{"id":"f3259ab5-285a-450f-9de3-a67b77387958","arxiv_id":"2501.18057","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a spider diffusion with controlled, local-time-dependent branch selection, the value function is characterized as the unique viscosity solution of a HJB system with a nonlinear local-time Kirchhoff boundary condition.","lead":"The paper introduces a new stochastic control problem for Walsh's spider diffusion, where the probability of picking a branch at the junction is controlled and depends on the local time spent at the junction. It proves that the optimal value function is the unique solution of a Hamilton-Jacobi-Bellman system with a nonlinear local-time Kirchhoff boundary condition.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2.7's vertex characterization rests entirely on the unproved local-time scaling estimate (24) from the in-preparation reference [38]; if that scaling fails, the nonlinear local-time Kirchhoff transmission is not recovered.","rationale":"The reader's weakest-assumption diagnosis is correct: the proof of Theorem 2.7 at the vertex hinges on estimate (24), and the paper explicitly delegates this estimate to [38], which is listed as in preparation. This is a genuine load-bearing gap rather than a stylistic issue. The structure of the argument is otherwise coherent: the weak DPP is adapted from [32] with the compactness of admissible rules proved in Section 3; the comparison theorem is sketched in Section 4 and traced to the accepted paper [40]; and the interior argument for the viscosity solution is standard. The vertex is the only place where the new local-time Kirchhoff transmission is actually derived, and (24) is the mechanism that converts the expected local-time increment into the ∂_l u term. Without a proof of (24), Theorem 2.7 is conditional on an external unpublished result. I also note a minor typographical issue: in the displayed martingale expression following the application of the DPP, the term S_i(u,l(u),ϑ) appears without ∂_x φ_i(u,0,l(u)); the surrounding text and Definition 2.3 make the intended expression unambiguous, so this is not a substantive objection. The verdict CONDITIONAL is appropriate: accept pending a complete proof of (24) or public availability of [38] with Proposition 8.1 verified.","tokens_in":38318,"tokens_out":6598,"duration_ms":80827,"concrete_test":"Independently prove (24) under assumption (H) from the martingale formulation (1) and the non-stickiness estimate (7). Concretely, for any admissible P and τ_h = inf{s ≥ t : x(s) = h}, derive E[l(τ_h) − l] = h(1 + o(1)) and E[τ_h − t] ≤ C h^2 by applying the Itô–Tanaka/occupation-time formula to the radial component and bounding drift and volatility terms using (H). If the derivation requires extra regularity of S_i or b_i in l near 0 beyond Lipschitz, that condition should be stated and added to (H); if instead one can exhibit coefficients satisfying (H) for which the limit is C h with C ≠ 1, then Theorem 2.7 is false as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 2.7 (Section 5, vertex case), the passage to the limit h→0 in the sub/supersolution inequalities is effected by (24): E^Q[τ_h − t*] ≤ C h^2 and h^{-1} E^Q[l(τ_h) − l*] → 1, where τ_h is the first exit from [0,h]. This estimate is quoted from [38, Prop. 8.1], which the reference list marks as \"in preparation\"; the present paper gives no proof. The inequalities (25)–(28) divide the vertex integral by h and let h^{-1} E^Q[l(τ_h) − l*] tend to 1. If the true asymptotic is C h with C ≠ 1, the recovered condition becomes ∂_l u + C · sup_ϑ {Σ_i S_i ∂_x u_i + h_0} = 0; if the local-time growth is not linear in h, no Kirchhoff-type transmission is obtained at all. Since Theorem 2.7 is precisely the claim that the value function solves (6), the central theorem's vertex part is unsupported until (24) is established. The other external inputs (DPP via [32], comparison via [40]) come from published or accepted works, whereas (24) is both unproved here and sourced to an unpublished preprint, making it the single most load-bearing gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper formulates a finite-horizon stochastic optimal scattering control problem for a Walsh spider diffusion whose spinning measure at the junction depends on the process's own local time and on time. The value function is defined as a supremum over suitably defined weak admissible rules, and the main results are: a weak dynamic programming principle (Theorem 2.4), a comparison theorem for continuous viscosity solutions of the associated backward HJB system with nonlinear local-time Kirchhoff transmission (Theorem 2.5), and a unique viscosity characterization of the value function by that HJB system (Theorem 2.7). The proofs rely heavily on prior work by the same group: the spider martingale problem in [37], the companion local-time analysis in [38], the linear parabolic theory in [39], and the comparison principle