{"id":"f1d70cb3-bd85-47be-8036-04b7ea153d3d","arxiv_id":"2501.17180","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Quasi-invariance of Gaussian measures under the BO-BBM flow is established for the full global well-posedness range s > 1/2, improving the previous threshold s > 1.","lead":"This paper proves that Gaussian measures on Sobolev spaces remain equivalent to their own evolution under the periodic Benjamin-Ono-BBM equation for all regularities where the equation is known to have global solutions. It closes a gap left by earlier work, which only covered a smaller regularity range.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.5 chooses τ_R so that Lemma 3.4 cannot control the L^p norm used in the iteration; the error is concrete and repairable but must be corrected.","rationale":"The reader identified (3.22), the convergence of truncated flows, as the weakest assumption. That is a genuine gap, but it is a standard argument in nonlinear evolution equations and can be supplied. The more load-bearing defect is in Proposition 3.5, the technical heart of the paper: the choice of τ_R in (3.14) makes the subsequent application of Lemma 3.4 invalid. This is not a disagreement with the overall assessment that the paper is valuable; the main strategy — combining the variational formula with an iteration for long-time higher integrability — is sound and the short-time estimates in Lemmas 3.3 and 3.4 are carefully done. The error is concrete, localized, and evidently repairable: replacing (3.14) by pτ_R = c0/R makes the estimates consistent and matches the final exponential factor in the proof. Because the printed proof contains a false step in a central proposition, the appropriate verdict is conditional acceptance rather than acceptance as-is. The reader's weakest assumption is different, so I disagree with that selection, while noting that (3.22) should also be addressed in revision.","tokens_in":14,"tokens_out":18133,"duration_ms":221377,"concrete_test":"Re-run Proposition 3.5 with the corrected choice τ_R := c0/(pR) instead of (3.14). Verify that (3.16) now satisfies the hypothesis of Lemma 3.4, that (3.18) follows with a bound of the form exp(C(R)(c0/R)^A t_j p R/c0), and that the subsequent estimates (3.20) and (3.21) still close to give a finite L^{q(t)} bound for each fixed t,R with q(t)→1 as |t|→∞. If the estimates close, the defect is the typo in (3.14) and the theorem stands after correction; if they do not close, Proposition 3.5 is false and Theorem 1.1's quantitative claim is not established.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Proposition 3.5, the proof fixes p>r1 and defines q(0)=r1(p-1)/(p-r1), then chooses τ_R via (3.14): q(0)τ_R = c0/R. The recursive inequality (3.16) requires a uniform bound on ||f^N_τ 1_BR||_{L^p(dμ_{s,N})} with p>q(0). Lemma 3.4 gives such a bound only under pτ_R ≤ c0/R. But (3.14) yields pτ_R = p c0/(q(0)R) > c0/R, so Lemma 3.4 does not apply at the point it is invoked to obtain (3.18). This is not a mere cosmetic issue: without a valid L^p bound for f^N_τ, the recursive estimate (3.18) collapses and the uniform-in-N bound (3.10) — and hence the quantitative statement in Theorem 1.1 — is unsupported as written. The final displayed exponential factor in Proposition 3.5, exp(~C(R)(R(p-1)t/c0+1)), is exactly what arises if one instead sets pτ_R = c0/R, so the printed (3.14) appears to be a typo for pτ_R = c0/R. The secondary gap flagged by the reader, the unproved 'standard argument' in (3.22), is also real but is a standard compactness/uniqueness argument and less serious than this algebraic inconsistency.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the periodic Benjamin-Ono-BBM equation and proves that the Gaussian measures μ_s, for every s > 1/2, are quasi-invariant under the flow. Theorem 1.1 states that the transported measure (Φ_t)_#μ_s is mutually absolutely continuous with respect to μ_s, and moreover the Radon-Nikodym derivative f_{t,R} satisfies a localised L^p bound with p = p(|t|,R) > 1 tending to 1 as |t| → ∞. The proof combines the variational formula for exponential integrability with a dyadic decomposition of the relevant trilinear operator and a long-time iteration argument adapted from Forlano–Tolomeo. The paper also proves a complementary lower bound (Proposition 1.2) showing that a natural 'quasi-invariance benchmark' quantity diverges for 0 < s ≤ 1/2, which supports the conjecture that quasi-invariance fails there.","tokens_in":22789,"tokens_out":11370,"duration_ms":83900,"significance":"If the proof is completed, this is a substantial improvement over the earlier quasi-invariance result of Genovese–Luca–Tzvetkov, which required s > 1. The paper reaches the full range s > 1/2 in which periodic BO-BBM is globally well-posed, and it does so in the critical-dispersion regime where the integrability exponent must necessarily deteriorate in time. The quantitative nature of the density bound (1.11) and