{"id":"0ce7bd28-4b57-4fa6-a16f-9c55e9b7a78c","arxiv_id":"2508.07727","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A self-contained geometric proof of the wall-crossing formula via path-lifting rules for spectral networks, including spiral domains, with the charge lattice generated by A0-laminations.","lead":"The paper proves the Kontsevich-Soibelman wall-crossing formula using only the geometry of quadratic differentials and spectral networks, settling a convergence question left open by Gaiotto, Moore and Neitzke. Its path-lifting framework works even when trajectories are dense, and it generates the charge lattice from laminations, giving a fully geometric proof instead of the earlier Donaldson-Thomas machinery.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1 hinges on the figure-verified detour enumeration in Lemmas 6.5–6.7; a missing or sign-wrong detour there would break F+ = K F− and hence the wall-crossing theorem.","rationale":"The reader and I identify the same load-bearing point. The paper’s own structure makes this precise: Theorem 1.3(iv) is Theorem 6.3, whose proof is “Proposition 6.8 + concatenation”; Proposition 6.8 is exactly the finite comparison at a rank-one active direction, and its only input is the figure-based enumeration in Lemmas 6.5–6.7. This is therefore the place where the central claim is least secure. The other pillars—formal completion, convergence of path liftings, approximation by A0-laminations—are supported by explicit propositions and arguments; the finite enumeration is not independently verified. This is not an external-consensus objection; it is a correctness risk internal to the proof. The paper honestly records its limitations (no simple poles, finite-area conditional), and the BPS invariants match known values, which is independent support. Because the concern is about an unverified finite case check rather than a demonstrated contradiction, the right disposition is to keep the reader’s CONDITIONAL verdict. If the proposed enumeration check passes, the main theorem is substantially confirmed; if it fails, Theorem 6.3 (and hence Theorem 1.1) would need revision. I would not escalate to REJECT on the present evidence.","tokens_in":44085,"tokens_out":19709,"duration_ms":234649,"concrete_test":"Write a small independent enumerator for the local models in Lemma 6.4: a type I saddle connection (Case 1), a two-boundary cylinder (Case 4a), and a toral cylinder (Case 4b). Input the detour rules of §4.1/§6.2 (monotone sequences, ±-turns, American/British driving rule) and generate all elementary detours at a single intersection point up to N cylinder twists (N=5), for both orientations of ℘ and both sheets. Compute each detour’s class in the signed homology groupoid and its intersections with L(bγ0) and L(2bγ0); compare with Lemmas 6.5–6.7. Then symbolically substitute the generated classes into the equations of Proposition 6.8 and verify F+ = KθF−. Any mismatch—missing detour family, extra detour, or sign error—would pinpoint the failure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing assumption is the completeness and sign correctness of the local case analysis proving Theorem 6.3. In the proof of Theorem 6.3, the global equality F+(℘,θ) = KθF−(℘,θ) is reduced by truncation and concatenation to the single-intersection statement Proposition 6.8. Proposition 6.8 is then verified case-by-case using the lists of ±-admissible detours and intersection numbers in Lemmas 6.5–6.7. Those lemmas are justified by inspection of Figures 11–14 (e.g., Lemma 6.5: “The proof … follows by inspection of Figure 11, checking that it contains all ±-admissible rays and checking the claimed intersection numbers”). The later algebraic identities in Proposition 6.8 are sign-sensitive: for instance, in Case (4b), the prefactor (1−[eγ0])−1(1−[eγ0]^2) = 1+[eγ0] depends both on ⟨L(γ0),D⟩ = −1 and ⟨L(2γ0),D⟩ = +1 for the relevant detours; if either sign is wrong, K acts by the wrong factor and F+ = KF− fails at that rank-one wall. Since Theorem 1.1’s proof invokes Theorem 6.3 through Proposition 6.10, a missing detour family or a sign error in Lemmas 6.5–6.7 invalidates the central claim. No alternative algebraic derivation of these lists is supplied.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper gives a self-contained proof of the Kontsevich–Soibelman wall-crossing formula for holomorphic quadratic differentials, working directly with spectral networks and path-lifting rules rather than motivic Donaldson–Thomas theory. The main theorem (Theorem 1.1) states that for infinite-area quadratic differentials without simple poles, the BPS automorphism S_Δ is covariantly constant along paths of homology-framed differentials whose boundary rays stay saddle-free, even if intermediate surfaces are not rank one. The proof constructs a completed signed path groupoid, defines path-lifting functions F(℘,θ) with convergence results, extends them to A0-laminations in order to generate the hat-homology lattice, and then compares the one-sided limits of the lifting at rank-one walls via a local detour computation. A finite-area version is stated only under an additional approximation hypothesis (Proposition 1.2), and the authors explicitly note that the finite-area wall-crossing statement is nearly void under non-active boundary rays.","tokens_in":44380,"tokens_out":7193,"duration_ms":92391,"significance":"If