in [40]. The central novelty is the Kirchhoff-type boundary condition at the vertex involving the local-time derivative and an optimized convex combination of spatial derivatives.","tokens_in":38498,"tokens_out":3207,"duration_ms":41029,"significance":"If the main theorem is established, this appears to be the first stochastic control characterization for spider diffusions in which the diffraction probability itself is controlled and depends on the local time. The paper also provides a compactness result for admissible rules and a weak DPP in the presence of vertex controls, which are nontrivial extensions of [32]. The comparison theorem would extend [40] to the parabolic backward setting. The work is transparent in its reliance on prior results and does not engage in data fitting or circularity by construction. The significance is, however, contingent on closing a central analytic gap: the local-time scaling estimate (24) is quoted from an unpublished source and is the key ingredient that recovers the nonlinear local-time Kirchhoff transmission in Theorem 2.7.","major_comments":[{"comment":"The vertex part of Theorem 2.7 is load-bearing on the estimate E^{Q}[(l(τ_h)-l*)/h] → 1 and E^{Q}[τ_h - t*] ≤ C h^2, quoted from [38, Proposition 8.1], which is listed as \"in preparation\" in the references. The present paper gives no proof of this estimate. The passage h → 0 in inequalities (25) and (28) divides by h and uses both the time bound and the local-time asymptotic; if the true asymptotic were C h with C ≠ 1, the recovered vertex condition would have a different coefficient, and if the growth were not linear, no Kirchhoff-type transmission would be obtained. Because Theorem 2.7 claims exactly the unique characterization by system (6), this gap is central and must be resolved either by proving the estimate in this paper or by supplying a fully citable proof of the quoted result.","section":"Section 5 (proof of Theorem 2.7), Eq. (24)"},{"comment":"The comparison theorem is only sketched. After reducing to the exponentially weighted system (8), the interior case is dismissed with a reference to standard viscosity arguments, and the vertex case is said to follow the proof of Theorem 2.2 of [40] with the statement that the time variable \"will have no impact.\" The construction of the ODE test functions φ and φ in (16)-(17), the uniform bounds on their local-time derivatives, and the limiting passages (20)-(21) are not proved here, despite being essential to the contradiction argument. Since Theorem 2.7 uses Theorem 2.5 for uniqueness, the comparison theorem should either be stated as a black-box theorem imported verbatim from a published or accepted source with all hypotheses verified, or its proof should be completed in this paper.","section":"Section 4 (proof of Theorem 2.5)"},{"comment":"Several results that underlie the compactness of admissible rules and the dynamic programming principle are stated without proof. Proposition 3.4 (path estimates) and Proposition 3.5 (non-stickiness) are quoted as adaptations from [37], while Propositions 3.7-3.9 (stability by conditioning and concatenation) are stated as direct consequences of [44]. The compactness theorem 3.6 and the proof of Theorem 2.4 depend on these properties. If these statements do not hold verbatim in the controlled setting with vertex-dependent spinning measures and local-time-dependent coefficients, the DPP is not established. Please either prove them or give precise references with theorem numbers and verify that the assumptions of the current paper match those references.","section":"Section 3.2 and 3.3 (Propositions 3.4, 3.5, 3.7, 3.8, 3.9)"},{"comment":"In the displayed inequalities used to pass to the limit at the vertex, the terms involving the spinning measure are written as S_i(u, l(u), ϑ) without the factor ∂_x φ_i, e.g. \"Σ_i S_i(u, l(u), ϑ) + h_0\" in (25) and a similar expression in (28). The correct vertex Hamiltonian in (6) contains Σ_i S_i(t,l,ϑ) ∂_x φ_i(t,0,l). This appears to be a typographical omission, but it occurs exactly in the limiting argument that proves the Kirchhoff transmission, so the displayed computation should be corrected and the limit rechecked with the factor present.","section":"Section 5, Eqs. (25) and (28)"}],"minor_comments":[{"comment":"In the last Lipschitz modulus of (R-iii), the expression uses |b_i(t,x,l,β_i) - b_i(t,x,l',β_i)|/|l-l'|, but the displayed bound is attributed to |h|; it should presumably involve h_i. Please correct the typo.","section":"Section 2.1, Assumption (H), (R-iii)"},{"comment":"The notation ℓ^{ν_0(·)}(·)^2 is confusing; it should be made clear that the square applies to the local-time increment process over the interval [t,·].","section":"Section 3.2, Proposition 3.4"},{"comment":"The phrase \"we will obtain as soon as h ց 0\" in the subsolution case is grammatically incomplete; it should read \"as h → 0, we obtain.