{"id":"26ddf260-b7f1-44b7-bbdd-c569d70bbb07","arxiv_id":"2608.12225","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A pathwise change-of-variable formula on manifolds for paths with finite p-th variation, with the p-th variation tensor as correction term, extending Föllmer's calculus to arbitrary order and to curved spaces.","lead":"This paper builds a path-by-path calculus for smooth functions of very irregular paths on curved spaces, using only an affine connection and no probabilistic or rough-path enhancements. It matters because it gives an Itô-type formula for manifold-valued processes such as fractional Brownian motion, with a higher-order variation tensor as the correction term.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: Theorem 4.1 is internally consistent under the stated V^p hypothesis; the main soft spot is only that V^p is strong and explicitly acknowledged.","rationale":"Reader's verdict ACCEPT with moderate confidence. My stress-test agrees: I examined the proof chain Theorem 4.1, the p-th variation definition, the independence and transformation propositions, the Taylor formula, and the fBm example. I found no gap that would invalidate the central claim. The weakest assumption, as the reader notes, is the V^p condition: it requires both the uniform p-energy bound and the weak-* convergence of the specific atomic tensor measures along the chosen partition sequence. This is genuinely stronger than continuity and is not implied by finite p-variation over all partitions. But the theorem is explicitly conditional on V^p, and Section 6 gives nontrivial, self-contained examples satisfying it (in particular Proposition 6.1 for fBm). The proof of Proposition 6.1 is careful and correct: the covariance estimate uses the absolute summability of rho_H for H=1/p<1/2, hypercontractivity gives the fourth-moment bound, and Borel-Cantelli plus a density argument yields almost-sure weak-* convergence and finite p-energy. I also checked the covariant Taylor remainder: the uniformity in (x,v) follows from compactness of the unit sphere bundle and continuity of geodesic flow and parallel transport; the remainder is o(|v|^p) as needed. The cancellation in Theorem 4.1 is valid: the remainder sum vanishes by uniform continuity and finite p-energy, and the p-th term converges by weak-* convergence. I therefore do not find a load-bearing concern that would change the acceptance.","tokens_in":27238,"tokens_out":41238,"duration_ms":363068,"concrete_test":"Recompute the Schied-Zhang example (Section 6.3) using the projected ambient increment P_x(y-x) in place of the logarithmic increment exp^{-1}_x(y) on S^2. Verify that the resulting atomic 3rd-variation measures have the same weak-* limit (27/256) U_t^{⊗3} dt and that the compensated first- and second-order sums in Theorem 4.1 still converge to sin^3(w(t)) - (27/2048)∫(9cos(3w(s))-cos(w(s)))ds. Since Prop 2.5 guarantees independence of the increment map, a discrepancy here would reveal a flaw in the independence lemma or in the remainder estimate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the independence lemma (Prop 2.5), the transformation rule (Prop 2.11), the covariant Taylor remainder (Thm 3.2), and the convergence argument in Thm 4.1, I find no internal inconsistency in the central change-of-variable formula. The only substantive assumption is the definition of V^p(M,π): finite p-energy plus weak-* convergence of the atomic p-increment measures along the same partition sequence. This is indeed not implied by continuity or by ordinary finite p-variation, so the theorem is genuinely conditional. However, the paper states this condition explicitly, proves the p-th variation tensor is independent of the admissible increment map, and provides several constructions (fBm with H=1/p, the Schied-Zhang fractal, and their manifold lifts) for which V^p holds along the relevant partitions. I therefore do not regard the restrictiveness of V^p as a defect of the proof. The proof of Prop 6.1 (Hermite expansion, hypercontractivity, Borel-Cantelli, density argument) also appears sound. The partition dependence noted in Remark 4.5 is a scope restriction, not an error.