{"id":"27dea199-c0ed-468a-b41f-7da90bd812c9","arxiv_id":"2504.17407","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Small forcing keeps KdV solitons near a time-varying traveling-wave family with high probability, and the amplitude's effective drift is given by an explicit stochastic differential equation.","lead":"This paper proves that solitons of the KdV equation remain close to a modulated traveling wave with high probability when the equation is forced by small deterministic and space-dependent random noise. It also derives explicit equations for how the soliton amplitude and position drift under the forcing.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"S1's noise regularity assumption is insufficient for the Hilbert-Schmidt bounds in Lemma 5.4, which underlie Proposition 5.1 and hence Theorem 1.1.","rationale":"The reader's own loaded concern was the uniform spectral-gap/amplitude-band condition C2, but that is an explicitly acknowledged timescale limitation of the method, not a gap in the argument for the stated range of parameters. The reader's stated reason for CONDITIONAL was the Theorem 1.2 flaw (dropped factor and unstated restriction in Lemma 9.3), which is real but does not affect the central stability theorem Theorem 1.1. The most load-bearing concern for the paper's central claim is instead the noise-regularity gap: under the stated S1, the Hilbert-Schmidt bounds used in every stochastic-convolution tail estimate are not guaranteed to be finite. This threatens the proof of Theorem 1.1 itself, not merely a secondary approximation result. It is an internally checkable technical condition rather than a disagreement with consensus. The concern is fixable by strengthening S1, and it is plausible the authors intended the stronger condition, so the appropriate verdict remains CONDITIONAL rather than REJECT; since the reader already recommended CONDITIONAL, the verdict is unchanged. I disagree with the reader's identification of the weakest assumption because the spectral-gap issue is a stated limitation while the noise-regularity issue is an unstated hypothesis failure.","tokens_in":43988,"tokens_out":29128,"duration_ms":271272,"concrete_test":"Let q be the inverse Fourier transform of the even, nonnegative function \\hat q(\\omega)=c(1+\\omega^2)^{-1}(\\log(e+\\omega^2))^{-3/4} (smoothed appropriately near the origin). (1) Verify q \\in H^1(\\mathbb R)\\cap L^1(\\mathbb R) by checking \\int(1+\\omega^2)|\\hat q(\\omega)|^2 d\\omega < \\infty and \\int|\\hat q(\\omega)|d\\omega < \\infty, plus enough decay of \\hat q'' for q\\in L^1. (2) Compute \\int\\omega^2 \\hat q(\\omega)d\\omega; it diverges. (3) For h(x)=e^{-wx}, evaluate the Hilbert-Schmidt norm in Lemma 5.4; it is infinite, showing that Lemma 5.4 cannot hold under S1. Alternatively, consult [9,28] to confirm the stricter covariance condition used there for well-posedness; if it is q_{1/2}\\in H^1, then S1 is mis-stated and Theorem 1.1's hypothesis must be amended.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.1 relies on the Gaussian tail bound in Proposition 5.1, whose noise estimate uses Lemma 5.4. Lemma 5.4 computes the Hilbert-Schmidt norm of the operator h T_{\\tilde\\xi} acting from L2_Q to H^1_w as ||h T_{\\tilde\\xi}\\cdot||_{HS}^2 = ||q_{1/2}||_{L^2}^2 ||h||_{H^1_w}^2 + ||q'_{1/2}||_{L^2}^2 ||h||_{L^2_w}^2. This is finite only if q_{1/2} \\in H^1, i.e. \\int(1+\\omega^2)\\hat q(\\omega)d\\omega < \\infty. The standing assumption S1 only requires q \\in H^1 \\cap L^1 with \\hat q \\ge 0. These conditions do not imply q_{1/2} \\in H^1: for a smoothed kernel with \\hat q(\\omega) = c(1+\\omega^2)^{-1}(\\log(e+\\omega^2))^{-3/4}, one has q \\in H^1\\cap L^1 and \\hat q\\ge 0, yet \\int\\omega^2 \\hat q(\\omega)d\\omega = \\infty, so q'_{1/2} \\notin L^2. Consequently the Hilbert-Schmidt bound in Lemma 5.4 fails for admissible S1 noise, the stochastic-convolution tail estimate in Proposition 5.1 is not justified, and the bootstrap in Proposition 8.1 leading to Theorem 1.1 is not established under S1 as stated. This is a missing regularity hypothesis in the statement of the main theorem, not merely a tightened constant: without q_{1/2}\\in H^1 the noise term may not even be Hilbert-Schmidt into the relevant H^1_w space. The fix is to strengthen S1 to require \\int(1+\\omega^2)\\hat q(\\omega)d\\omega < \\infty (or equivalently q_{1/2}\\in H^1), and to verify that Lemma 2.1's cited well-posedness results indeed hold under the strengthened assumption.