\"","section":"Section 5, proof of Theorem 2.7"},{"comment":"Reference [38] is marked \"In preparation\" and reference [40] is marked accepted but without a volume/year; since crucial results are imported from these works, the author should update the reference list to published or fully accessible versions, or state which parts of [38] are available.","section":"References"},{"comment":"The discussion of the local-time-independent case is informal and does not state a theorem; if this section is intended as a mathematical contribution, the claims in items (a)-(c) should be formulated as propositions with hypotheses and proofs or precise references.","section":"Section 6"}],"recommendation":"major_revision","confidential_remarks":"The paper's central claim depends on an estimate quoted from an in-preparation companion paper by the same group. Even granting good faith, this creates a verification problem for the reader and the referee: Theorem 2.7 cannot be certified from the submitted manuscript alone. The other external inputs ([37], [39], [40]) are either published or accepted, so the focus should be on either proving (24) in this paper or making the companion paper available in a citable form. The topic is appropriate for a mathematical analysis or stochastic control journal, and the result is likely significant if the gap can be closed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the problem is new and worth a referee's time, but the paper is not self-contained at its key vertex step.\n\nThe genuinely new thing here is the formulation: a stochastic control problem where the controller picks the diffraction probabilities at the spider's junction as a function of the current local time, with running and terminal rewards. That is not in the prior spider literature, which is about stopping or MFG with fixed spinning measures. The authors are also straight about the architecture: the weak DPP is from El Karoui–Nguyen–Jeanblanc, the comparison theorem is the author's own [40] adapted to the parabolic case, and the process estimates come from [37]. That is legitimate, but it means the reader is signing up for a chain of external results.\n\nThe central theorem, Thm 2.7, says the value function is the unique viscosity solution of the HJB system with the nonlinear local-time Kirchhoff transmission. The interior part is standard. The vertex part is where the real action is, and it depends on estimate (24): E[τ_h − t*] ≤ C h^2 and h^{-1}E[l(τ_h) − l*] → 1. That estimate is quoted from [38], which is marked in preparation. It is not proved here, and it is not a cosmetic gap: dividing by h and passing to the limit is exactly how the ∂_l u term appears in the Kirchhoff condition. If the local-time increment scaled differently, the boundary condition would come out with the wrong constant or not at all. So the characterization theorem hangs on an unpublished bootstrap.\n\nSome smaller soft spots: Propositions 3.4–3.5 and the stability-by-conditioning/concatenation results are stated without proof. Those are less troubling because they are routine adaptations and the paper says so. The comparison proof in Section 4 is a sketch, but [40] is accepted and the adaptation is plausible.\n\nOverall, this is a solid preprint with a real new formulation and an honest map of its dependencies. It is not yet a standalone proof of Theorem 2.7. The fix is also clear: either include a proof of (24) or wait for [38] to appear and be verified.\n\nWho is this for? People working on stochastic control on networks and spider diffusions. It deserves a serious referee; a good referee could check whether (24) follows from the martingale problem in [37] or needs genuinely new work.\n\nRecommendation: send it to peer review, with the expectation that the (24) gap gets closed or the paper comes with [38].","headline":"New control problem, honest about its debts, but the vertex argument rests on an unproved local-time estimate from an unpublished preprint.","tokens_in":39128,"tokens_out":2335,"would_cite":true,"duration_ms":26766,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["34B45","35R02","49L20","49L25","60G46","78A45","93E20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The value function of a spider diffusion controlled at its junction is the unique viscosity solution of a Hamilton–Jacobi–Bellman system with a nonlinear local-time Kirchhoff boundary condition.","keywords":["stochastic scattering control","Walsh's spider diffusion","nonlinear local-time Kirchhoff boundary condition","Hamilton-Jacobi-Bellman system","viscosity solutions","comparison principle","dynamic programming principle","local time"],"falsifier":"Simulate the spider with a deliberately sticky spinning measure at the vertex (or any measure violating the non-stickiness assumption) and measure, for small h, the ratio (1/h)E[l(τ_h)-l] where τ_h is the first time the radial part reaches h; if the limit is not 1, or if E[τ_h-t] does not decay like h², then the boundary