{"id":"a5e48515-b339-48b3-b379-fa62390a4052","arxiv_id":"2507.15445","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Given a cyclic A-infinity category, a set of objects and a splitting of the non-commutative Hodge filtration, an L-infinity quasi-isomorphism trivializing the open-closed string field theory algebra is constructed.","lead":"This paper constructs an open-closed formality morphism for Calabi-Yau categories, extending a known closed-string result to include open-string sectors. The result is a step toward defining open-closed Gromov-Witten invariants categorically and quantizing large N open string field theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Graph bijection in Theorem 5.1.6 ignores loop/genus multiplicities; Theorem 5.2.1 fails for m=2, so the proof of Theorem A rests on a false identity.","rationale":"The reader's weakest assumption concerned the regrading of the closed formality morphism from [CT24] and asserted graph compatibilities. I agree with the latter concern but locate a sharper, internal failure. The graph bijection of Theorem 5.1.6 does not account for the multiplicity of loop/genus labels when a vertex is split; in the minimal case of a two-leaf corolla the source set A has a g-independent number of elements while the target set C has g+1 distributions. This is not a disagreement with an external convention: it is a counting contradiction inside the definitions of Section 4. As a consequence, the m=2 case of Theorem 5.2.1 is numerically false by a γ^1 coefficient check. Since Theorem 5.2.1 is the sole input to Lemma 5.0.10, the L∞ relation for Ψ^oc is not established. The paper's Theorem A may be salvageable with a corrected graph identity that tracks genus labels, but the present proof does not contain it. The verdict should therefore be REJECT rather than CONDITIONAL: the required correction is central, not a local sign fix.","tokens_in":21486,"tokens_out":33601,"duration_ms":377980,"concrete_test":"Verify Theorem 5.2.1 directly for m=2 with x,z of word-length one. Compute both sides from the graph sums (4.0.2) and rule 4.0.8, and project to the Sym^2(H) component. The left side equals (Σ_{g≥0}γ^g)x·z; the right side's product term equals (Σ_{g≥0}γ^g)^2 x·z, while γK^2_s(x,z) is a scalar in Sym^0(H). The γ^1 coefficients therefore differ (1 versus 2, up to the common leaf-label normalization), so the identity fails. Equivalently, count A_{g,2}^{1,1} and C_{g,2}^{1,1} for g>0: the former is g-independent, the latter has g+1 loop-defect distributions, contradicting Theorem 5.1.6.","verdict_should_be":"REJECT","load_bearing_attack":"The proof of Theorem A is reduced in Section 5.2 to Theorem 5.2.1, which is derived from the graph bijection ψ of Theorem 5.1.6. That bijection is not valid as stated because the vertex loop-defect labels, which are part of the 'labeled graphs' of Section 4, are not tracked by the half-edge splitting. Fix m=2, k1=k2=1, n=2 and take G to be the one-vertex graph with two leaves. For each total betti number g, the number of split data in A_{g,2}^{1,1} is independent of g (one or two, depending on leaf labeling), while the corresponding target set C_{g,2}^{1,1} contains a distinct pair of one-vertex, one-leaf graphs for every loop-defect distribution (g1,g2) with g1+g2=g, giving g+1 times as many elements. Hence no bijection ψ exists for g>0. This failure propagates to Theorem 5.2.1. For x,z of word-length 1, definition 4.0.8 gives K^1_s(y) = (Σ_{g≥0} γ^g)y, so the left side of the m=2 case has γ^1-coefficient x·z in Sym^2(H), while K^1_s(x)K^1_s(z) has coefficient 2x·z and γK^2_s(x,z) lies in Sym^0(H), so it cannot repair the mismatch. The proof of Lemma 5.0.10, and therefore of Theorem A, collapses.