{"id":"6d354ec6-c2a0-4783-bf97-fedfc26a6652","arxiv_id":"2608.04937","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Mean-field doubly reflected forward-backward SDEs with optional barriers admit unique solutions on small time horizons for Lp data, and globally for p=2 under a monotonicity condition.","lead":"This paper proves existence and uniqueness for a new class of stochastic equations that combine mean-field interaction, forward-backward coupling, two price barriers, and data with finite p-th moment. The result gives a mathematical foundation for game-option valuation in large populations when payoffs can jump or lack right-continuity.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.3's proof is internally inconsistent: Eq. (4.2) drops the e^{-KT} terminal weight and has the wrong sign on the K-term, so the global contraction estimate does not follow as written.","rationale":"The reader's conditional verdict is appropriate: the short-time theorem appears structurally sound, while the global theorem has a genuine proof gap. The reader identified the sign typo in Eq. (4.2), which I agree with, but the more load-bearing issue is the unweighted terminal term: it is not controlled by the weighted estimate (4.9), and the missing factor e^{KT} is not covered by assumption C. This is an internal inconsistency in the proof of Theorem 4.3, not a disagreement with the surrounding theory. I do not recommend rejecting the paper because the local contraction argument in Theorem 4.2 is not compromised, and the global argument may be repairable by writing the terminal term with e^{-KT} and adjusting the constants. However, the paper should not be treated as final until the corrected identity is supplied and the resulting contraction constant is shown to satisfy the stated smallness condition. The B6/Mokobodzki concern raised by the reader remains a separate applicability caveat, but it is an assumption rather than a demonstrated proof gap. The concrete test I propose settles whether the global claim has a valid proof as currently formulated.","tokens_in":19984,"tokens_out":39110,"duration_ms":335446,"concrete_test":"Re-derive Eq. (4.2) from Itô's formula with U_t=e^{-Kt}(Y^1_t-Y^2_t)^2, keeping the terminal term e^{-KT}|δh|^2 and the +K∫ term. Then rerun the chain (4.5)-(4.10) with a terminal bound in the weighted norm. If the resulting contraction constant contains a factor e^{KT} not controlled by C(ii), Theorem 4.3's proof fails as written. This check is purely algebraic and needs no simulation; the corrected identity determines whether the arbitrary-horizon claim survives.","verdict_should_be":"UNCHANGED","load_bearing_attack":"For p=2, Theorem 4.3 claims a contraction on the exponentially weighted space S^{2,K}_F and therefore a global-in-time result. The proof hinges on Eq. (4.2). A direct Itô expansion for U_t = e^{-Kt}|Y^1_t-Y^2_t|^2 gives, on the left, e^{-Kt}|δY_t|^2 + ∫_t^T e^{-Kr}|δZ_r|^2 dr + K∫_t^T e^{-Kr}|δY_r|^2 dr, and on the right, the terminal term e^{-KT}|h(x^1_T,·)-h(x^2_T,·)|^2, not |δh|^2. The manuscript writes (-K)∫ e^{-Kr}|δY_r|^2 dr and an unweighted |δh|^2. The unweighted terminal term is the serious problem: inequality (4.7) contains |δh|^2, but (4.9) only bounds Ee^{-KT}|δh|^2. To pass from one to the other one must insert e^{KT}, and C(ii) imposes no control on e^{KT}(2l+(T∨1)+1/20). Thus, as written, the contraction constant in (4.10) is not derived. The sign error alone might be cosmetic, but the missing weight is essential for an arbitrary-horizon claim and is not a harmless constant absorption. Theorem 4.2, which uses only the small-time contraction, is not affected by this objection.