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the stochastic Korteweg-de Vries equation with multiplicative, spatially homogeneous noise and deterministic forcing. It introduces a modulation decomposition around the soliton family, derives an effective SDE for the soliton amplitude, and proves two main results: Theorem 1.1, a weighted H^1 exit-time bound showing that the solution remains close to the modulated soliton with high probability, and Theorem 1.2, a bound on the time during which the actual amplitude is approximated by the reduced SDE. The proof strategy partitions time into intervals of length Delta_T, uses Pego-Weinstein semigroup estimates in exponentially weighted spaces, controls the unweighted L^2 growth by energy arguments, and combines these with Gaussian tail estimates for stochastic convolutions.","tokens_in":44427,"tokens_out":14052,"duration_ms":129015,"significance":"If the results are correct as stated, the paper is a significant extension of previous soliton-stability results for stochastic KdV, since it allows the deterministic forcing to produce O(1) amplitude modulation rather than only small fluctuations. The proof is substantial and does not rely on fitting or tuning of constants: the modulation system is derived explicitly and the stability argument is carried out through a bootstrap over local intervals. The main technical tool, namely freezing the linearized operator on short intervals and using the resulting spectral gap, is natural and is presented in considerable detail. However, two load-bearing issues need to be resolved before the stated theorems can be accepted.","major_comments":[{"comment":"The standing assumption S1 only requires q ∈ H^1 ∩ L^1 with ŵhat q ≥ 0. These conditions do not imply q_{1/2} ∈ H^1. For example, ŵhat q(ω) = c(1+ω^2)^{-1}(log(e+ω^2))^{-3/4} satisfies q ∈ H^1 ∩ L^1 and ŵhat q ≥ 0, but ∫ ω^2 ŵhat q(ω)dω = ∞. The Hilbert-Schmidt identity displayed in Lemma 5.4 contains the term ‖q'_{1/2}‖_{L^2}^2 ‖h‖_{L^2_w}^2, which is infinite for such q and generic h ∈ H^1_w. Since Lemma 5.4 is used in Lemma 5.5 and then in Proposition 5.1, the proof of Theorem 1.1 is not justified under S1 as stated. A strengthened hypothesis such as ∫(1+ω^2)ŵhat q(ω)dω < ∞ (equivalently q_{1/2} ∈ H^1) appears necessary; the authors should add it to S1 and verify that the cited well-posedness results hold under this strengthened assumption.","section":"§2 (S1); §5, Lemma 5.4"},{"comment":"The conclusion (1.10) is not supported by the proof. Lemma 9.3 produces, after imposing the restriction σ√T ≤ 1/(2C_{21}) inside its proof, a bound of the form C_{22} T e^{σ^2 T} η^2/λ. In the proof of Theorem 1.2 this is combined with Theorem 1.1, yielding in particular terms with an explicit factor T; the final step then drops every T factor and also drops the e^{σ^2 T} factor when claiming (1.10). For T ≥ 1 this is not a legitimate manipulation. Moreover, the left-hand side of (1.10) is nondecreasing in T while the right-hand side is independent of T, so the asserted bound cannot hold for large T unless the approximation error never exits the interval; the paper provides no argument for such a conclusion. Theorem 1.2 should be restated with the T-dependent bound actually proved, together with the hypotheses on σ√T, or the claim should be removed.","section":"§9, Theorem 1.2 and Lemma 9.3"}],"minor_comments":[{"comment":"In the displayed definition of Q, the integral is written as ∫ q(x−y) f(y) dx; the integration variable should be dy.","section":"§2, S1"},{"comment":"The proof states 'For ease of exposition, we consider T ≥ 1 for which T/ΔT ∈ N'. The reduction to such T is not explained; either the statement should mention that T can be enlarged to the next multiple of ΔT, or the final partial interval should be handled explicitly.","section":"§8, Proof of Theorem 1.1"},{"comment":"The mild Itô formula computation is very dense. A short remark identifying the exact version of the mild Itô formula used (reference [6, Theorem 1]) and explaining how the formal strong-form computation is justified would improve readability.","section":"Appendix B, Proof of Lemma 3.2"},{"comment":"There is a typo in the abstract: 'Kortew eg-de Vries' should read 'Korteweg-de Vries'. Several occurrences of the notation c_{ap} and Ω_{ap} are typeset inconsistently (appearing as 'cap' and 'Omega ap').","section":"Abstract and title"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious contribution and the main stability argument for Theorem 1.1 appears substantial and largely self-contained. However, the regularity gap in S1 affects the proof of the main theorem, and Theorem 1.2 as stated is not supported by the supplied proof. I would advise the editor to require a corrected statement of Theorem 1.2, either with the T-dependent bound and the necessary restrictions on σ√T, or with the claim removed from the main results."