{"id":"66cfdfbc-3bbf-4c94-af14-3e8a2a11b47c","arxiv_id":"2502.06754","paper_version":4,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Conditioning two cable-graph points to lie in the same Brownian loop-soup cluster adds an odd-numbered Poisson cloud of Brownian excursions between them, yielding an exact law for the conditional cluster and its GFF analogue.","lead":"A new theorem says conditioning two points in a cable-graph Brownian loop-soup to be connected is the same, in law, as adding an odd number of Brownian excursions between them. This gives an exact conditional description of Gaussian free field sign-clusters and a simple incipient infinite cluster for all dimensions d >= 3.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section 3's Dynkin-style Laplace transform appears to drop a factor of 2 in the Gaussian shift; if so, the second proof of Theorem 2 is invalid as written.","rationale":"The reader's verdict identifies the trace/Markov-chain approximation in Section 2 as the weakest assumption and accepts the paper partly because the Section 3 Laplace-transform proof is independent. My stress test finds a more specific and potentially more serious issue in that independent proof: the Gaussian shift calculation appears to be off by a factor of 2 and by a factor of 2 in the killing rate, unless the GFF covariance is G/2, which is not the paper's convention. If this is correct, the second proof of Theorem 2 does not establish the claimed identification with the Poisson excursion process, and the first proof's sketched convergence in Lemma 8 becomes the only route, leaving the central theorem not fully proven as written. The theorem may well be true, and the issue may be a normalization typo, but it is load-bearing: without a corrected derivation, the paper's central claim is not rigorously supported by the two independent proofs claimed. A concrete computation on the one-edge cable graph will settle whether the factor is present. Unless the test shows consistency, the verdict should be CONDITIONAL pending a corrected or clarified derivation.","tokens_in":35726,"tokens_out":39078,"duration_ms":335587,"concrete_test":"Re-derive Eq. (3) of Section 3 for the one-edge cable graph [0,1] with x=0, y=1. Compute both sides explicitly: the Gaussian-shift Laplace transform of (Γ0+Φ1+Φ2)^2 using Cov(Γ0)=G, and the Laplace transform of the PPP of excursions with intensity abν_{0,1} using the explicit excursion measure on an interval. Check whether the coefficient of ∫∫G^k k k Φ1Φ2 in the cross term is 1 or 2, and whether the Green's function appearing is G^k or G^{2k}. Independently, verify that the constant in Corollary 12 is C=1 under the paper's stated normalization; a factor-of-2 error would change C or the mean of the Poisson random variable.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Section 3, the paper expands (Γ0+Φ)^2 and rewrites E[exp(-∫k(Γ0+Φ)^2)] as E_k(Γ0^2) × exp(-∫Φ^2 k) × E[exp(-2∫Γ^k Φ k)], where Γ^k is the GFF reweighted by exp(-∫Γ0^2 k). For a Gaussian Γ^k with covariance G^k, the last factor equals exp(2∫∫G^k k k ΦΦ), not exp(∫∫G^k k k ΦΦ) as written. More importantly, the reweighted field has covariance (G^{-1}+2k)^{-1}, i.e. the Green's function with killing rate 2k, not k, under the paper's own convention that the GFF covariance is G. The formula in the paper is only correct if the GFF covariance is G/2 and the killing rate is k, which contradicts the stated normalization (Λ = Γ^2 with Cov(Γ)=G). Consequently the identification of the cross term (3) with the Laplace transform of the Poisson process of excursions joining x1 and x2, and hence the parity lemma and switching property derived in Section 3, rests on a normalization identity that is not established and appears false as written. Since the Section 2 proof relies on the sketched discrete approximation in Lemma 8, the central theorem currently lacks a fully rigorous written proof unless this factor is reconciled.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper states and proves a \"switching identity\" for critical Brownian loop-soups and the Gaussian free field on cable graphs. The main statement, Theorem 2, says that conditionally on two points x and y being in the same loop-soup cluster and on the occupation times at x and y taking values a^2 and b^2, the occupation field has the law of the sum of four independent inputs: a loop-soup avoiding x and y, Poisson processes of excursions away from x and away from y with intensities proportional to a^2 and b^2, and a Poisson process of excursions joining x and y whose total number is conditioned to be odd. The paper gives three routes to this result: a discrete Markov-chain parity lemma, a Laplace-transform/Dynkin computation, and a random-current-style switching along circuits. It then derives consequences: a