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to construct an open-closed formality L∞-quasi-isomorphism for a cyclic A∞-category C with a splitting of the non-commutative Hodge filtration and a full subcategory Λ, trivializing the closed operations while keeping the open operations intact (Theorem A, Corollary 5.2.14). The construction is a graph-dependent morphism on tensor products of closed and open string field theory BD algebras. The proof reduces Theorem A to a combinatorial identity for the Taylor components of the closed formality morphism, Theorem 5.2.1, which is derived from a graph bijection in Theorem 5.1.6. The paper also sketches a conditional quantization statement (Theorem C) under an external Assumption (*) that is delegated to work in preparation [AT25].","tokens_in":21803,"tokens_out":9941,"duration_ms":111800,"significance":"If Theorem A were correct, it would provide a natural open-closed analogue of the closed formality morphism of Caldararu–Tu and Amorim–Tu, with potential applications to categorical open Gromov–Witten invariants and to quantization of large-N open string field theory. The paper is clearly organized, makes an explicit effort to track shifts and gradings, and isolates the combinatorial core of the argument. However, the central combinatorial bijection on which Theorem A rests is false as stated, so the main theorem is not established in this version. The conditional quantization statement is also explicitly dependent on an unverified assumption and on in-preparation work.","major_comments":[{"comment":"The asserted bijection ψ is not valid because it does not track the vertex loop-defect labels introduced in Section 4. For m=2, k1=k2=1, n=2, take G to be a one-vertex graph with two leaves and vertex defect g. The domain set A_{g,2}^{1,1} has cardinality independent of g (one or two elements, depending on leaf-labeling conventions), while the target set C_{g,2}^{1,1} contains a distinct pair of one-vertex, one-leaf graphs for every decomposition g1+g2=g, hence g+1 elements for g>0. No bijection can exist for such g. Moreover, Construction 5.1.1 does not specify how the loop-defect label of the split vertex is distributed between the old and new vertices, so ψ is not even well-defined as a map of the labeled graphs introduced in Section 4. Since Theorem 5.1.6 is the basis for Theorem 5.2.1, the proof chain leading to Theorem A is broken.","section":"Section 5.1, Theorem 5.1.6"},{"comment":"The proof of Theorem 5.2.1 asserts after equations (5.2.8) and (5.2.9) that “the evaluation maps are compatible with this bijection” and with the identifications (5.1.4) and (5.1.5). This compatibility is not demonstrated, and in the m=2 example above it is false: the sum over C_{g,2}^{1,1} contains more terms than the original sum over A_{g,2}^{1,1} for g>0, so the regrouped sums have different numbers of contributions with the same γ-weight. Consequently, the equality claimed in Theorem 5.2.1 does not follow from Theorem 5.1.6, and Lemma 5.0.10 is not proven.","section":"Section 5.2, proof of Theorem 5.2.1"},{"comment":"The reduction of Theorem 5.0.6 to Lemma 5.0.10 assumes that the closed L∞-morphism K_s of [CT24, Theorem 7.1] and [AT22, equation (26)] satisfies identity (5.0.9) verbatim under the paper's different grading and shift conventions. The paper notes that it “carefully keeps track of the Z-grading” but does not reproduce or prove this compatibility. Since (5.0.9) is used to cancel terms in equation (5.0.7), any sign or γ-degree mismatch would invalidate the claimed equivalence between Theorem 5.0.6 and Lemma 5.0.10. This step is load-bearing and cannot be checked from the present text.","section":"Section 5, equations (5.0.8) and (5.0.9)"}],"minor_comments":[{"comment":"The condition “0≤ g, n < 0” is nonsensical; it should read “0≤g and 0<n” or the like, since n denotes the number of leaves.","section":"Section 5.1, Definition 5.1.3"},{"comment":"The proof uses the notation ψ(pG,f,I,Jq) before the bijection ψ of Theorem 5.1.6 has been extended to triples; the domain of the map should be made explicit.","section":"Section 5.2, proof of Theorem 5.2.1"},{"comment":"The statement writes x1,...,xm ∈ F^c(C)r6−2ds, but the proof treats xi and xj as elements of Sym^{k1} and Sym^{k2} respectively; the word-length decomposition should be stated in the theorem.","section":"Section 5.2, statement of Theorem 5.2.1"},{"comment":"There are several typographical issues, including “thereom” before Theorem 5.0.6 and inconsistent non-ASCII characters in expressions such as “p2d´5q-twisted”; a careful proofreading pass is needed.","section":"Throughout"}],"recommendation":"reject","confidential_remarks":"The