaunt 2-categories.","tokens_in":18189,"feed_emoji":"⬜","tokens_out":11006,"duration_ms":103973,"temperature":0.7,"pith_summary":"This paper proves the final unresolved conjecture in the $(\\infty,2)$-categorical foundations of Gaitsgory and Rozenblyum: for any $(\\infty,2)$-categories $C,D,E$ there is a natural equivalence $\\mathrm{Map}(C^h\\times D^v,\\mathrm{Sq}(E))\\simeq \\mathrm{Map}(C\\otimes D,E)$, which is equivalently the statement $\\mathrm{Gr}(C^h\\times D^v)\\simeq C\\otimes D$. It also establishes the universal property of the squares functor: $\\mathrm{Sq}(C)$ is the double $\\infty$-category freely obtained from the vertical inclusion $C^v$ by adjoining companions, with a dual statement for the horizontal inclusion. The result closes the last of eight conjectures from [GR17], and the uniqueness of the comparison follows because the Gray tensor product has no nontrivial automorphism. A reader should care because the Gray tensor product is the central device for controlling lax phenomena in higher category theory, and this gives it a description purely in terms of squares and double categories.","feed_headline":"Squares functor settles last Gaitsgory-Rozenblyum conjecture","feed_subtitle":"Theorem B equates mapping spaces for squares and the Gray tensor product, closing the last of eight conjectures.","key_machinery":"The machine at the center is the squares functor $\\mathrm{Sq}$, which sends an $(\\infty,2)$-category $C$ to the double $\\infty$-category whose objects are the objects of $C$, whose horizontal and vertical arrows are the arrows of $C$, and whose 2-cells are lax commutative squares in $C$; its left adjoint $\\mathrm{Gr}$ realizes a double $\\infty$-category as an $(\\infty,2)$-category. Around this sit the directed Čech nerve of a filtration, which is the relative form of squares, and the notion of a companion: a horizontal arrow $F$ companion to a vertical arrow $f$ is witnessed by a unit and a counit satisfying triangle identities. The proof reduces arbitrary $(\\infty,2)$-categories to globular sums and assembles $\\mathrm{Gr}(C^h\\times D^v)$ from pushouts, using Maehara's Gray tensor product on space-valued presheaves over $\\Theta_2$ to identify the resulting colimit with $C\\otimes D$.","core_discovery":"The central claim is Theorem B: the squares functor and the Gray tensor product are governed by the same universal property. Concretely, for $(\\infty,2)$-categories $C,D,E$, mapping out of the product of the horizontal inclusion of $C$ and the vertical inclusion of $D$ into the squares of $E$ is naturally equivalent to mapping out of $C\\otimes D$ into $E$. Since $\\mathrm{Sq}$ has a left adjoint $\\mathrm{Gr}$, this is equivalent to $\\mathrm{Gr}(C^h\\times D^v)\\simeq C\\otimes D$, settling [GR17, Proposition 10.4.5.4]. The paper also proves Theorem A, the universal property of squares: functors $\\mathrm{Sq}(C)\\to Q$ are exactly functors $C^v\\to Q$ that send every arrow of $C$ to a vertical arrow of $Q$ admitting a companion, and the analogous statement for horizontal arrows holds under a local completeness assumption. Because Proposition 2.18 shows the Gray tensor product has no nontrivial endomorphisms, the comparison map originally written down by Gaitsgory and Rozenblyum must coincide with the one produced here.","pith_inferences":["By the same density argument, the equivalence $\\mathrm{Gr}(C^h\\times D^v)\\simeq C\\otimes D$ is likely to extend to $(\\infty,n)$-categories with an $n$-dimensional squares functor, but the paper does not claim this.","Because the proof invokes effectivity of generalized double $\\infty$-categories and the companion theorems as black boxes, a fully self-contained proof would need those results checked independently; the paper does not provide that check.","The uniqueness of the comparison map suggests that any construction of a Gray tensor product satisfying the same universal property will automatically agree with Maehara's, which may simplify future comparisons among tensor products."],"forward_implications":["All eight Gaitsgory–Rozenblyum conjectures about the Gray tensor product and the squares functor are now resolved.","The Gray tensor product can be described as $\\mathrm{Gr}(C^h\\times D^v)$, giving a double-categorical description of the tensor product.","Theorem A gives a usable mapping-space criterion: maps out of $\\mathrm{Sq}(C)$ are detected by maps out of $C^v$ that preserve companions, and dually for $C^h$.","Any comparison between different models of the Gray tensor product is unique, because the bifunctor has no nontrivial endomorphisms."],"supporting_citations":[{"why":"states the eight conjectures and defines the squares functor, Gray tensor product, and the comparison map that Theorem B proves.","marker":"[GR17]"},{"why":"supplies effectivity of generalized double ∞-categories, used in Theorem 3.10, and the tensor product ⊗L compared in Proposition 2.22.","marker":"[Lou25]"},{"why":"provides the companion theorems and the constructions of the horizontal and vertical inclusions C^h and C^v used in Theorems 3.14 and 4.1.","marker":"[Rui25a]"},{"why":"defines the Gray tensor product on 2-quasi-categories and proves it preserves colimits and matches the gaunt tensor product on globular sums.","marker":"[Mae21]"},{"why":"supplies colimit computations in presheaf categories used in Lemmas 2.20 and 4.8.","marker":"[Lou24]"},{"why":"formulated the universal property of squares as a conjecture that Theorem A proves.","marker":"[Rui25b]"}],"fun_headline_variants":["Squares functor proves last Gaitsgory-Rozenblyum conjecture","Universal property of squares functor clinches final GR conjecture","Squares functor's universal property closes G-R's last open conjecture","Theorem B unifies squares and Gray tensor, solving GR's last conjecture","Squares functor's universal property settles final Gaitsgory-Rozenblyum conjecture"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof takes several results from the authors' earlier preprints as black boxes—effectivity of generalized double ∞-categories, the companion theorems, and the identification of the Gray tensor product used here with Gaitsgory–Rozenblyum's—so the central equivalence stands only if those cited theorems are correct.","fun_headline_variants_meta":{"raw":{"variants":["Squares functor proves last Gaitsgory-Rozenblyum conjecture","Universal property of squares functor clinches final GR conjecture","Squares functor's universal property closes G-R's last open conjecture","Theorem B unifies squares and Gray tensor, solving GR's last conjecture","Squares functor's universal property settles final Gaitsgory-Rozenblyum conjecture"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000751,"raw_usage":{"total_tokens":3313,"prompt_tokens":882,"completion_tokens":2431,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":498,"completion_tokens_details":{"reasoning_tokens":2331}},"tokens_in":498,"tokens_out":2431,"duration_ms":19643,"temperature":1.0,"reasoning_tokens":2331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:31:52.281915+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The quickest check is to compute both sides for the first non-trivial shapes: take $C=[1]$, $D=[1]$, and $E=[1;1]$, and compare $\\mathrm{Map}(C^h\\times D^v,\\mathrm{Sq}(E))$ with $\\mathrm{Map}(C\\otimes D,E)$; if they are not equivalent, Theorem B is false. The paper's own reduction says it suffices to check globular sums, so a reader could also verify the pushout decompositions of Lemmas 4.6 and 4.8 directly in the category of gaunt 2-categories.","supporting_citations":[],"review_version":1}