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a genuine step forward: the first rigorous stability theorem for KdV solitons under stochastic forcing that allows the amplitude to change by an O(1) factor over the time horizon. The piecewise local modulation setup is clever, the deterministic/stochastic decomposition is worked out in detail, and the proof of Theorem 1.1 through Section 8 is internally coherent. The citations to [27] and [28] are to companion works, not to the target result, and there is no circularity in the argument.\n\nTwo problems stand out, and one of them is in the main theorem. First, S1's assumption q ∈ H^1 ∩ L^1 is not enough for Lemma 5.4. The Hilbert-Schmidt identity in that lemma contains a term involving the derivative of q_{1/2}; it is finite only if the integral of ω^2 times the Fourier transform of q is finite. The stress-test example—hat q(ω) = c(1+ω^2)^{-1}(log(e+ω^2))^{-3/4}—satisfies S1 but has infinite second moment, so the operator hT_ξ is not Hilbert-Schmidt into H^1_w in general. The inequality in Lemma 5.4, and hence Lemma 5.5 and Proposition 5.1, fail as stated. The fix is straightforward: strengthen S1 to require that integral to be finite, i.e. q_{1/2} ∈ H^1. This is a missing hypothesis in the statement of the main theorem, not a cosmetic tightening.\n\nSecond, Theorem 1.2 is not supported. Lemma 9.3 carries an unstated restriction σ√T ≤ const from the Gronwall step, and its bound contains an e^{σ^2T} factor. In the proof of Theorem 1.2, the T factor is then dropped when arriving at (1.10). For fixed λ > 0 and large T, the exit probability for tap should tend to 1, so a T-independent bound cannot hold. This looks fixable only by restating the theorem with a time restriction or a T-dependence; as written, it is false.\n\nThe core idea is good and the main stability result will likely survive once the regularity assumption is added. Both gaps, however, need to be addressed before the statements can be trusted. I agree with the reader's conditional verdict, and the stress-test concern is real.\n\nThis paper is for people working on stochastic dispersive PDEs and phase-tracking methods. It deserves a serious referee, but the refereeing process should demand a corrected S1 and a repaired Theorem 1.2.","headline":"Real advance in stochastic KdV soliton stability, but S1 needs more regularity and Theorem 1.2's bound drops a T factor.","tokens_in":724,"tokens_out":1907,"would_cite":false,"duration_ms":62374,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H15","35Q53","35C08"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that stochastically forced KdV solitons remain close to the traveling-wave family on timescales where the deterministic forcing changes the amplitude by an O(1) factor.","keywords":["stochastic traveling waves","Korteweg-de Vries equation","stability","forcing","multiplicative noise","soliton modulation","weighted Sobolev spaces","translation-invariant noise"],"falsifier":"Take f≡1, choose any even q∈H¹∩L¹ with nonnegative Fourier transform, integrate (1.1) numerically from φ_{c*} and measure P[t_st(η)<T] at T=$σ^{{-1}}$ for small σ; the theorem predicts this probability is at most Cσ log(1/σ)+$Ce^{{-δη²/σ²}}$, so observing a probability bounded away from zero as σ→0 at fixed η would falsify Theorem 1.1. A cheaper test targets Proposition 7.1: with the same setup, measure the first time the amplitude exits [cmin,cmax]; the bound says this exit probability before T=$σ^{{-1}}$ is at most $e^{{-δ17/σ}}$, which should be tiny.","tokens_in":43781,"feed_emoji":"🌊","tokens_out":9164,"duration_ms":88151,"temperature":0.7,"pith_summary":"Stochastically and deterministically forced Korteweg-de Vries solitons are shown to stay close to the family of traveling waves, even while the effective soliton amplitude drifts by a nontrivial amount. The central bound is an exit-time estimate: with small noise σ and small forcing ε, the probability that the perturbation leaves a weighted H¹ neighborhood before time T is at most CTσ²log(1/σ)+$CTe^{{-δη²/σ²}}$. Earlier stochastic stability results applied only on times small compared to $σ^{{-2}}$, where amplitude motion was