graph-counting counterexample to Theorem 5.1.6 is decisive and appears to invalidate the proof of Theorem A. If the author can repair the construction by incorporating loop-defect distributions or by changing the definition of the domain sets, a substantially revised version could be reconsidered; as it stands, the main theorem is not established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main theorem is unproved. The proof of Theorem A reduces to a graph identity, Theorem 5.2.1, which rests on the bijection in Theorem 5.1.6. That bijection is false as stated because it never accounts for the loop-defect labels on vertices. The \"labeled graphs\" of Section 4 carry a vertex labeling g: V→Z_{≥0}, and the betti number sums these labels with rank H1. When Construction 5.1.1 splits a vertex into two, it says nothing about how the loop defect divides; when the inverse merges two vertices, it says nothing about adding the two loop defects. The cardinality argument is decisive: for m=2, k1=k2=1, n=2, start with the one-vertex, two-leaf graph with total betti g. The set A has one or two elements, independent of g. The target side B is empty, and C contains a distinct pair of one-vertex, one-leaf graphs for each (g1,g2) with g1+g2=g, so g+1 elements. No bijection exists for g>0. Consequently Theorem 5.2.1 fails for m=2: for x,z of word-length 1, K^1(xz) has γ^1 coefficient xz, while K^1(x)K^1(z) has coefficient 2xz and γK^2(x,z) contributes only in Sym^0. The identity cannot hold.\n\nThis is a real pity, because the paper does useful work. The idea of extending the closed formality morphism of [CT24] to the open-closed tensor product is natural, and the explicit BD-algebra setup, the careful shifts, and the LQT context are clearly explained. Theorem C is honestly conditional on Assumption (*), which is flagged as depending on [AT25]. The graph-theoretic interlude is meant to be the core auxiliary result, and it is where the error appears.\n\nThe reader's earlier concern about regrading [CT24] is secondary. Even granting the closed formality under the paper's shifts, the open-closed proof fails at the level of graph combinatorics. This is not a minor gap; it is the load-bearing identity. The statement of Theorem A may be salvageable with a revised graph construction that splits loop defects, but the present draft does not contain a valid proof.\n\nI would not cite this in its current form, and I would not send it back for minor revision. That said, the direction is important enough that a serious referee could usefully lay out the necessary repair. If the authors fix the loop-defect issue and rework Theorem 5.2.1, the paper could become a solid contribution.\n\nRecommendation: send it to a referee who is willing to check the graph combinatorics in detail, with the expectation of a major revision. The flaw is specific and demonstrable, and the surrounding framework is worth the referee time.","headline":"The graph bijection underpinning the main theorem doesn't track loop defects, so the proof of Theorem A collapses; the open-closed idea is plausible but this draft isn't ready.","tokens_in":22346,"tokens_out":7406,"would_cite":false,"duration_ms":73738,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["16E40","14N35","53D45"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that, after choosing a splitting of the non-commutative Hodge filtration, the open-closed string field theory of a Calabi-Yau category is $L_\\infty$ quasi-isomorphic to a much simpler complex, a necessary step for…","keywords":["categorical enumerative invariants","open-closed string field theory","L∞ quasi-isomorphism","Beilinson-Drinfeld algebras","Calabi-Yau A∞ categories","non-commutative Hodge filtration","circle action formality","cyclic A∞ categories"],"falsifier":"Take a simple example—say, a dimension-$d$ cyclic $A_\\infty$-category with trivial compositions and the identity splitting—and compute both sides of the $L_\\infty$ relation (2.4.7) for the proposed map on two inputs. A sign or degree mismatch in the regraded closed formality map would show up as a nonzero difference, refuting Theorem