{"id":"842fae2a-0973-4d4b-90a1-32004e593eb7","arxiv_id":"2607.18223","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Tree-level massive de Sitter correlators are constructed by gluing Lauricella-type vertex functions according to graph combinatorics, and the hypergeometric content collapses to rational functions once the dynamical propagators are removed.","lead":"This paper shows that all tree-level massive cosmological correlators can be built from universal hypergeometric building blocks — Lauricella vertex functions — glued together by a graph-combinatorial residue algorithm, and that stripping away the dynamical propagators collapses the remaining hypergeometric structure to rational functions. It delivers explicit series formulas for arbitrary tree graphs (chains and stars) and a family of new identities among Lauricella function","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The spectral gluing algorithm's pole-cancellation step is proven only for 2F1 vertex functions (Sec. 4.1); for multivariable Lauricella F_C^{(n)} the needed analytic continuation and reflection identities are asserted, so the universal series may be incomplete.","rationale":"The reader's weakest assumption identifies the general analytic-continuation and pole-cancellation machinery. I agree that this is the critical region, but I single out the specific pole-cancellation of the \\(\\pm\\pm\\) spectral integrals as the most load-bearing step: the algorithm's universal applicability to arbitrary tree graphs depends on it, and the paper only demonstrates it for \\(F_C^{(1)}=2F_1\\) through the connection formula (A.11). The multivariable case is asserted without proof, and the supporting reflection identity (Lemma A.1) is new and only formally derived. This is a fixable gap rather than a proven failure — the two-site chain (where only \\(F_C^{(1)}\\) appears) matches the literature, and the two-massive-leg vertex numerics pass in the tested regions. However, without an explicit proof or numerical verification for \\(n\\ge2\\), the central claim of a universal spectral gluing algorithm remains conditional. My concrete test directly checks the cancellation for a three-site chain with a degree-2 vertex in the region where the naive contour argument would break, which would settle whether the gap is real. Since the reader already rendered a CONDITIONAL verdict and my concern is a specific instance of the same gap, the verdict should remain unchanged.","tokens_in":68418,"tokens_out":6186,"duration_ms":54323,"concrete_test":"Evaluate the three-site chain analytic piece \\(\\hat I^P_{+++}\\) (e.g., Eq. (B.19) for the region \\(u_{12}>u_{21}, u_{32}>u_{23}\\)) in a kinematic configuration where \\(u_{12}u_{21} > 4\\) (e.g., \\(X_1=1, X_2=0.6, Y_{12}=1.2\\) so that \\(u_{21}=2, u_{12}=1.2\\)) and compare against direct numerical integration of the double spectral integral (B.17) using the Cauchy principal-value spectral density and vertex functions evaluated by high-precision quadrature (without series regularization). If the gluing series and the numerical integral disagree beyond truncation error, the pole cancellation fails for \\(n=2\\). Repeat with a three-massive-leg internal vertex to test \\(F_C^{(3)}\\).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Sec. 4.1 evaluates the gluing of two vertex functions by decomposing into four frequency sectors. The paper asserts that the equal-sign contributions \\(\\hat I^P_{++}\\) and \\(\\hat I^P_{--}\\) vanish, either via 'apparent poles' (Lemma 4.1) or via analytic continuation when \\(u_{ij}u_{ji}>4\\). For the latter, the text explicitly states: 'For the case of \\(F_C^{(1)}=2F_1\\) one can see this by using a connection formula (A.11)' — no proof is provided for general \\(F_C^{(n)}\\) with \\(n\\ge 2\\). The maximally-nested algorithm of Sec. 4.2 (off-shell ansatz, path-matrix on-shell solution, Eqs. (4.13)–(4.18)) and hence the universal claim for arbitrary tree graphs relies on this cancellation to reduce the spectral integral to the mixed ±∓ residues. If the cancellation fails for a vertex with two or more off-shell legs, the resulting series representation is not the true correlator. Furthermore, the Lauricella reflection formula (Lemma A.1) used to prove Lemma 4.1 is proven only at the level of formal series manipulations, without specifying the region of \\(z\\) where the limit is uniform or addressing boundary terms in analytic continuation. These gaps mean the algorithmic output is not established for generic graphs with internal vertices of degree ≥ 2.