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an intrinsic pathwise calculus for continuous paths on smooth manifolds with finite p-th order variation along a fixed partition sequence. It introduces admissible increment maps, defines the p-th variation tensor as a weak-* limit of atomic symmetric tensor measures, proves independence of the increment map and naturality under smooth maps, establishes a uniform covariant Taylor expansion, and derives a p-th order change-of-variable formula (Theorem 4.1) expressing f(X_T)-f(X_0) as the limit of covariant left Taylor sums of orders 1,...,p-1 plus a contraction of the p-th covariant derivative with the p-th variation tensor. It then interprets the formula in terms of higher-order tangent geometry (Pohl jets, connection splitting) and gives examples: Riemannian Brownian motion (p=2), exponential lifts of fractional Brownian motion with H=1/p, a Schied-Zhang fractal on the sphere, constant-curvature spaces, and fractional rotations on SO(3).","tokens_in":27447,"tokens_out":24750,"duration_ms":254333,"significance":"The central theorem is substantial: it gives a genuinely intrinsic higher-order analogue of Föllmer's pathwise Itô calculus on manifolds without rough-path enhancements. The main line of proof is complete: the covariant Taylor remainder is controlled uniformly by the finite p-energy, the highest-order term converges by the defining weak-* convergence of the variation tensor, and the remainder vanishes. The independence lemma (Prop 2.5), transformation rule (Prop 2.11), and the almost-sure variation computation for fractional Brownian motion (Prop 6.1, proved in Appendix A) are concrete and carefully argued. The paper is honest about the main hypothesis: membership in V^p(M,π) is a substantive condition on the path-partition pair and is not implied by continuity or ordinary finite p-variation; this is acknowledged and illustrated by nontrivial examples. I found no circularity: no constants are fitted and the p=2 case is checked against Föllmer's formula.","major_comments":[],"minor_comments":[{"comment":"The definition of V^p(M,π) leaves implicit the role of the admissible increment map; it should state explicitly that finiteness of p-energy and weak-* convergence are required for some admissible increment map, with Proposition 2.5 then ensuring independence of the choice.","section":"Section 2.2, Definition 2.3"},{"comment":"The line 'Expand eFA in the multivariate Hermite basis' is garbled; it should read 'Expand the mean-zero polynomial F̃_A(z)=F_A(z)-A(m_p) in the multivariate Hermite basis'.","section":"Appendix A, proof of Proposition 6.1"},{"comment":"The definition of I_t^∇ for arbitrary t in [0,T] should specify how Theorem 4.1 is applied on the subinterval [0,t]; the natural choice is the induced partition sequence obtained by inserting t, and one should note that the restriction of X to [0,t] belongs to V^p(M,π^t).","section":"Section 4.2, Proposition 4.6"},{"comment":"The proof that D^(p,∇)_{π,X}(f) depends only on the reduced p-jet is compressed into a single reference to Proposition 5.1; please spell out that the connection-dependent covariant derivatives are the components of the reduced p-jet under the splitting and that the pathwise limiting construction factors through the jet section.","section":"Section 5.4, Proposition 5.3"},{"comment":"The claim that the absolute cubic variation of the Schied-Zhang path is finite and non-zero should be supported by a precise citation or a short argument, since the quoted theorem states the signed variation and the finiteness of total variation is needed for the weak-* convergence in V^p.","section":"Section 6.3"}],"recommendation":"minor_revision","confidential_remarks":"No concerns about scope, novelty, or citation practice; the self-citations to Cont-Perkowski and related pathwise calculus literature are appropriate. I would be comfortable with acceptance once the minor clarifications above are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Rama Cont's paper gives the manifold analogue of the p-th variation calculus he and Perkowski built in Euclidean space. That is a genuinely missing piece: Föllmer gives Itô calculus pathwise in R^d, Schwartz–Émery give stochastic second-order geometry, and rough-path manifold constructions need enhancements. This paper supplies the intrinsic object—the p-th variation tensor as a symmetric tensor measure along the path—and proves a change-of-variable formula for C^p functions. I read the proof carefully; it holds together. The covariant Taylor expansion with uniform remainder is done cleanly; the telescoping and vanishing remainder are standard but handled with the right hypotheses. The invariance under admissible increment maps (Prop 2.5) and the transformation rule under smooth maps (Prop 2.11) are proved in full. The p=2 check against Föllmer's formula is a nice external benchmark.\n\nThe main soft spot is the definition of V^p(M,π): finite p-energy plus weak-* convergence of the atomic p-increment measure along the same partition sequence. That is a strong assumption and not implied by continuity or ordinary finite p-variation. The paper says so explicitly and gives examples (fBm with H=1/p, the Schied–Zhang fractal, and their manifold lifts) where it holds. So it is a real restriction on scope, but not a flaw in the theorem. The partition dependence noted in Remark 4.5 is standard for Föllmer-type calculus and does not invalidate anything.