negligible; here the deterministic forcing is allowed to change the amplitude by an O(1) factor within the stability window. The paper also proves that the amplitude is accurately described, up to small-probability error, by an explicit effective stochastic differential equation.","feed_headline":"KdV solitons survive noise long enough to change size","feed_subtitle":"A new proof bounds how often noise knocks a KdV soliton off the traveling-wave family while its amplitude drifts","key_machinery":"The argument is carried by a variational-phase modulation decomposition together with a frozen-linearization step. The solution is written as a modulated soliton plus a remainder v, with orthogonality conditions ⟨v,φ_{c(t)}⟩=⟨v,ζ_{c(t)}⟩=0, where ζ_c=∫_{-∞}^x ∂_c φ_c dy; the modulation parameters c(t) and ξ(t) obey stochastic differential equations. On each short interval [T,T+ΔT] the paper freezes the amplitude at c(T) and studies a local remainder v_T against the semigroup generated by the linearized KdV operator L_{c(T)}. In exponentially weighted L² spaces this semigroup has a spectral gap of size w(c-w²) after removing two neutral modes, giving the exponential decay estimates (2.3)-(2.4) that damp the local remainder. Gaussian tail estimates for stochastic convolutions control the noise integrals on each interval, and a bootstrap over the partition proves the global exit-time bound.","core_discovery":"Under assumptions S1-S2, for any initial soliton φ_{c*} and any admissible deterministic forcing with ε∫|f|≤E, the modulation decomposition u(t,x+ξ(t))=φ_{c(t)}(x)+v(t,x) satisfies the exit-time bound P[t_st(η)<T]≤CTσ²log(1/σ)+$CTe^{{-δη²/σ²}}$, where t_st(η) is the first time the weighted H¹_w norm of v exceeds η. The two terms have distinct origins: the exponential term comes from linear damping on exponentially weighted spaces, while the σ²log(1/σ) term comes from the Itô drift growth of the unweighted L² norm combined with amplitude fluctuations. A companion theorem shows the reduced amplitude SDE (1.7) tracks the true amplitude for times up to T with error probability at most Cσ²/λ log(1/σ). The result establishes orbital stability of KdV solitons under combined deterministic and stochastic multiplicative forcing on timescales where the forced amplitude change is O(1).","pith_inferences":["An implication the authors leave implicit is that the log(1/σ) penalty comes from unweighted energy control; replacing that step by direct pointwise dispersive estimates should extend the result to times of order σ^{-2} without changing the rest of the architecture.","The frozen-linearization scheme should transfer to other solitary-wave PDEs whose linearized operators have a spectral gap in weighted spaces, since the components specific to KdV are the explicit soliton family and the L¹_w nonlinearity estimate.","Theorem 1.2 makes the reduced amplitude SDE a predictive tool: given the noise kernel q, one can compute g_Q and estimate when amplitude fluctuations become observable, before dispersive radiation has grown."],"forward_implications":["The bound P[t_st(η)<T]≤CTσ²log(1/σ)+CTe^{-δη²/σ²} makes stable soliton motion a high-probability event for T small compared with σ^{-2}log(1/σ)^{-1}.","The reduced amplitude SDE is quantitatively reliable: the probability that |c(t)-c_ap(t)| exceeds λ before time T is controlled by (Cσ²/λ)log(1/σ), so one can predict amplitude drift without resolving the dispersive radiation.","The deterministic forcing is allowed to change the amplitude by an O(1) factor during the stability window, meaning the result covers soliton amplification and decay rather than only infinitesimal fluctuations.","Stability can be propagated over many short intervals of length ΔT without needing a uniform-in-time linearization of the time-varying soliton, because each interval uses the frozen semigroup at its left endpoint."],"supporting_citations":[{"why":"Supplies the exponentially weighted space semigroup decay estimates (2.3)-(2.4) that form the linear stability backbone.","marker":"[25]"},{"why":"Provides the L¹_w treatment of the KdV nonlinearity used to bound the deterministic forcing term in Corollary 4.6.","marker":"[23]"},{"why":"Sets up the noise covariance S1 and supplies the formal modulation system whose leading-order amplitude SDE (1.7) is made rigorous here.","marker":"[27]"},{"why":"Gives the deterministic forcing analysis and energy estimates that the present paper extends to space-dependent noise.","marker":"[28]"},{"why":"Earlier energy-based stochastic