correct, this is a substantial contribution: it provides a proof of the Kontsevich–Soibelman wall-crossing formula entirely within the geometry of quadratic differentials, resolves convergence questions for spectral networks with dense trajectories and spiral domains, and confirms the richness of the GMN path-lifting rules. The intermediate result that lifts of A0-laminations generate the hat-homology lattice is of independent interest. The paper is carefully structured and does not rely on fitted parameters or external DT input. The main risk to rigor is that the central wall-crossing comparison is reduced to a finite enumeration of detour configurations that is verified by inspection of figures rather than by a systematic proof; this is the load-bearing gap discussed below.","major_comments":[{"comment":"The completeness and sign correctness of the local detour enumeration is load-bearing for Theorem 6.3, and hence for Theorem 1.1. The proofs of Lemmas 6.5–6.7 are justified 'by inspection' of Figures 11–14, and Proposition 6.8 applies them case-by-case. A missing admissible ray family or a sign error in an intersection number would change the BPS automorphism K and break F^+=KF^-; for instance, Eq. (53) uses both ⟨L(γ0),D⟩ and ⟨L(2γ0),D⟩, and a single sign flip would replace the prefactor (1+[eγ0]) by (1−[eγ0]). I request that the inspection argument be replaced by a systematic enumeration: a complete list of ±-admissible ray types for each configuration, their turn sequences, and all intersection numbers, or a machine-checked verification. As written, the central proof is not fully verifiable without trusting the completeness of the pictures.","section":"§6.2, Lemmas 6.5–6.7 and Proposition 6.8"},{"comment":"The proof of the one-sided limits relies on Lemma 6.2, which asserts a bijection between detours in nearby directions and ±-admissible detours in the limit. While the rectangle argument for a single ray is plausible, the passage from rays to full detours involves collisions of starting points and the American/British driving rule, which is only discussed heuristically via Figure 10. Since Proposition 6.1 supplies the limits in Theorem 1.3(iv) and is used in Proposition 6.10, please expand this step into a precise statement of the bijection at the level of detour sequences, including multiplicities and the handling of colliding starting points.","section":"§6.1, Lemma 6.2 and Proposition 6.1"}],"minor_comments":[{"comment":"The order of the product in the definition of S_Δ is inconsistent: the introduction says the product is taken in clockwise order, while Eq. (54) says counterclockwise order. Since the factors K_θ need not commute in general, please harmonize the convention.","section":"Introduction vs. §6.3"},{"comment":"Typo: 'suppose the there are sufficiently many based path lifting functions' should read 'suppose there are sufficiently many based path lifting functions'.","section":"Proposition 1.2"},{"comment":"The text refers to an 'irrational torus end', but the classification in Lemma 6.4 only speaks of a spiral domain whose interior is a torus; the word 'irrational' is unexplained and potentially misleading.","section":"§6.2, proof of Case (4b)"},{"comment":"In the statement, 'defines an element F(℘,θ)' should be 'F(L,θ)' for the lamination lift.","section":"Proposition 5.3"},{"comment":"There is a typographical error in the formula: 'c([γ_e]^{-1}])' has an extra closing bracket. Please also clarify the value of c in terms of the base point.","section":"Lemma 5.10(iv)"},{"comment":"The product notation with upper and lower indices (e.g. 'Π_{i=1}^k ...') is hard to read; please clarify the summation indices and the ranges of i,j in the displayed formulas.","section":"Lemma 5.6"}],"recommendation":"major_revision","confidential_remarks":"The paper is long and technical, and the main theorem is likely correct in spirit. The decisive issue is the figure-verified enumeration in Lemmas 6.5–6.7; if the authors can supply a rigorous enumeration or a reproducible computational check, I would be willing to accept. The finite-area part is explicitly conditional and honestly discussed, so it is not a bar to publication. The introduction/body inconsistency about the order of the S_Δ product should be fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nYou should know this about arXiv:2508.07727: if correct, it is the first fully self-contained proof of the Kontsevich–Soibelman wall-crossing formula inside quadratic differential theory, with the convergence issues from [GMN13b, Question 7] settled for infinite-area surfaces. The interesting new pieces are the profinite convergence framework for path liftings, the extension of the lifting rule to A0-laminations (which generates the hat-homology lattice), and the proof that one-sided limits at a rank-one wall are related by the BPS automorphism. The paper is honest about its boundaries: Theorem 1.1 excludes simple poles, and the finite-area statement (Prop. 1.2) is explicitly conditional. The abstract, however, promises more than the theorem delivers, and should be narrowed.