condition in Theorem 2.7 is not the correct characterization for that process.","tokens_in":37984,"feed_emoji":"🕷️","tokens_out":7739,"duration_ms":82079,"temperature":0.7,"pith_summary":"This paper establishes a complete well-posedness theory for a new stochastic control problem: steering a Walsh spider diffusion by choosing, at each visit to the junction, the probability measure with which the process is scattered onto the branches. The value function of this scattering control problem is shown to be characterized as the unique continuous viscosity solution of a Hamilton–Jacobi–Bellman (HJB) system posed on a star-shaped network, with a new boundary condition at the junction called the nonlinear local-time Kirchhoff transmission. The boundary condition arises from the controller's ability to choose the diffraction measure as a function of the process's own local time at the vertex. A comparison theorem, a weak dynamic programming principle, and compactness of admissible rules are proved, yielding the unique characterization. If correct, the result turns an apparently singular control problem at a network junction into a tractable PDE problem.","feed_headline":"Optimal spider scattering obeys one Kirchhoff-type HJB law","feed_subtitle":"A single boundary transmission at the vertex links the optimal diffraction law to the spider's local time.","key_machinery":"The machinery is the Walsh spider HJB system with the nonlinear local-time Kirchhoff transmission at the vertex: alongside the ray-wise HJB equations, the value u must satisfy ∂_l u(t,0,l) + sup_ϑ { Σ_i S_i(t,l,ϑ) ∂_x u_i(t,0,l) + h_0(t,l,ϑ) } = 0, where l is the accumulated local time and (S_i) is a controlled probability vector on the branches. The proof of well-posedness rests on three supports: the weak dynamic programming principle in the spirit of [32], which needs a compactness theorem for admissible rules at the junction; a comparison theorem adapted from [40] that handles the nonlinear boundary transmission; and the non-stickiness estimate (24) from [38]—expected local time increase ≈ h before the radial coordinate first exits [0,h]—which is what converts the dynamic programming principle into the boundary equation in the limit h → 0.","core_discovery":"The central discovery is that the value function defined through the martingale formulation of the controlled spider is the unique continuous viscosity solution of the Walsh spider backward HJB system. On each ray, the usual HJB equation holds; at the vertex, a new transmission condition appears: ∂_l u(t,0,l) + sup_ϑ { Σ_i S_i(t,l,ϑ) ∂_x u_i(t,0,l) + h_0(t,l,ϑ) } = 0, where l is the local time accumulated at the junction and S(t,l,ϑ) is the controlled diffraction probability. The proof passes to the limit at the vertex using the estimate that local time grows linearly before the radial process first exits a small interval, and it relies on a comparison theorem for this class of HJB systems. The paper also proves the weak dynamic programming principle under junction controls and the compactness of the admissible rules, so that the supremum defining the value function is attained.","pith_inferences":["If the local-time growth estimate (24) can be verified for sticky or degenerate spinning measures, the same HJB characterization may extend to processes that spend positive time at the vertex; the paper explicitly leaves the degenerate case to future work.","The physics analogy drawn in the introduction suggests a concrete experimental test: a particle moving on a star-shaped photonic network whose ray affinities are updated by the local time at the junction should exhibit diffraction probabilities close to the ones maximizing the reward functional; such a prediction is not derived in the paper.","The comparison theorem opens a path to numerical solution of the problem via monotone schemes, though the paper does not report any computational experiment.","Because the HJB system is posed with l as an external variable, the same technique may apply to other reflected or boundary-controlled diffusions where the control acts only through the local time."],"forward_implications":["The optimal diffraction probabilities at the junction are determined by the unique solution of a PDE system; solving the HJB system once gives the value function everywhere without simulating the controlled process.","The comparison theorem ensures that any approximation scheme whose limits are viscosity solutions of the HJB system converges to the true value function.","Because the value function is continuous and admissible rules are compact, optimal controls exist in the space of relaxed controls; the paper identifies that optimal feedback controls at the vertex can be built from the jets of u, a point deferred to future work.","For problems whose data do not depend on local time, the same arguments yield a simpler well-posedness result for a nonlinear Kirchhoff transmission without the local-time