A.","tokens_in":21213,"feed_emoji":"🔁","tokens_out":11062,"duration_ms":106919,"temperature":0.7,"pith_summary":"The paper aims to extend the circle-action formality trivialization from closed to open-closed string field theory. It claims that, for a cyclic Calabi-Yau category together with a collection of objects and a splitting of the non-commutative Hodge filtration, the open-closed string field theory is $L_\\infty$ quasi-isomorphic to a dramatically simpler complex in which the closed sector retains only the Hochschild differential. This simplification is what makes it feasible to read off open-closed enumerative invariants and to quantize the large-$N$ open string field theory. The author frames the result as the open-closed analogue of the closed formality theorem, and as a step toward categorical open-closed Gromov-Witten invariants.","feed_headline":"Open-closed string field theory becomes homotopy trivial","feed_subtitle":"The closed sector shrinks to a bare differential, a step toward open-closed Gromov-Witten invariants.","key_machinery":"The machinery is the closed circle-action formality morphism $K_s$, extended to a map on the tensor product with the open sector. The paper's new map is defined on symmetric powers by sending $(x_1\\otimes y_1)\\cdots(x_n\\otimes y_n)$ to $K_s^n(x_1\\cdots x_n)\\otimes(y_1\\cdots y_n)$ up to Koszul signs, with Taylor coefficients $K_s^n$ built from marked graphs with loop defects whose edges are contracted using the symmetric bilinear form $H_s^{\\mathrm{sym}}$ determined by the Hodge-filtration splitting $s$. The proof that this is an $L_\\infty$ quasi-isomorphism turns on a combinatorial lemma (Theorem 5.1.6): a bijection between a moduli space of graphs on $m-1$ vertices with a partition of the first vertex's half-edges and a union of two moduli spaces of graphs on $m$ vertices (connected, and two-component). This bijection, together with the Leibniz rule in the open BD algebra, yields the key identity Theorem 5.2.1, which makes the map commute with the differentials. In short, the engine is a graph-moduli bijection that reconciles contracting two closed inputs at once with distributing them into separate closed factors.","core_discovery":"The central discovery is a formality theorem for open-closed string field theory: for a dimension $d$ cyclic $A_\\infty$-category $\\mathcal{C}$, a splitting $s$ of the non-commutative Hodge filtration, and a full subcategory $\\Lambda$, there exists an $L_\\infty$ quasi-isomorphism $\\Psi^{\\mathrm{oc}}_s : F^c(\\mathcal{C}) \\otimes F^{\\mathrm{op}}(\\Lambda)[5-2d] \\dashrightarrow F^c(\\mathcal{C})_{\\mathrm{Triv}} \\otimes F^{\\mathrm{op}}(\\Lambda)[5-2d]$. Here $F^c(\\mathcal{C})$ is the closed string field theory Beilinson-Drinfeld algebra built from Hochschild chains, $F^{\\mathrm{op}}(\\Lambda)$ is the open sector built from cyclic cochains on $\\Lambda$, and $F^c(\\mathcal{C})_{\\mathrm{Triv}}$ retains only the Hochschild differential on the closed side. The morphism is constructed explicitly by a graph calculus: the Taylor components are built by contracting half-edges according to marked graphs using the symmetric bilinear form coming from the splitting $s$. The proof reduces to a combinatorial identity (Theorem 5.2.1) about the closed formality morphism's Taylor components, which is proved via a bijection between certain moduli spaces of graphs. The author argues this is the open-closed analogue of the closed circle-action trivialization and an ingredient toward quantizing large-$N$ open string field theory, conditional on an assumption about the existence of a Maurer-Cartan element whose open part is the cyclic potential.","pith_inferences":["Not in the paper: the same graph-moduli bijection might give a direct proof that the open-closed formality morphism is compatible with the full $L_\\infty$ structure without relying on the regraded closed formality theorem, which would remove the paper's main sign-convention vulnerability.","Not in the paper: the invariants extracted from $\\Psi^{\\mathrm{oc}}_s$ could depend on the chosen Hodge-filtration splitting; if so, studying this dependence would connect to wall-crossing phenomena in open Gromov-Witten theory, mirroring the closed case.","Not in