the accompanying negative result in Proposition 1.2 are valuable additions. The paper is generally careful, with detailed dyadic estimates and explicit use of the Gaussian structure, and it builds on recent techniques from Coe–Tolomeo and Forlano–Tolomeo, giving clear credit to those works.","major_comments":[{"comment":"The proof sets q(0)τ_R = c0/R with q(0) = r1(p-1)/(p-r1). The recursive inequality (3.16) requires a uniform bound on ‖f^N_τ 1_BR‖_{L^p} for p > q(0). Lemma 3.4, however, only supplies an L^p bound under the condition p τ_R ≤ c0/R. Since (3.14) implies p τ_R = p c0/(q(0)R) > c0/R, Lemma 3.4 cannot be invoked at the point where (3.18) is obtained. The final exponential bound in Proposition 3.5, namely exp(∼C(R)(R(p-1)t/c0 + 1)), matches the choice p τ_R = c0/R, indicating that (3.14) is a typo. The author should replace (3.14) by p τ_R = c0/R and then verify the subsequent estimates, in particular (3.20) and the treatment of general t, under this corrected scaling.","section":"§3, Proposition 3.5, Eq. (3.14)"},{"comment":"The limit lim_{N→∞} ‖Φ_t(u0) − Φ^N_t(Π_{≤N}u0)‖_{C([-T,T];H^{σ'})} = 0 is asserted to follow from global well-posedness and 'a standard argument', but no proof or reference is supplied. This convergence is essential for the dominated-convergence steps (3.24) and (3.25) that identify the limiting object f_{t,R} as the Radon-Nikodym derivative. The author should provide a proof, for instance by combining the continuity of the solution map for (1.1) with the fact that Φ^N_t(Π_{≤N}u0) solves the projected equation (3.1), or give a precise reference for this approximation statement.","section":"§3, Proof of Theorem 1.1, Eq. (3.22)"}],"minor_comments":[{"comment":"In the paragraph for Case 4, the displayed estimate has inconsistent notation: the left-hand side is written as Q^{(1)}_{s,N}(Y_{N_1}, V_{N_2}, Y_{N_3}) while the right-hand side uses \\v_{N_1}, \\Y_{N_2}, \\v_{N_3}. The intended case is (Y_{N_1}, V_{N_2}, V_{N_3}); the notation should be corrected to avoid confusion.","section":"§4.1, Case 4"},{"comment":"In the chain of estimates for the L^p norm of f^N_t 1_{B_R}, the displayed identity involving the subtraction 'μ_{s,N}(B_R)' after the exponential integral is not an identity. The middle step with the subtraction should be removed or replaced by an inequality; the final bound is unaffected.","section":"§3, Lemma 3.4 proof"},{"comment":"The statement that the global well-posedness at σ = 1/2 from [22] 'extends without additional complication to the periodic setting' would benefit from a brief explanation, since the entire paper works on the torus and this is the only place where the real-line result is invoked.","section":"§1"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a genuinely interesting result and the overall methodology is sound. The main issue is the algebraic inconsistency in Proposition 3.5: as written, the application of Lemma 3.4 is invalid, although the final formulas strongly suggest a simple typo. This is a load-bearing point and must be corrected. The unproved convergence (3.22) is standard but should be documented. With these fixes the paper would be a strong contribution to the quasi-invariance literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the paper carefully after seeing the stress-test note, and the stress-test is right about Proposition 3.5.\n\nThe main theorem extends quasi-invariance of Gaussian measures for periodic BO-BBM from s > 1 to the full well-posedness range s > 1/2. That matches the threshold the previous Genovese-Luca-Tzvetkov paper left open. The strategy is sensible: use the Coe-Tolomeo variational approach for exponential integrability and the Forlano-Tolomeo iteration to go from short to long times, with the energy cut-off BR doing the localization. The dyadic estimates in Section 4 are written out in real detail, and Proposition 1.2 gives new lower bounds suggesting singularity for s <= 1/2. None of that looks circular; the reliance on the author's own earlier tools is legitimate, since those results are cited as building blocks rather than assumed here.\n\nThe soft spot is in Proposition 3.5. Equation (3.14) chooses tau_R via q(0) tau_R = c0/R, with q(0) = r1(p-1)/(p-r1). But the recursive inequality (3.16) starts from ||f_tau 1_BR||_{L^p}, where p > q(0), and Lemma 3.4 only applies when p tau_R <= c0/R. With the printed choice, p tau_R = (p/q(0)) c0/R > c0/R, so Lemma 3.4 cannot control the very norm the iteration needs. This is a real gap in the proof as written. It is also clearly repairable: set p tau_R = c0/R instead. Then q(0) tau_R < c0/R, so the initial L^p bound and the later q(t) bounds both satisfy the Lemma 3.4 condition, and the iteration goes through. The final exponential factor in the proposition is consistent with that corrected choice. The author should fix this and re-check the constants.