\n\nThe main theorem is proven by reducing the global wall-crossing to a local statement, Proposition 6.8, about detours at a single intersection with the spectral network. That reduction is elegant. The soft spot is exactly where the stress-test says it is: the local statement rests on the case-by-case enumeration of ±-admissible detours in Lemmas 6.5–6.7, and those lemmas are verified by inspection of Figures 11–14. The subsequent algebra is sign-sensitive—for instance in Case (4b) the prefactor (1−[eγ0])−1(1−[eγ0]^2)=1+[eγ0] depends on both intersection numbers being right. A missing detour family or a sign error there would break F+=KF−, and the lists are not otherwise derived. There is no reason to suspect such an error, but this is a load-bearing verification-by-picture, and independent checking is needed.\n\nThe paper is not circular: the BPS invariants are computed from saddle connections and ring domains via the geometry, matching known values, and there are no fitted parameters. The citation pattern is fine. My overall take is close to the reader's: the architecture is sound and the result is significant, but the proof is not yet fully auditable at the crucial local step.\n\nWho is this for? Anyone working on spectral networks, quadratic differentials and DT theory. It deserves a serious referee, but the referee should be asked to check the detour enumeration with care, and the authors should be asked to either replace figure-inspection by a systematic enumeration or at least record the verification in a way that does not depend on the reader seeing what the authors see. I would engage with it, but not rely on it for downstream results until the local case analysis is tightened.\n\nBest.","headline":"A serious and significant proof of wall-crossing via spectral networks, but its load-bearing local detour enumeration is verified by pictures rather than algebra and should be independently checked.","tokens_in":44947,"tokens_out":2304,"would_cite":true,"duration_ms":28460,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["30F30","32G15","14N35","37F34"],"pacs":[],"model":"deepseek-v4-flash","headline":"Spectral networks alone prove the wall-crossing formula","keywords":["quadratic differentials","spectral networks","wall-crossing formula","BPS invariants","path lifting","A0-laminations","hat-homology","saddle connections"],"falsifier":"Work out a concrete rank-one configuration (e.g. a saddle connection between two simple zeros on a genus-two surface) and explicitly enumerate all ±-admissible rays and their intersection numbers with the BPS cycle $L(\\hat\\gamma_0)$, either by hand or by computer; if any admissible detour family is missing from Figures 11–14 or any intersection number in Lemmas 6.5–6.7 is wrong, then the identity $F^+(\\wp,\\theta_0)=K_{\\theta_0}F^-(\\wp,\\theta_0)$ fails for a single-crossing path. Alternatively, compute $S_\\Delta$ around a closed loop in the space of framed quadratic differentials crossing a ran","tokens_in":43892,"feed_emoji":"🕸️","tokens_out":7522,"duration_ms":79954,"temperature":0.7,"pith_summary":"This paper proves the Kontsevich–Soibelman wall-crossing formula entirely through quadratic differentials and their spectral networks, with no input from Donaldson–Thomas theory. It shows that the product of BPS automorphisms over the active directions in a sector is covariantly constant as the quadratic differential moves through moduli space, as long as the sector's boundary rays never hit a saddle connection. The proof handles infinite-area surfaces, including spiral domains where trajectories accumulate, by working in a profinite completion truncated by central charge. An intermediate result extends the path-lifting rule to A0-laminations and shows their lifts generate the hat-homology lattice, which is what makes the BPS automorphism act on the right completed module.","feed_headline":"Spectral networks alone prove the wall-crossing formula","feed_subtitle":"A self-contained proof from quadratic differentials, including the spiral domains where prior methods fail.","key_machinery":"The completed signed path groupoid $\\widehat{\\Pi}_\\Delta$ and its homological shadow $\\widehat{H}_\\Delta$: formal linear combinations of path (or homology) classes truncated by the absolute value of the central charge $Z$, which makes infinite detour sums converge even when trajectories are dense. The path-lifting rule $F(\\wp,\\theta)$ adds to the trivial lift of $\\wp$ all elementary detours that follow rays of the spectral network $W_\\theta$ to a zero, loop around it, and return on the other sheet; the extension to A0-laminations with a preferred-sheet rule yields enough elements to approximate every hat-homology class. The wall-crossing automorphism $K_\\theta$ acts on $\\widehat{H}_\\Delta$ b","core_discovery":"Theorem 1.1: for an infinite-area half-translation surface $(X,q)$ whose singularity signature has no simple poles, the BPS automorphism $S_\\Delta$ is covariantly constant along any path of homology-framed quadratic differentials whose boundary rays stay non-active. The proof constructs a path-lifting rule $F(\\wp,\\theta)$ that assigns to each path (or A0-lamination) a formal sum of lifts of all possible detours along the spectral network, valued in a completed signed path groupoid. At a rank-one active direction $\\theta_0$ the one-sided limits satisfy $F^+(\\wp,\\theta_0)=K_{\\theta_0}F^-(\\wp,\\theta_0)$, where $K_{\\theta_0}$ is the BPS automorphism built from saddle connections and ring-domain","pith_inferences":["The same completed-groupoid machinery could