stability result that limited times to σ^{-2} with negligible amplitude motion; the present result goes beyond it.","marker":"[10]"},{"why":"Maximal inequalities for stochastic convolutions used in Theorem 5.3 to obtain Gaussian tail bounds on noise integrals.","marker":"[24]"},{"why":"Burkholder-Davis-Gundy estimates used to control stochastic integrals in the unweighted energy and amplitude-approximation proofs.","marker":"[26]"},{"why":"Well-posedness of the stochastic KdV equation with multiplicative noise, quoted in Lemma 2.1 for mild solutions.","marker":"[8]"},{"why":"Phase-tracking by frozen linearization at intermediate times, adapted here to the time-varying soliton family.","marker":"[20]"}],"fun_headline_variants":["KdV solitons resist noise and keep traveling","Stochastic forcing shifts soliton amplitude, not stability","Noise-tolerant solitons: probabilistic stability bounds","Solitons stay near family despite random forcing","How long solitons survive noise: a new proof"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof needs the soliton amplitude c(t) to remain inside a fixed band [cmin,cmax] with w<√cmin/3, so that the spectral gap b<w(c-w²) stays strictly positive over the whole time interval; if the noise or forcing pushes c(t) out of this band, the linear decay estimates lose their constants and the iterative stability argument collapses.","fun_headline_variants_meta":{"raw":{"variants":["KdV solitons resist noise and keep traveling","Stochastic forcing shifts soliton amplitude, not stability","Noise-tolerant solitons: probabilistic stability bounds","Solitons stay near family despite random forcing","How long solitons survive noise: a new proof"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000252,"raw_usage":{"total_tokens":1519,"prompt_tokens":864,"completion_tokens":655,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":480,"completion_tokens_details":{"reasoning_tokens":579}},"tokens_in":480,"tokens_out":655,"duration_ms":6400,"temperature":1.0,"reasoning_tokens":579,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:42:35.284745+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take f≡1, choose any even q∈H¹∩L¹ with nonnegative Fourier transform, integrate (1.1) numerically from φ_{c*} and measure P[t_st(η)<T] at T=$σ^{{-1}}$ for small σ; the theorem predicts this probability is at most Cσ log(1/σ)+$Ce^{{-δη²/σ²}}$, so observing a probability bounded away from zero as σ→0 at fixed η would falsify Theorem 1.1. A cheaper test targets Proposition 7.1: with the same setup, measure the first time the amplitude exits [cmin,cmax]; the bound says this exit probability before T=$σ^{{-1}}$ is at most $e^{{-δ17/σ}}$, which should be tiny.","supporting_citations":[{"cited_title":"Pego and M.I","cited_arxiv_id":null,"evidence_quote":"Supplies the exponentially weighted space semigroup decay estimates (2.3)-(2.4) that form the linear stability backbone."},{"cited_title":"Mizumachi and N","cited_arxiv_id":null,"evidence_quote":"Provides the L¹_w treatment of the KdV nonlinearity used to bound the deterministic forcing term in Corollary 4.6."},{"cited_title":"Westdorp and H.J","cited_arxiv_id":null,"evidence_quote":"Sets up the noise covariance S1 and supplies the formal modulation system whose leading-order amplitude SDE (1.7) is made rigorous here."},{"cited_title":"Westdorp and H.J","cited_arxiv_id":null,"evidence_quote":"Gives the deterministic forcing analysis and energy estimates that the present paper extends to space-dependent noise."},{"cited_title":"De Bouard and A","cited_arxiv_id":null,"evidence_quote":"Earlier energy-based stochastic stability result that limited times to σ^{-2} with negligible amplitude motion; the present result goes beyond it."},{"cited_title":"van Neerven and M.C","cited_arxiv_id":null,"evidence_quote":"Maximal inequalities for stochastic convolutions used in Theorem 5.3 to obtain Gaussian tail bounds on noise integrals."},{"cited_title":"Veraar and L","cited_arxiv_id":null,"evidence_quote":"Burkholder-Davis-Gundy estimates used to control stochastic integrals in the unweighted energy and amplitude-approximation proofs."},{"cited_title":"De Bouard and A","cited_arxiv_id":null,"evidence_quote":"Well-posedness of the stochastic KdV equation with multiplicative noise, quoted in Lemma 2.1 for mild solutions."},{"cited_title":"MacLaurin","cited_arxiv_id":null,"evidence_quote":"Phase-tracking by frozen linearization at intermediate times, adapted here to the time-varying soliton family."}],"review_version":1}