the paper: since the paper works with cyclic (strict) Calabi-Yau categories, a natural extension would be to weak proper Calabi-Yau categories, using known strictification results; the closed case has already been treated at that level of generality.","Not in the paper: the author stops short of computing actual numerical invariants from the trivialized complex; a concrete low-genus or one-object example would test whether the proposed formalism is computationally accessible."],"forward_implications":["Choosing a splitting of the non-commutative Hodge filtration, the open-closed string field theory of a cyclic Calabi-Yau category with a set of objects is $L_\\infty$ quasi-isomorphic to a complex whose closed sector is just Hochschild chains with its Hochschild differential; this is Theorem A.","Via the trivialized complex, the image of the Maurer-Cartan element of the theory gives coefficients that are conjecturally the open-closed Gromov-Witten invariants of the corresponding Fukaya category, mirroring the closed case.","If Assumption (*) holds, Theorem C constructs a quantization of the cyclic $A_\\infty$ algebra of an object, i.e., a Maurer-Cartan element in the open string field theory whose dequantization is the cyclic potential; the quantization receives contributions from the closed sector.","The new graph-moduli bijection (Theorem 5.1.6) relating graphs with different numbers of vertices is presented as a combinatorial result of independent interest."],"supporting_citations":[{"why":"Supplies the closed formality morphism $K_s$ whose Taylor components define the new map.","marker":"[CT24]"},{"why":"Provides equation (26) recording that the closed Taylor maps form an $L_\\infty$ morphism, a step used to reduce the proof to Lemma 5.0.10.","marker":"[AT22]"},{"why":"Introduces marked graphs and the moduli spaces $\\check{\\Gamma}(g,n)_m$ used in all Feynman contractions.","marker":"[GK98]"},{"why":"Gives the tensor-product formula for BD algebras (Lemma 2.1.4) that produces the open-closed bracket in the $L_\\infty$ relations.","marker":"[LZ14]"},{"why":"Defines the string-field-theory observables $F^c(\\mathcal{C})$ and the open-closed Maurer-Cartan framework the paper works within.","marker":"[Cos09]"},{"why":"Defines the open string field theory algebra $F^{\\mathrm{op}}(\\Lambda)$ on cyclic cochains, the open factor in the tensor product.","marker":"[Ulm25]"}],"fun_headline_variants":["Open-closed SFT rendered homotopy trivial","Formality theorem for open-closed string field theory","Graph calculus proves open-closed formality","Circle-action formality morphism for open-closed invariants","Homotopy trivialization in open-closed SFT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof takes as given that the closed formality morphism from earlier work remains an $L_\\infty$ quasi-isomorphism after the paper's regrading and shift conventions, and that the graph-evaluation maps are compatible with the new combinatorial bijection; if those signs and compatibilities fail, the reduction to the key lemma collapses.","fun_headline_variants_meta":{"raw":{"variants":["Open-closed SFT rendered homotopy trivial","Formality theorem for open-closed string field theory","Graph calculus proves open-closed formality","Circle-action formality morphism for open-closed invariants","Homotopy trivialization in open-closed SFT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000301,"raw_usage":{"total_tokens":1766,"prompt_tokens":1005,"completion_tokens":761,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":621,"completion_tokens_details":{"reasoning_tokens":684}},"tokens_in":621,"tokens_out":761,"duration_ms":7933,"temperature":1.0,"reasoning_tokens":684,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:32:01.408846+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a simple example—say, a dimension-$d$ cyclic $A_\\infty$-category with trivial compositions and the identity splitting—and compute both sides of the $L_\\infty$ relation (2.4.7) for the proposed map on two inputs. A sign or degree mismatch in the regraded closed formality map would show up as a nonzero difference, refuting Theorem A.","supporting_citations":[],"review_version":1}