boundary-point version (Theorem 1), an incipient-infinite-cluster measure (Theorem 3), interlacement versions, parity identities for windings of loop-soup clusters, and multiple-point estimates for loop-soup percolation.","tokens_in":35956,"tokens_out":36560,"duration_ms":338533,"significance":"If Theorem 2 is correct, it is a genuinely striking and sharp description: conditioning on a connection is exactly an odd-path insertion into an unconditioned configuration. The statement is clean, falsifiable, and has direct consequences that are not accessible by prior methods, including a simple construction of the IIC measure in all d >= 3 and new parity identities for loop-soup clusters. The paper also correctly connects the result to the rewiring property of [66], to random-current switching in the Ising model, and to Pitman--Yor decompositions of Bessel bridges. The multiple proof strategies are a strength, and the consequences in Sections 5--6 are appropriately stated as applications. However, as written, the Laplace-transform proof in Section 3 contains a load-bearing Gaussian computation error, and the first proof in Section 2 relies on a parity lemma whose convergence argument is only sketched. The central claim is likely correct, but the manuscript needs a corrected and/or completed proof before it can be accepted.","major_comments":[{"comment":"For a centered Gaussian field Gamma^k with covariance G^k, E[exp(-2 int Gamma^k(x) k(x) Phi(x) dx)] equals exp(2 int int G^k(x,y) k(x) k(y) Phi(x) Phi(y) dx dy), not exp(int int G^k kk Phi Phi) as written. Moreover, under the paper's own normalization (Cov(Gamma)=G and Lambda=Gamma^2), the field obtained by reweighting with exp(-int Gamma_0^2 k) has covariance (G^{-1}+2k)^{-1}, which is the Green function with killing rate 2k, not k, in the usual Brownian convention. The identification of the cross term (3) with the Laplace transform of the Poisson process of excursions joining x1 and x2 relies on this computation, so the Section 3 proof of the parity lemma and of the switching property is not valid as written. Please correct the factor and state the normalization of G^k explicitly, or restructure Section 3 so that the claimed identity is derived from a correct Gaussian calculation.","section":"Section 3, displayed formula after \"By inspecting the variance...\""},{"comment":"The passage from Formula (2) to the claimed even-Poisson law is not justified. Substituting A ~ a^2 K, B ~ b^2 K and p(x,y)=alpha/K in (2) gives, for a jump count 2t, weights proportional to (alpha a b)^{2t}/t! (up to factors independent of t), whereas a Poisson variable conditioned to be even has weights proportional to mu^{2t}/(2t)!. The displayed asymptotics therefore do not identify the limiting conditional distribution. The bracketing argument with P_1 <= N <= P_2 also does not by itself prove the conditional law. Since Lemma 6 and hence the first proof of Theorem 2 depend on this step, the Section 2 route needs a complete proof or a precise reference to a result that contains this convergence.","section":"Section 2, Lemma 8 and its proof"}],"minor_comments":[{"comment":"In the displayed computation of E[1_{x1 connected to x2} exp(-int Gamma^2 k) | ...], the denominator e^m - e^{-m} should be e^m + e^{-m}; as printed, the equality to sinh(m(k))/cosh(m) is algebraically false.","section":"Section 3, proof of the switching property"},{"comment":"The notation 'Gamma(partial_1)' and 'Gamma(partial_n)' should be Gamma(x_1) and Gamma(x_2), and the comparison of Gaussian densities should be at (a_1,a_2) and (a_1,-a_2), not (a_1,-a_1).","section":"Section 3, proof of the parity lemma"},{"comment":"Several consequences, including the explicit bijection via 'peeling' and the proof of Lupu's intensity doubling conjecture, are deferred to papers listed as in preparation ([46], [68], [69], [13]). The reader would benefit from a sentence making explicit which statements are conditional on those forthcoming works.","section":"Sections 4.4 and 5"},{"comment":"There are several typographical slips: 'loose their full independence' should be 'lose their full independence'; in Section 3, 'm(k)' is used before its definition in the same sentence; and the proof of Lemma 6 refers to 'x' and 'y' without restating their role.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The theorem is very likely correct and important, and the paper is well within the scope of math.PR. My recommendation is driven by the unresolved status of the written proofs: the Section 3 computation has a clear factor-2 error, and the Section 2 parity lemma is only sketched at a load-bearing point. Both issues appear fixable within the manuscript's scope, so I do not recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague — this is the real thing, mostly. The general switching identity for cable-graph loop-soups (Theorem 2) is genuinely new; earlier work had special cases, but the statement that conditioning on a connection is exactly an odd Poisson insertion of excursions is a substantial and satisfying unification. The consequences are strong and clean: the IIC description for all d≥3, including d=6, and the random-even-subgraph/parity corollaries. The paper is honest about what is new and what was already known, and the citations to Pitman–Yor, [35], and [1] are fair.