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a spectral-gluing framework for tree-level massive cosmological correlators. It first defines universal one-time-integral \"vertex functions\" and evaluates them as type-C Lauricella functions F_C^(n) (Sec. 3, Eqs. (3.40)-(3.41)). It then proposes a combinatorial algorithm (Sec. 4) that evaluates the spectral integrals of nested propagators by collecting residues fixed by the incidence matrix Q and path matrix P=Q^{-1} of a rooted tree (Eqs. (4.13)-(4.18)). Explicit series are given for two-site, three-site, N-site chain and N-site star graphs (Sec. 5). Acting with graph annihilators is shown to remove propagator factors, leading to \"magical identities\" that reduce products/sums of Lauricella functions to rational functions (Sec. 6, Eq. (6.28)). The paper claims that the hypergeometric complexity of massive correlators is governed entirely by graph topology.","tokens_in":68680,"tokens_out":13577,"duration_ms":124921,"significance":"If the main results hold, this is a substantial step: it provides a systematic, graph-topological construction of tree-level massive correlators, with explicit partially resummed series, and it uncovers a new family of identities connecting multivariable Lauricella functions to rational functions. The paper is commendably concrete: it includes numerical cross-checks, recovers the known two-site result of Ref. [46], credits Ref. [66] for the two-site magical identity, and argues for uniform transcendental weight. The construction is parameter-free in the sense that the series coefficients are fixed by graph incidence data, not fitted. However, the advertised universality currently outruns the proof: the central pole-cancellation step is justified only for F_C^(1)=2F1, the analytic continuation for multivariable Lauricella functions is incomplete, and marginal vertices are explicitly excluded. These gaps prevent the paper from fully establishing the claim that all tree-level massive correlators are governed by the asserted rigid structure.","major_comments":[{"comment":"The reduction of the gluing integral to the mixed ±∓ residues depends on the assertion that bI^P_{++}=bI^P_{--}=0 for arbitrary vertex functions. The text states that for u_{ij}u_{ji}>4 one can see this using a connection formula (A.11), but adds explicitly: 'For the case of F_C^{(1)}=2F1'. In generic trees, internal vertices have degree ≥2, so the relevant vertex functions are F_C^{(n)} with n≥2 (e.g. Appell F4 for a two-massive-leg vertex). No analytic continuation of these multivariable Lauricella functions in the required kinematic region is supplied. Since the off-shell ansatz (4.13) and the path-matrix on-shell condition (4.17)-(4.18) depend on this cancellation, the algorithmic output is not established to equal the true correlator for generic graphs. This is a load-bearing gap in the universal claim.","section":"Sec. 4.1, Eqs. (4.5)-(4.10)"},{"comment":"The title and abstract claim 'all tree-level massive cosmological correlators', but marginal vertices (X=0) are not covered. Eq. (3.74) places the physical region for marginal vertices outside the domain in which type-C Lauricella analytic continuations are available, and the text explicitly says 'We will discuss this in a future work.' Marginal vertices occur in legitimate tree-level diagrams (e.g. any internal vertex with no external conformally-coupled leg). The precise class of graphs covered should be stated, and the headline claim should be softened accordingly, or the marginal-vertex continuation must be provided.","section":"Sec. 3.5, Eq. (3.74); Sec. 7"},{"comment":"The proof of sufficiency of the inequalities (3.63) is flawed. The exhibited rank-one construction does not generally satisfy 1^T G1≤1: the step 'the last step uses (3.63) for i=j*' is invalid when u_{j*} - sum_{i≠j*} u_i is negative with large magnitude. For example, u=(0.9,1,1) satisfies (3.63) but the proposed collinear configuration gives 1^T G1=(0.9-2)^2=1.21>1. The proposition may still be true (for instance by arranging non-collinear vectors so the resultant vanishes when no single vector dominates), but the proof as written needs repair. Since this proposition underlies the characterization of the Euclidean kinematic region, a correct proof should be supplied.","section":"Sec. 3.4, Proposition 1"},{"comment":"The analytic continuation into region (IV) of the two-massive-leg vertex relies on a Levin u-transform (3.34) applied to a formally divergent series. This is a numerical extrapolation technique, but no theorem or rigorous bound is given showing that the transformed sequence converges to the value of the original integral (3.19). The vertex functions are subsequently used as analytic building