\n\nThe other thing I would ask for is a proof or reference for the truncated-flow convergence in (3.22). It is probably a standard compactness plus uniqueness argument, but it is load-bearing for identifying the limiting density as the Radon-Nikodym derivative.\n\nNet: the result is likely correct and the proof is almost there. It deserves a serious referee and should be accepted after the tau_R fix and a short justification of (3.22).","headline":"The paper makes a real threshold improvement in quasi-invariance for BO-BBM, but Proposition 3.5 contains a concrete exponent mismatch that must be fixed before the proof is complete.","tokens_in":23370,"tokens_out":3679,"would_cite":true,"duration_ms":32174,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q53","76B15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the Gaussian measure class is preserved by the periodic BO-BBM flow for every Sobolev regularity s > 1/2, matching the equation's global well-posedness range.","keywords":["Benjamin-Ono-BBM equation","quasi-invariance","Gaussian measures","Sobolev regularity","critical dispersion","exponential integrability","variational formula","trilinear symbol"],"falsifier":"Fix a regularity $\\frac12\\frac12$, the uniform exponential-integrability bound of Lemma 3.3 is false and Theorem 1.1 collapses; the paper's own calculation shows this divergence only for $01/2$, the same range where the equation is known to be globally well-posed. This matters because earlier quasi-invariance results required $s>1$, leaving a whole low-regularity regime open despite the existence of a global flow. The paper also proves a quantitative version: after inserting a conserved-energy cut-off, the transported density lies in $L^p$ for some $p=p(|t|,R)>1$ that tends to $1$ as $|t|\\to\\infty$, with bounds uniform in the frequency truncation. The main technical burden is the critical dispersion, which the paper handles by combining a variational formula for exponential moments with an iteration of short-time density bounds.","feed_headline":"Gaussian measures survive BO-BBM flow down to s > 1/2","feed_subtitle":"Extends quasi-invariance from s > 1 to the full range where the equation is globally well-posed.","key_machinery":"The central object is the trilinear operator $Q_{s,N}(u)$, defined through the symmetrized decomposition (3.5)--(3.8), which measures the infinitesimal change of the Gaussian Sobolev norm under the truncated flow. Its symbol $\\Psi_s(n_1,n_2,n_3)$ is controlled by Lemma 3.1, which places the derivative on the lowest-frequency factor and gives the needed gain. The exponential-integrability lemma (Lemma 3.3) is proved by a variational representation for exponential functionals, bounding $\\lambda|Q_{s,N}(Y_N+V_N)|$ against $\\frac12\\|V_N\\|^2_{H^{s+1/2}}$ via dyadic decompositions and random-oscillation estimates. Short-time $L^p$ bounds on the transported density follow (Lemma 3.4), and a recursive exponent-iteration argument (Proposition 3.5) extends them to arbitrarily long times.","core_discovery":"Theorem 1.1 establishes that for every $s>\\frac12$ and every $t\\in\\mathbb{R}$, the pushforward of the Gaussian measure $\\mu_s$ by the BO-BBM flow is mutually absolutely continuous with respect to $\\mu_s$. Moreover, for each $t>0$ and $R>0$ there is an integrability exponent $p=p(|t|,R)>1$, decaying to $1$ as $|t|\\to\\infty$, such that the Radon-Nikodym derivative $f_{t,R}$ of the localized transported measure lies in $L^p(d\\mu_{s,R})$ with a bound uniform in the frequency truncation. The proof reaches this by writing the log-density as an integral of a trilinear operator $Q_{s,N}$, symmetrizing that operator so the derivative falls on the lowest frequency, proving a uniform exponential-integrability bound for $Q_{s,N}$ against a cut-off ball, and then iterating short-time $L^p$ estimates to arbitrary times. A separate calculation gives a lower bound on a variance criterion that grows like $N^{1-2s}$ for $0\\frac12$, the uniform exponential-integrability bound would need an extra cancellation, suggesting the true threshold is lower than the well-posedness range.","The same two-step recipe—symmetrized trilinear symbol, variational exponential moments, and iteration of short-time bounds—could be tested on other equations with borderline dispersion, such as fractional BBM at the critical exponent, to see whether the same $p(|t|,R)\\to1$ pattern appears.","Removing the energy cut-off would require identifying the density without localization; a first testable step would be to check whether $f_{t,R}$ converges as $R\\to\\infty$ on bounded sets of the Gaussian support, which the paper leaves open.","The paper's divergence computation for $s\\le\\frac12$ suggests a possible numerical experiment: simulate the finite-dimensional truncated densities $f_t^N$ for small $s$ and observe whether their $L^p$ norms grow with $N$, which would corroborate the expected singularity."],"forward_implications":["Quasi-invariance now holds on the same Sobolev regularity range