derivative, as discussed in Section 6.","The characterization supplies a candidate verification theorem: to certify a control is optimal it suffices to check that the candidate value function is a viscosity solution of the HJB system with the local-time Kirchhoff condition."],"supporting_citations":[{"why":"Constructs the spider diffusion with local-time-dependent spinning measure and proves weak uniqueness, so the set of admissible rules is nonempty.","marker":"[37]"},{"why":"Provides the key estimate (24): expected local time grows like h before the radial part exits [0,h], used to derive the vertex boundary condition.","marker":"[38]"},{"why":"Supplies the comparison theorem for Walsh spider HJB systems with nonlinear local-time Kirchhoff transmission, on which the uniqueness proof is built.","marker":"[40]"},{"why":"Gives the weak dynamic programming principle framework and measurable selection lemmas adapted in the paper.","marker":"[32]"},{"why":"Proves well-posedness of the linear parabolic system with local-time Kirchhoff boundary, used for uniqueness and regularity in the linear case.","marker":"[39]"},{"why":"Provides the conditioning/concatenation and martingale problem tools used to prove compactness of admissible rules.","marker":"[44]"}],"fun_headline_variants":["Spider diffusion's optimal scattering: one Kirchhoff rule at the junction","Stochastic spider control picks a local-time-dependent diffraction law","New HJB boundary condition solves optimal spider scattering","Spider diffusion's optimal junction rule: HJB with local-time coupling","Weak dynamic programming yields unique spider scattering optimum"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole vertex argument rests on the estimate (24) from [38]—that before the radial coordinate first exits [0,h] the expected local-time increase is approximately h and the expected time is at most a constant times h²—so if local time grows sublinearly or superlinearly in that window, the derived Kirchhoff boundary condition fails.","fun_headline_variants_meta":{"raw":{"variants":["Spider diffusion's optimal scattering: one Kirchhoff rule at the junction","Stochastic spider control picks a local-time-dependent diffraction law","New HJB boundary condition solves optimal spider scattering","Spider diffusion's optimal junction rule: HJB with local-time coupling","Weak dynamic programming yields unique spider scattering optimum"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00066,"raw_usage":{"total_tokens":3031,"prompt_tokens":968,"completion_tokens":2063,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":584,"completion_tokens_details":{"reasoning_tokens":1995}},"tokens_in":584,"tokens_out":2063,"duration_ms":17548,"temperature":1.0,"reasoning_tokens":1995,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:50:25.094170+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the spider with a deliberately sticky spinning measure at the vertex (or any measure violating the non-stickiness assumption) and measure, for small h, the ratio (1/h)E[l(τ_h)-l] where τ_h is the first time the radial part reaches h; if the limit is not 1, or if E[τ_h-t] does not decay like h², then the boundary condition in Theorem 2.7 is not the correct characterization for that process.","supporting_citations":[{"cited_title":"Martingale problem for a Walsh ’s spider diffusion with spinning measure selected from its own local time","cited_arxiv_id":null,"evidence_quote":"Constructs the spider diffusion with local-time-dependent spinning measure and proves weak uniqueness, so the set of admissible rules is nonempty."},{"cited_title":"On spider diffusion having a spi nning measure selected from its own local-time","cited_arxiv_id":null,"evidence_quote":"Provides the key estimate (24): expected local time grows like h before the radial part exits [0,h], used to derive the vertex boundary condition."},{"cited_title":"Comparison principle for Walsh's spider HJB equations with non linear local time Kirchhoff's boundary transmission","cited_arxiv_id":"2312.01362","evidence_quote":"Supplies the comparison theorem for Walsh spider HJB systems with nonlinear local-time Kirchhoff transmission, on which the uniqueness proof is built."},{"cited_title":"Compac tification methods in the control of degenerate diffusions: Existence of an optimal control","cited_arxiv_id":null,"evidence_quote":"Gives the weak dynamic programming principle framework and measurable selection lemmas adapted in the paper."},{"cited_title":"and I.Ohavi","cited_arxiv_id":null,"evidence_quote":"Proves well-posedness of the linear parabolic system with local-time Kirchhoff boundary, used for uniqueness and regularity in the linear case."},{"cited_title":"Multidimensional Diffus ion Processes, (2004)","cited_arxiv_id":null,"evidence_quote":"Provides the conditioning/concatenation and martingale problem tools used to prove compactness of admissible rules."}],"review_version":1}