blocks in the spectral gluing algorithm, so this is not a merely cosmetic issue. A rigorous convergence statement, or a reformulation in which the integral definition is primary and the series is only an auxiliary representation, is needed.","section":"Sec. 3.2, Eq. (3.34) and Fig. 4"}],"minor_comments":[{"comment":"In the factorized reduction formula for p=1/2, the upper parameters of the two 2F1 factors appear to contain a spurious '+k'; since (2p-1)/2 = 0, the F4 sum collapses to k=0, so these '+k' should be absent. Please check.","section":"Sec. 3.2, Eq. (3.31)"},{"comment":"The central vertex of the N-site star is drawn with no external line, yet the formulas use a nonzero X0. Please clarify whether the computation assumes X0>0, and whether the X0=0 marginal case is deferred to Sec. 3.5.","section":"Sec. 5.4, Fig. 7(d)"},{"comment":"The Lauricella reflection formula is proven by formal manipulation of the defining series, without stating the domain of z or justifying the interchange of the limit c→1-m with the infinite sum. A statement of uniform convergence on compact subsets of the Lauricella convergence domain should be added.","section":"Appendix A.3, Lemma A.1"},{"comment":"The claimed absolute convergence of the on-shell series is justified only by the soft-limit ('geometric type series') heuristic. The full series contains Lauricella functions whose growth in the summation indices m_i is not controlled. A direct convergence argument or an explicit bound is needed.","section":"Sec. 4.2.3"},{"comment":"The heading 'Reflexion formula' should be 'Reflection formula'. Also, the denominator Γ[+iν,-iν] in Eq. (4.6) is notationally confusing because the measure factor Nν has already been displayed; clarifying the cancellation would help the reader.","section":"General notation"}],"recommendation":"major_revision","confidential_remarks":"This is a strong and potentially important paper, but the advertised universality is currently supported only under incomplete analytic-continuation assumptions. The key blocker is the pole-cancellation argument in Sec. 4.1, which is proven for 2F1 but asserted for general F_C^(n); this directly affects the validity of the algorithm for generic internal vertices. The marginal-vertex exclusion also makes the 'all tree-level' claim too broad. I see no grounds for rejection: the framework is coherent, the two-site checks are convincing, and the combinatorial structure is elegant. The requested revision should either complete the missing continuations or carefully delimit the class of graphs and kinematic regions for which the theorem is proven."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the spectral gluing algorithm is the real deal: the incidence/path-matrix construction in Sec. 4.2 is clean, the two-site result matches [46], and the numerics for vertex functions check out. Second, the 'all tree-level' claim is more than the paper proves. The pole-cancellation step for the equal-sign spectral integrals is demonstrated only for 2F1; for multivariable F_C^(n) it is asserted. That is a load-bearing step for internal vertices of degree ≥2, exactly the case the universal algorithm is designed for.\n\nWhat is genuinely new: the universal vertex functions as F_C^(n), the graph-combinatorial gluing algorithm, the closed-form N-site chain and star series, the reflection formula Lemma A.1 (claimed new, plausible), and the family of 'magical' identities. The paper is honest about its limits: marginal vertices are explicitly deferred (Sec. 3.5), and the analytic continuation of F4 into region (IV) relies on a Levin u-transform of a formally divergent series—a numerical regularisation, not a proof. I agree with the reader's conditional verdict; the gaps are fixable rather than fatal.\n\nThe softest spot is Sec. 4.1. The text says 'for the case of F_C^(1)=2F1 one can see this by using a connection formula (A.11)' and gives no argument for general F_C^(n). Lemma 4.1, which removes the apparent poles, rests on the reflection formula A.1 proven only at the level of formal series, with no control of uniformity or boundary terms. The general magical identities further depend on the differential system (6.8), whose derivation in App. C I could not check because it was not in the text I reviewed. These are not internal contradictions, but they mean the universal series representation is not established for generic graphs yet.\n\nWho should read it: anyone computing massive cosmological tree correlators, and people working on Lauricella-function identities. The combinatorial perspective is valuable even if some analytic details get tightened later.