produce an abelianization of $SL(2,\\mathbb{C})$-local systems along spectral networks with dense trajectories, as the authors plan; if successful, this would supply a missing link between non-abelianization and Joyce structures.","A formal, computer-checked enumeration of the admissible detours in Lemmas 6.5–6.7 would remove the main residual doubt in the proof and could be adapted to higher-rank spectral networks, where junctions create trajectories born at interior points and new infinite families.","The paper's exclusion of simple poles looks technical rather than structural; if the approximation statement of Proposition 5.11 can be extended to simple poles, the same argument would give the formula in finite area, where the non-active-boundary condition is very restrictive.","The proof of Proposition 6.10 by induction on Cantor–Bendixon rank suggests that wall-crossing could be organized purely by the accumulation structure of active directions, independent of the specific dynamics."],"forward_implications":["The Kontsevich–Soibelman formula is a theorem about quadratic differentials alone, not a corollary of motivic Donaldson–Thomas theory (in the infinite-area case).","The path-lifting functions $F(\\wp,\\theta)$ are well-defined and homotopy-invariant even when the spectral network has spiral domains, resolving convergence questions left open in earlier treatments.","The extension to A0-laminations shows that lifts of laminations are topologically dense in the hat-homology lattice, with supports controlled inside a fixed cone translate.","The one-sided limits $F^\\pm(\\wp,\\theta_0)$ are genuine limits in the Z-adic topology, and the wall-crossing relation holds for every rank-one direction, for both paths and laminations.","In finite area the theorem is conditional on an approximation statement; under unique ergodicity the non-active-boundary condition forces the path to be a Teichmüller geodesic ray, making the statement nearly void."],"supporting_citations":[{"why":"Supplies the spectral-network path-lifting-with-detours idea and the wall-crossing picture that the paper formalizes and proves convergent.","marker":"[GMN13b]"},{"why":"Original statement of the wall-crossing formula and the profinite/twisted-torus setting used for the BPS automorphisms.","marker":"[KS08]"},{"why":"Gives quadratic differentials as stability conditions, the hat-homology lattice with its triangulation basis, and the classification of rank-one directions in infinite area.","marker":"[BS15]"},{"why":"Proposes the extension of path lifting to A0-laminations to generate lattices; Section 5 follows this proposal.","marker":"[GMN13a]"},{"why":"Defines A0-laminations and their triangulation coordinates on ciliated surfaces, used in the approximation argument.","marker":"[FG07]"},{"why":"Prior wall-crossing proof via Fock–Goncharov coordinates, which covers only the B2-locus and relies on Donaldson–Thomas theory elsewhere; the paper aims to supersede this.","marker":"[All23]"},{"why":"Provides the finite-area identification with CY3 stability conditions and the rank-one classification including simple poles.","marker":"[Hai24]"},{"why":"Supplies the twisted-torus and completed Poisson-algebra conventions the paper adopts for the BPS automorphism.","marker":"[Bri19]"},{"why":"Gives the canonical lift of homology classes and the sign rule behind the winding ideal and the lift formulas.","marker":"[Joh80]"}],"fun_headline_variants":["Spectral networks alone prove wall-crossing","Wall-crossing via spectral networks, no DT theory","Path-lifting rules yield wall-crossing formula","Spectral networks settle wall-crossing formula"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The local case analysis of all possible ±-admissible detour paths near a rank-one saddle connection is complete and sign-correct; it is verified by inspecting finitely many figures rather than by a systematic algebraic enumeration, and a missing configuration or sign error would break the wall-crossing identity $F^+=K_\\theta F^-$.","fun_headline_variants_meta":{"raw":{"variants":["Spectral networks alone prove wall-crossing","Wall-crossing via spectral networks, no DT theory","Path-lifting rules yield wall-crossing formula","Spectral networks settle wall-crossing formula"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1332,"prompt_tokens":657,"completion_tokens":675,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":401,"completion_tokens_details":{"reasoning_tokens":617}},"tokens_in":401,"tokens_out":675,"duration_ms":6951,"temperature":1.0,"reasoning_tokens":617,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T21:54:08.312303+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Work out a concrete rank-one configuration (e.g. a saddle connection between two simple zeros on a genus-two surface) and explicitly enumerate all ±-admissible rays and their intersection numbers with the BPS cycle $L(\\hat\\gamma_0)$, either by hand or by computer; if any admissible detour family is missing from Figures 11–14 or any intersection number in Lemmas 6.5–6.7 is wrong, then the identity $F^+(\\wp,\\theta_0)=K_{\\theta_0}F^-(\\wp,\\theta_0)$ fails for a single-crossing path. Alternatively, compute $S_\\Delta$ around a closed loop in the space of framed quadratic differentials crossing a ran","supporting_citations":[],"review_version":1}