\n\nThe Section 2 route via the parity lemma and the rewiring property is credible, and the discrete Markov-chain approximation in Lemma 8 is plausible, though the K→∞ limit is only sketched. That is a soft spot, but not a fatal one.\n\nThe bigger soft spot is Section 3. The Dynkin-style computation looks to me like it drops a factor of two. Reweighting Γ0 by exp(-∫Γ0²k) should move the covariance to (G^{-1}+2k)^{-1}, not to (G^{-1}+k)^{-1}. As printed, the cross term (3) has exp(∫∫G^k k k ΦΦ) where the correct object would have a 2 and G^{2k}. If that is right, the Section 3 proof as written does not establish the parity lemma, and the claim of two independent proofs should be downgraded to one and a half. This is the kind of normalization slip that is easy to fix, and the fact that the same identity is obtained in Section 2 and via [66]/[44] makes me confident the theorem itself is true.\n\nThe paper also contains a clearly labeled heuristic section (4.4) about an explicit bijection; the author does not pretend it is a proof, and it should not count as one.\n\nBottom line: this is a significant paper with a real new theorem and useful consequences. It deserves a serious referee. The referee should check the Section 3 normalization carefully and ask for the Lemma 8 convergence details to be written up. I would accept it for peer review and would likely cite it once the normalization is cleaned up.","headline":"Genuinely new switching identity with strong consequences; Section 3 has a likely factor-of-two normalization slip, but the theorem is probably true and deserves serious review.","tokens_in":645,"tokens_out":1647,"would_cite":true,"duration_ms":170403,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J65","60K35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Conditioning two points on a cable graph to belong to the same Brownian loop-soup cluster is the same, in law, as adding an odd number of independent Brownian excursions to an unconditioned loop-soup.","keywords":["Brownian loop-soup","cable graph","Gaussian free field","switching identity","percolation","incipient infinite cluster","random even subgraphs","Markov loops"],"falsifier":"On a small cable graph such as a single edge with endpoints $x$ and $y$, compute the conditional Laplace transform of the occupation-time field given $x\\leftrightarrow y$, $\\Lambda(x)=a^2$ and $\\Lambda(y)=b^2$; the switching identity requires it to factor as the product of the Laplace transforms of an unconditioned loop-soup in the interior, two Poisson excursion processes away from the endpoints, and an odd-conditioned Poisson process of joining excursions. If the odd-conditioned factor does not appear with the exact intensity $ab\\nu_{x,y}$, or if the four factors are not independent, the identity is false; the one-dimensional single-edge case corresponds to a classical Bessel bridge decomposition, so a discrepancy there would falsify the general theorem.","tokens_in":35484,"feed_emoji":"🌀","tokens_out":11648,"duration_ms":84776,"temperature":0.7,"pith_summary":"The paper establishes a switching identity for critical Brownian loop-soups on cable graphs, equivalently for the square of the Gaussian free field. The claim is that conditioning two points x and y to lie in the same loop cluster is, in law, the same as taking an unconditioned loop-soup and overlaying an odd number of independent Brownian excursions joining x and y. The identity holds at the level of occupation-time fields, not just cluster topology, and extends to the boundary case where the two points are on the boundary of the same cluster, where a single excursion suffices. A reader should care because it converts a difficult conditional object into a sum of simple independent pieces, giving explicit laws for incipient infinite clusters, simplifications of recent results, and new tools for multi-point and high-dimensional questions.","feed_headline":"Odd path insertion reproduces conditioned loop-soup law","feed_subtitle":"The law of the conditioned cluster is an odd number of Brownian excursions over an unconditioned soup.","key_machinery":"The central object is the Brownian loop-soup, a Poisson point process of unrooted Brownian loops on a cable graph (a metric graph whose edges are segments), whose occupation-time field has the same law as the square of the cable-graph