\n\nMy recommendation: send it to a serious referee. Ask the referee to focus on Sec. 4.1 and the continuation claims, and to request either a proof for F_C^(n) with n≥2 or an explicit conjecture with numerical checks on a graph with an internal degree-3 vertex. The paper deserves referee time; it just needs a revision that closes or clearly fences off the general-continuation gap.","headline":"Real algorithm, honest gaps: the spectral gluing construction is likely correct, but the universal claim outruns the analytic-continuation proof, which is demonstrated only for 2F1.","tokens_in":69305,"tokens_out":2972,"would_cite":true,"duration_ms":25190,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["33C65","33C70","81T20","83F05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that every tree-level massive cosmological correlator can be built from universal type-C Lauricella vertex functions glued by a spectral algorithm governed solely by the graph's incidence and path matrices, and that after","keywords":["cosmological correlators","de Sitter space","Lauricella functions","hypergeometric functions","spectral representation","tree-level graphs","graph combinatorics","magical identities"],"falsifier":"Compute, by direct numerical time integration, the two-massive-leg vertex V^{(2)}_{+,µ1,µ2}(u1,u2;p) deep inside region (IV), for example along u1=u2=0.8 with p=1.5, and compare against the paper's reduction-formula series: the claimed continuation must match the integral to numerical precision. Separately, evaluate the spectral gluing integral for two vertices in the regime u_ij u_ji > 4; the paper claims the equal-frequency pieces vanish after analytic continuation, so any nonzero residue series there would falsify the algorithm. For marginal vertices, numerically integrate W^{(3)}_{+,µ}(u;p","tokens_in":68156,"feed_emoji":"🌌","tokens_out":8653,"duration_ms":84267,"temperature":0.7,"pith_summary":"This paper sets out to prove that the hypergeometric complexity of massive cosmological correlators at tree level is not intrinsic: every tree graph decomposes into universal single-time-integral 'vertex functions' that are type-C Lauricella functions, and the full correlator is reconstructed by a spectral gluing algorithm whose only input is elementary graph data—the incidence matrix of the rooted tree and its inverse path matrix. The intended payoff is practical and conceptual: any tree-level correlator of scalar fields with generic masses receives an explicit convergent multiple-series representation in which internal energies are resummed, and these series solve the correlator's boundary differential equations as expansions in eigenfunctions. The central claim is that applying graph annihilators—differential operators that strip away the dynamical propagators—yields 'magical identities' in which the remaining hypergeometric kinematic dependence collapses to a rational function (Eq. 6.28), so the apparent transcendentality of massive correlators is kinematical rather than dynamical. A sympathetic reader would care because this turns a zoo of intractable multivariable special functions into graph combinatorics and suggests a geometric principle behind de Sitter observables.","feed_headline":"Every tree-level massive correlator is glued from Lauricella blocks","feed_subtitle":"A spectral gluing algorithm reads residues from the path matrix; stripping propagators gives rational functions.","key_machinery":"The central objects are the vertex functions V^{(n)}: single-time integrals over de Sitter time of the external plane wave and n massive Hankel mode functions. Their Mellin–Barnes evaluation expresses each as a sum over ±iμ_j frequency modes of the regularized type-C Lauricella function F_C^{(n)} (multivariable generalizations of the hypergeometric 2F1 function; Eqs. 3.40–3.41). These are the universal building blocks. The spectral propagator (2.14) represents the time-ordered bulk-to-bulk propagator as an integral over off-shell mass ν with density ρ_ν(μ)=1/(ν²–μ²)_{iϵ}, factorizing time integrals and making gluing a residue-collection problem. The combinatorial core is the pair (Q, P): the","core_discovery":"At the paper's core is the claim that every tree-level massive correlator is governed by one rigid structure. The authors define a vertex function V^{(n)}_{+,µ}(u;p) as a single time integral of n Hankel functions, show via Mellin–Barnes that it is a sum of 2^n frequency modes each given by a regularized type-C Lauricella function F_C^{(n)} (Eqs. 3.40–3.41). They then introduce a spectral representation of the time-ordered bulk-to-bulk propagator that factorizes time integrals and trades nesting for