where the equation is known to be globally well-posed, so the gap between the dynamics and the measure-transport theory is closed.","For any fixed time and energy cut-off, the localized transported density has some $L^p$ integrability uniformly in the frequency truncation, making the transported measure a well-defined $L^p$ object rather than merely absolutely continuous.","The integrability exponent must tend to $1$ as $|t|\\to\\infty$, a loss attributed to critical dispersion; this predicts that no uniform-in-time higher integrability should be expected for this model.","For $s\\le\\frac12$, the variance criterion $QI_{s,N}(t)$ diverges polynomially (or logarithmically at $s=\\frac12$), giving concrete evidence for the paper's conjecture that the transported measure becomes singular in that range."],"supporting_citations":[{"why":"Establishes the baseline quasi-invariance result for s>1 that this paper extends to s>1/2.","marker":"[19]"},{"why":"Supplies the variational/exponential-integrability method and the symmetrisation idea for the trilinear symbol.","marker":"[11]"},{"why":"Provides the iteration scheme that turns short-time L^p bounds into long-time bounds.","marker":"[16]"},{"why":"Gives the global well-posedness in H^sigma for sigma>1/2 used to define the full flow and the truncation convergence.","marker":"[25]"},{"why":"Gives endpoint global well-posedness at sigma=1/2, identifying the full range that the theorem targets.","marker":"[22]"},{"why":"Gives the variational representation for exponential functionals used in the proof of the exponential-integrability lemma.","marker":"[7]"},{"why":"Provides an alternative variational formula for Laplace transforms used alongside the Boue-Dupuis representation.","marker":"[42]"},{"why":"Supplies the simplified variational formula invoked directly in (4.1) for the exponential moment of Q_{s,N}.","marker":"[15]"}],"fun_headline_variants":["Quasi-invariance for BO-BBM reaches s > 1/2","BO-BBM quasi-invariance now for all s > 1/2","Gaussian measures quasi-invariant down to s > 1/2","Critical dispersion tamed: BO-BBM quasi-invariance for s > 1/2"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the finite-dimensional truncated flows converge to the full flow uniformly on the Gaussian-measure support, exactly as asserted in (3.22); the paper cites global well-posedness and calls this standard, but gives neither a proof nor a reference.","fun_headline_variants_meta":{"raw":{"variants":["Quasi-invariance for BO-BBM reaches s > 1/2","BO-BBM quasi-invariance now for all s > 1/2","Gaussian measures quasi-invariant down to s > 1/2","Critical dispersion tamed: BO-BBM quasi-invariance for s > 1/2"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000213,"raw_usage":{"total_tokens":1406,"prompt_tokens":914,"completion_tokens":492,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":530,"completion_tokens_details":{"reasoning_tokens":404}},"tokens_in":530,"tokens_out":492,"duration_ms":4770,"temperature":1.0,"reasoning_tokens":404,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T14:50:21.042564+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix a regularity $\\frac12\\frac12$, the uniform exponential-integrability bound of Lemma 3.3 is false and Theorem 1.1 collapses; the paper's own calculation shows this divergence only for $01 that this paper extends to s>1/2."},{"cited_title":"Sharp quasi-invariance threshold for the cubic Szeg\\H{o} equation","cited_arxiv_id":"2404.14950","evidence_quote":"Supplies the variational/exponential-integrability method and the symmetrisation idea for the trilinear symbol."},{"cited_title":"Quasi-invariance of the Gaussian measure for the two-dimensional stochastic cubic nonlinear wave equation","cited_arxiv_id":"2409.20451","evidence_quote":"Provides the iteration scheme that turns short-time L^p bounds into long-time bounds."},{"cited_title":"Mammeri, Long time bounds for the periodic Benjamin-Ono-BBM equatio n, Nonlinear Anal","cited_arxiv_id":null,"evidence_quote":"Gives the global well-posedness in H^sigma for sigma>1/2 used to define the full flow and the truncation convergence."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives endpoint global well-posedness at sigma=1/2, identifying the full range that the theorem targets."},{"cited_title":"Bou´ e, P","cited_arxiv_id":null,"evidence_quote":"Gives the variational representation for exponential functionals used in the proof of the exponential-integrability lemma."},{"cited_title":"¨Ust¨ unel,Variational calculation of Laplace transforms via entropy on Wiener space and applications , J","cited_arxiv_id":null,"evidence_quote":"Provides an alternative variational formula for Laplace transforms used alongside the Boue-Dupuis representation."},{"cited_title":"Forlano, L","cited_arxiv_id":null,"evidence_quote":"Supplies the simplified variational formula invoked directly in (4.1) for the exponential moment of Q_{s,N}."}],"review_version":1}