\n\nThe examples are useful: fBm lifted to S^2, H^2, SO(3) show how curvature and the exponential map appear in the correction terms. The fractal cubic-variation example on S^2 is a nice sanity check that odd-order variation can be non-trivial. Proposition 6.1 (almost-sure variation tensor of fBm with H=1/p) has a full proof via Hermite expansions and Borel–Cantelli; I see no gap.\n\nWho is this for? Anyone working on pathwise stochastic analysis on manifolds, or on rough-path-free approaches to irregular paths. The paper deserves a serious referee. I would accept it for review; my own recommendation would be accept after minor revision—mostly clarifying that the V^p condition is not verified for generic paths and maybe hedging the word \"arbitrary regularity\" in the abstract, since the regularity is controlled by p-energy, not truly arbitrary.","headline":"A clean, self-contained pathwise Itô calculus on manifolds for arbitrary p; the central theorem is sound, and the strong V^p hypothesis is honestly flagged rather than a hidden flaw.","tokens_in":27967,"tokens_out":2800,"would_cite":true,"duration_ms":25621,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H10","58A20","58C20","58C25","58C35","49Q15","60L99"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves an intrinsic change-of-variable (Itô-type) formula for smooth functions of continuous manifold-valued paths with finite p-th order variation along a partition sequence, with the p-th variation tensor as the sole…","keywords":["pathwise Itô calculus","higher-order variation","manifold-valued paths","p-th variation tensor","covariant Taylor expansion","affine connection","fractional Brownian motion","jet bundles"],"falsifier":"A concrete test: take the Schied–Zhang cubic-variation path w lifted to the sphere via the exponential map, then compute the weak-* limits of ∑($exp^{{-1}}$_{X_{t_i}}(X_{t_{i+1}}))^{⊗3} δ_{t_i} and of ∑(P_{X_{t_i}}(X_{t_{i+1}}-X_{t_i}))^{⊗3} δ_{t_i} along the same ternary partitions, using orthogonal projection as a second admissible increment map; Proposition 2.5 predicts identical limits, so a numerical discrepancy would falsify the independence claim that underlies the change-of-variable formula.","tokens_in":27026,"feed_emoji":"📐","tokens_out":6773,"duration_ms":67483,"temperature":0.7,"pith_summary":"The paper develops an intrinsic, purely pathwise calculus for smooth functions of continuous paths on a smooth manifold, with no probabilistic structure attached. Its central result is a change-of-variable formula for any test function f of class C^p: along a path X with finite p-th order variation, f(X_T)-f(X_0) equals the limit of covariant left Taylor sums of orders 1 through p-1 plus (1/p!) times the integral of the p-th covariant derivative of f against a p-th variation tensor [X]^p_π. For p=2 this reduces to Föllmer's pathwise Itô formula on Euclidean space, now on manifolds, and the construction requires only an affine connection, not rough-path enhancements. A sympathetic reader would care because it extends Itô-type calculus to arbitrarily rough manifold-valued paths, including fractional Brownian lifts and deterministic fractal paths, and it clarifies the geometric content of the Itô correction term.","feed_headline":"A change-of-variable formula for rough paths on manifolds","feed_subtitle":"Generalizes Föllmer's pathwise Itô calculus to any order p using geodesic increments and a connection.","key_machinery":"The load-bearing object is the p-th variation tensor [X]^p_π: a symmetric p-contravariant tensor-valued Radon measure along the path, obtained as the weak-* limit of atomic measures ∑ (Δ_i^π X)^{⊗p} δ_{t_i}, where increments are taken via an admissible increment map such as the logarithmic increment $exp^{{-1}}$_{X_{t_i}}(X_{t_{i+1}}). The paper proves the limiting measure is independent of the increment map, so the object is intrinsic to (X,π). The companion mechanism is the covariant Taylor expansion along geodesics: f(exp_x v)=Σ_{k=0}^p (1/k!)∇^{(k)}f_x($v^{{⊗k}}$) + o(|v|^p), uniformly on compact sets, which turns each discrete increment into a p-th order differential expression and identifies the p-th variation tensor as the canonical highest-order correction. A connection supplies the splitting of higher-order tangent vectors that makes the lower-order compensated sums well-defined.","core_discovery":"On the paper's own terms, the discovery is Theorem 4.1: for any integer p≥2, any continuous path X in V^p(M,π) — meaning the p-th order increment sums are uniformly bounded and the atomic measures of p-fold tensor increments converge weak-* to a symmetric tensor measure [X]^p_π along the same partition sequence — and any f∈C^p(M), the covariant left Taylor sums of orders 1,...,p-1 converge, and f(X_T)-f(X_0) equals their limit plus (1/p!)