Gaussian free field. The load-bearing mechanism is the parity lemma: conditionally on $\\Lambda(x)=a^2$ and $\\Lambda(y)=b^2$, the loops meeting $\\{x,y\\}$ decompose into excursions, and the number of excursions joining $x$ to $y$ is a Poisson random variable conditioned to be even. Comparing this even description with the unconditioned Poisson description, and reweighting by the event that the two signs of the GFF agree, gives the switching: conditioning on $x\\leftrightarrow y$ makes the number of joining excursions odd. The paper supplies three proofs of this mechanism: one via discrete Markov-chain approximations, one via Laplace transforms in the spirit of Dynkin's isomorphism, and one via random even subgraphs in the Ising random current representation.","core_discovery":"Theorem 2 states that, conditionally on $x \\leftrightarrow y$, $\\Lambda(x)=a^2$ and $\\Lambda(y)=b^2$, the critical loop-soup occupation time $\\Lambda=\\Gamma^2$ has the same law as the sum of the occupation times of four independent inputs: an unconditioned critical loop-soup in $G\\setminus\\{x,y\\}$; a Poisson process of excursions away from $x$ with intensity $a^2$ times the excursion measure; a Poisson process of excursions away from $y$ with intensity $b^2$ times the excursion measure; and a Poisson process of excursions joining $x$ and $y$ with intensity $ab$ times the excursion measure, conditioned so that the number of joining excursions is odd. The boundary version (Theorem 1) says that conditioning two points to be on the boundary of the same cluster is equivalent to overlaying one independent Brownian excursion between them. As a consequence, the incipient infinite cluster measure in $\\mathbb{Z}^d$ for $d\\ge 3$ exists and is described by an unconditioned loop-soup reweighed by the square root of its local time at the origin plus one independent Brownian excursion from the origin to infinity (Theorem 3).","pith_inferences":["If the switching identity survives to the continuum scaling limit, then in dimensions $d=3,4,5$ the continuum loop-soup clusters would inherit an even simpler conditional law: conditioning two points to be connected would amount to adding a single Brownian excursion, as in the two-dimensional results cited in the paper.","The explicit overlay description of the incipient infinite cluster suggests that spectral properties of the cluster (e.g., Alexander–Orbach type exponents) could be studied by analysing a Brownian excursion in the random environment created by an unconditioned loop-soup.","A finite-graph Monte Carlo check could test the parity mechanism directly: sample loop-soups conditioned on $\\Lambda(x)=a^2$, $\\Lambda(y)=b^2$ and $x\\leftrightarrow y$, and count the number of independent excursion bridges joining $x$ and $y$; the switching identity predicts a Poisson law with mean $ab$ times the mass of $\\nu_{x,y}$, conditioned to be odd."],"forward_implications":["The incipient infinite cluster of the loop-soup in $\\mathbb{Z}^d$ for $d\\ge 3$ exists and has the explicit description of a loop-soup reweighed by the square root of its local time at the origin, plus one independent Brownian excursion from the origin to infinity.","Conditioned versions of the incipient infinite cluster have the same form: fixing a boundary point, or fixing the occupation time at the origin, yields the overlay of an unconditioned or conditioned loop-soup with one excursion to infinity.","For a loop-soup superposed on a Brownian interlacement, conditioning the origin to be connected to infinity gives an odd number of Brownian excursions from the origin to infinity, with the interlacement conditioned to avoid the origin.","Conditionally on the occupation-time field, the parity of loop crossings forms a uniform random even subgraph, which yields independent fair coins for the winding parity of clusters and proves the intensity doubling conjecture for loop-soups in dimensions $d\\ge 7$.","The switching property gives new upper and lower bounds for multi-point connection probabilities and for the size of the largest clusters in dimensions $d=3,4,5$, complementing recent results obtained by renormalization arguments."],"supporting_citations":[{"why":"It establishes that the occupation-time field of the Brownian loop-soup has the same law as the square of the Gaussian free field.","marker":"[36]"},{"why":"It identifies loop-soup clusters with sign-clusters of the cable-graph GFF and gives the two-point connection probability used throughout.","marker":"[39]"},{"why":"It supplies the spatial Markov/rewiring property and the parity result (Proposition 7) on which the parity lemma is based.","marker":"[66]"},{"why":"It provides the discrete loop-soup jump-count formula (Proposition 2.46) and the background used in the Markov-chain approximation proof of Lemma 