spectral integrals; the spectral gluing algorithm evaluates these by residues. The algebraic heart is the off-shell ansatz (4.13): for a rooted tree with incidence matrix Q, the on-shell conditio","pith_inferences":["If the magical identities are as universal as claimed, they imply a form of analytic bootstrap: the full hypergeometric dependence of a massive graph is fixed by its rational contraction at the total-energy pole, so one could try to prove the identities by matching discontinuities around the singular loci rather than by direct summation.","The incidence/path-matrix mechanism points to a natural loop-level extension: for graphs with cycles, the analogue of the path matrix would be a cycle-space or matroid object; the tree-level rooted-decoration construction is the first step toward a purely combinatorial loop integrand.","The marginal-vertex gap (Sec. 3.5) is the cleanest place to stress-test the framework: one could numerically integrate W^{(3)} in its physical region and compare to the F_C^{(2)} series; if the series diverges or mismatches, the algorithm needs a new continuation prescription before it can handle folded vertices such as those in the triple-exchange bispectrum."],"forward_implications":["For every tree graph with principal-series massive internal lines, the maximally-nested analytic contribution is written as a multiple series over m∈N^{V-1} with terms fixed by the path matrix; no nested time integral needs to be evaluated.","The same differential operators D_{ij} that define the graph's differential system also contract edges: applying all of them maps any tree-level massive correlator to a single rational function of the external energies, so the correlator satisfies an explicit hierarchy of differential identities.","Because the magical identities survive with Lauricella factors removed, the rational soft-limit expression fixes the full hypergeometric series; this gives an infinite family of non-trivial identities among generalized hypergeometric functions that hold for general twists and internal masses.","If the algorithm's convergence statements hold, changing the root vertex yields distinct but valid representations for disjoint kinematic regions, so any given correlator can be evaluated by patching finitely many root expansions."],"fun_headline_variants":["Spectral gluing: tree-level massive correlators from Lauricella blocks","Massive correlators: hidden rational simplicity via spectral gluing","All tree-level massive correlators glued from hypergeometric functions","Spectral gluing algorithm unifies massive correlators","Tree-level massive correlators: rational after stripping propagators"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The construction stands or falls on the assumption that the analytic-continuation and pole-cancellation machinery is globally valid in every kinematic region—specifically that the bI^P_{±±} spectral integrals vanish for general multivariable Lauricella functions (shown only for 2F1), that the regularization of the divergent series correctly continues the two-massive-leg vertex in region (IV), and that marginal vertices, for which the authors state no analytic continuation of","fun_headline_variants_meta":{"raw":{"variants":["Spectral gluing: tree-level massive correlators from Lauricella blocks","Massive correlators: hidden rational simplicity via spectral gluing","All tree-level massive correlators glued from hypergeometric functions","Spectral gluing algorithm unifies massive correlators","Tree-level massive correlators: rational after stripping propagators"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1212,"prompt_tokens":764,"completion_tokens":448,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":508,"completion_tokens_details":{"reasoning_tokens":376}},"tokens_in":508,"tokens_out":448,"duration_ms":4375,"temperature":1.0,"reasoning_tokens":376,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T15:38:33.824190+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, by direct numerical time integration, the two-massive-leg vertex V^{(2)}_{+,µ1,µ2}(u1,u2;p) deep inside region (IV), for example along u1=u2=0.8 with p=1.5, and compare against the paper's reduction-formula series: the claimed continuation must match the integral to numerical precision. Separately, evaluate the spectral gluing integral for two vertices in the regime u_ij u_ji > 4; the paper claims the equal-frequency pieces vanish after analytic continuation, so any nonzero residue series there would falsify the algorithm. For marginal vertices, numerically integrate W^{(3)}_{+,µ}(u;p","supporting_citations":[],"review_version":1}