∫⟨∇^{(p)}f(X_t), d[X]^p_{π,t}⟩. The p-th variation tensor is defined intrinsically, independent of the admissible increment map used to form increments, and transforms under smooth maps by the p-th tensor power of the differential. When p=2 this is exactly Föllmer's pathwise Itô formula lifted to manifolds; for general p it gives a higher-order Itô-type calculus valid for paths of arbitrary low regularity.","pith_inferences":["The partition-dependence of V^p(M,π) suggests a natural test question: whether the change-of-variable formula is invariant across partition sequences when the p-th variation tensor exists along each, a property the paper does not establish.","Because the construction uses only geodesics and symmetrized covariant derivatives, one could run the entire theory with a connection carrying torsion and expect identical formulas; the paper's SO(3) example hints at this, but the general torsional case is not written out.","The examples are mostly exponential lifts of Gaussian or fractal drivers; a reader interested in the scope could test the formulas on other deterministic non-semimartingales, such as sums of Weierstrass-type oscillations, to see whether weak-* convergence of tensor measures is the effective regularity condition."],"forward_implications":["For p=2, the formula recovers Föllmer's pathwise Itô formula on manifolds, with the second variation tensor playing the role of the quadratic variation.","The p-th variation tensor of a transformed path Φ∘X is the fiberwise pushforward (DΦ)^{⊗p}[X]^p_π, making the change-of-variable formula functorial under smooth maps.","The pathwise integral defined by the compensated lower-order sums inherits an isometry property: for even p, its p-th variation is ⟨(df)^{⊗p}, d[X]^p_π⟩.","For fractional Brownian motion with Hurst parameter H=1/p lifted by the exponential map, the p-th variation density is (D exp)^{⊗p} m_p dt, producing explicit higher-order corrections that encode curvature on S^2, H^2, and the nonlinear structure of SO(3).","A deterministic fractal path on the sphere with non-zero cubic variation satisfies the formula with an explicit cubic correction term."],"supporting_citations":[{"why":"Supplies Itô's geodesic polygonal approximations and logarithmic increments used to define the p-th variation tensor.","marker":"[25]"},{"why":"Provides the Euclidean higher-order pathwise Itô calculus that the paper extends to manifolds.","marker":"[11]"},{"why":"Gives the p=2 pathwise Itô formula which the theorem recovers on manifolds.","marker":"[20]"},{"why":"Introduces Schwartz's second-order differential geometry, of which Section 5 gives a higher-order analogue.","marker":"[35]"},{"why":"Provides Émery's intrinsic Itô calculus on manifolds and the p=2 comparison for the correction term.","marker":"[17]"},{"why":"Supplies the local vector measure formalism used to interpret the p-th variation tensor.","marker":"[7]"},{"why":"Constructs the deterministic fractal path with non-zero signed cubic variation used in Section 6.3.","marker":"[34]"},{"why":"Establishes the 1/H-variation of fractional Brownian motion that Proposition 6.1 extends to the p-th variation tensor.","marker":"[31]"}],"fun_headline_variants":["Higher-order Itô calculus for rough manifold paths","Pathwise Itô formula on manifolds for any order p","Extending Föllmer's calculus to arbitrary variation order","Change of variable for very rough manifold paths","Pathwise Itô at any order on manifolds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything in the main theorem rests on the assumption that the path's p-th order increment sums are uniformly bounded and the atomic p-tensor measures converge weak-* along the chosen partition sequence; continuity and ordinary finite p-variation alone do not guarantee this.","fun_headline_variants_meta":{"raw":{"variants":["Higher-order Itô calculus for rough manifold paths","Pathwise Itô formula on manifolds for any order p","Extending Föllmer's calculus to arbitrary variation order","Change of variable for very rough manifold paths","Pathwise Itô at any order on manifolds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001261,"raw_usage":{"total_tokens":5197,"prompt_tokens":1008,"completion_tokens":4189,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":624,"completion_tokens_details":{"reasoning_tokens":4113}},"tokens_in":624,"tokens_out":4189,"duration_ms":30511,"temperature":1.0,"reasoning_tokens":4113,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:13:23.588849+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: take the Schied–Zhang cubic-variation path w lifted to the sphere via the exponential map, then compute the weak-* limits of ∑($exp^{{-1}}$_{X_{t_i}}(X_{t_{i+1}}))^{⊗3} δ_{t_i} and of ∑(P_{X_{t_i}}(X_{t_{i+1}}-X_{t_i}))^{⊗3} δ_{t_i} along the same ternary partitions, using orthogonal projection as a second admissible increment map; Proposition 2.5 predicts identical limits, so a numerical discrepancy would falsify the independence claim that underlies the change-of-variable formula.","supporting_citations":[],"review_version":1}