8.","marker":"[70]"},{"why":"It relates loop-soups to Ising random currents and random even subgraphs, underlying Proposition 4 and the third proof of the switching identity.","marker":"[44]"},{"why":"It gives the one-edge Bessel bridge decomposition that supplies the switching identity along a single edge, used in Section 4.","marker":"[51]"},{"why":"It provides the recent existence result for the incipient infinite cluster that the switching property reproves, extends to d=6, and describes explicitly.","marker":"[12]"},{"why":"It develops the general Markov loop-soup framework in which occupation fields and excursion decompositions are expressed.","marker":"[37]"}],"fun_headline_variants":["Odd excursions replace conditioned loop-soup law","Conditioned loop-soup: odd Brownian excursions added","Switching identity: odd paths join loop-soup clusters","Odd path insertion yields conditioned loop-soup law","Law of conditioned clusters: soup plus odd excursions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The most exposed premise is that the fine discrete approximations of the cable-graph loop-soup converge to the continuous process in the way needed for the parity lemma to pass to the limit; the Laplace-transform proof in Section 3 provides an independent route that does not rely on that approximation.","fun_headline_variants_meta":{"raw":{"variants":["Odd excursions replace conditioned loop-soup law","Conditioned loop-soup: odd Brownian excursions added","Switching identity: odd paths join loop-soup clusters","Odd path insertion yields conditioned loop-soup law","Law of conditioned clusters: soup plus odd excursions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000194,"raw_usage":{"total_tokens":1336,"prompt_tokens":912,"completion_tokens":424,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":350}},"tokens_in":528,"tokens_out":424,"duration_ms":3930,"temperature":1.0,"reasoning_tokens":350,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T14:25:34.034002+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On a small cable graph such as a single edge with endpoints $x$ and $y$, compute the conditional Laplace transform of the occupation-time field given $x\\leftrightarrow y$, $\\Lambda(x)=a^2$ and $\\Lambda(y)=b^2$; the switching identity requires it to factor as the product of the Laplace transforms of an unconditioned loop-soup in the interior, two Poisson excursion processes away from the endpoints, and an odd-conditioned Poisson process of joining excursions. If the odd-conditioned factor does not appear with the exact intensity $ab\\nu_{x,y}$, or if the four factors are not independent, the identity is false; the one-dimensional single-edge case corresponds to a classical Bessel bridge decomposition, so a discrepancy there would falsify the general theorem.","supporting_citations":[{"cited_title":"Markov loops and renormalization","cited_arxiv_id":null,"evidence_quote":"It establishes that the occupation-time field of the Brownian loop-soup has the same law as the square of the Gaussian free field."},{"cited_title":"From loop clusters and random interlacements to the free field","cited_arxiv_id":null,"evidence_quote":"It identifies loop-soup clusters with sign-clusters of the cable-graph GFF and gives the two-point connection probability used throughout."},{"cited_title":"On the spatial Markov property of soups of unoriented and oriented loops","cited_arxiv_id":null,"evidence_quote":"It supplies the spatial Markov/rewiring property and the parity result (Proposition 7) on which the parity lemma is based."},{"cited_title":"Lecture notes on the Gaussian free field , volume 28 of Cours Sp´ ecialis´ es","cited_arxiv_id":null,"evidence_quote":"It provides the discrete loop-soup jump-count formula (Proposition 2.46) and the background used in the Markov-chain approximation proof of Lemma 8."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It relates loop-soups to Ising random currents and random even subgraphs, underlying Proposition 4 and the third proof of the switching identity."},{"cited_title":"A decomposition of Bessel Bridges","cited_arxiv_id":null,"evidence_quote":"It gives the one-edge Bessel bridge decomposition that supplies the switching identity along a single edge, used in Section 4."},{"cited_title":"Incipient infinite clusters and volume growth for Gaussian free fields and loop soups on metric graphs","cited_arxiv_id":"2412.05709","evidence_quote":"It provides the recent existence result for the incipient infinite cluster that the switching property reproves, extends to d=6, and describes explicitly."},{"cited_title":"Markov paths, loops and fields , L.N","cited_arxiv_id":null,"evidence_quote":"It develops the general Markov loop-soup framework in which occupation fields and excursion decompositions are expressed."}],"review_version":1}