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves existence and uniqueness for mean-field doubly reflected forward-backward SDEs with two optional barriers and L^p data, p in (1,2]. The system couples a forward SDE whose drift may depend on the joint law of (X,Y,Z) with a doubly reflected backward SDE whose generator also depends on that law. The main results are Theorem 4.2, a short-time existence and uniqueness theorem under smallness conditions on the horizon, and Theorem 4.3, a global-in-time theorem for p=2 under an additional monotonicity condition and an exponentially weighted norm. The proofs combine a contraction argument for the mean-field forward SDE (Proposition 2.2), a stability estimate for the mean-field reflected backward SDE (Proposition 3.2), and a fixed-point composition Phi composed with Psi on the forward component X.","tokens_in":20307,"tokens_out":9953,"duration_ms":86955,"significance":"If the main theorems hold, the paper fills a genuine gap in the literature: it combines mean-field dependence, forward-backward coupling, two reflecting barriers, optional barrier regularity, and law dependence on the martingale integrand Z in a single L^p framework. The fixed-point strategy is natural and the reliance on prior published results by Klimsiak, Rzymowski, and Slominski for the optional-barrier RBSDE component is appropriate. The paper also gives a plausible mean-field Dynkin game interpretation, while carefully marking the game-theoretic verification as beyond its scope. The significance is tempered by the fact that the global-in-time theorem, which is one of the two central claims, has a gap in the proof as written; the short-time theorem appears to be sound in its main lines.","major_comments":[{"comment":"The Itô expansion in Eq. (4.2) is not correct as written. For U_t = e^{-Kt}|Y^1_t - Y^2_t|^2, the correct identity contains a positive K term on the left and an exponentially weighted terminal term: e^{-Kt}|δY_t|^2 + ∫_t^T e^{-Kr}|δZ_r|^2 dr + K∫_t^T e^{-Kr}|δY_r|^2 dr = e^{-KT}|h(x^1_T,L(x^1_T))-h(x^2_T,L(x^2_T))|^2 + ... . The manuscript writes (-K)∫ e^{-Kr}|δY_r|^2 dr and an unweighted terminal term |h(...)-h(...)|^2. The sign error is repairable using condition (C)(i), but the missing e^{-KT} weight is not cosmetic: it propagates into (4.5), (4.6), and (4.7), where the terminal contribution appears as the unweighted random variable |δh|^2. Inequality (4.9) only controls E e^{-KT}|δh|^2; passing from the unweighted term in (4.7) to the contraction estimate (4.10) would require an uncontrolled factor e^{KT}, and condition (C)(ii) gives no control on 20 e^{KT}(2l+(T∨1)+1/20). Thus the contraction estimate for Theorem 4.3 is not derived as written.","section":"Section 4, Eq. (4.2)"},{"comment":"Even after restoring the missing e^{-KT} on the terminal term, inequality (4.9) does not follow from (B1') with the constant 2l. Since W_2(L(x^1_T),L(x^2_T)) ≤ ||x^1_T - x^2_T||_{L^2}, the Lipschitz condition (B1') gives |h(x^1_T,L(x^1_T))-h(x^2_T,L(x^2_T))| ≤ l(|δx_T| + ||δx_T||_{L^2}); after squaring and taking expectations one obtains a bound of order 4l^2 Esup e^{-Kt}|δx_t|^2, not 2l Esup e^{-Kt}|δx_t|^2. The constants in condition (C)(ii) and in inequality (4.10) therefore need to be adjusted consistently, and the assumption as stated does not currently imply the displayed contraction constant.","section":"Section 4, Eq. (4.9)"},{"comment":"The proof of Theorem 3.4 contains the sentence 'In the general case C≥1, we divide [0,T] into finitely many sufficiently small intervals' as the entire argument for removing the smallness condition on the contraction constant. For a mean-field reflected backward equation with optional barriers, gluing on subintervals requires checking terminal compatibility at each interior endpoint, preservation of the strong Mokobodzki condition on each subinterval, and consistency of the law fixed point across subintervals. These steps are not supplied, and the quoted existence result [19, Theorem 3.9] is invoked on the whole interval. The small-time contraction case is fine, but the general case of Theorem 3.4 needs a detailed argument rather than a one-sentence assertion.","section":"Section 3, Theorem 3.4"}],"minor_comments":[{"comment":"The left-limit notation is corrupted in the displayed text: strings such as '/leftr⫯g⊸tl⫯ne →X_s' appear instead of a proper notation such as X_{s-} or limsup_{r↑s} X_r; please repair these symbols so that the reflection conditions in Definitions 3.1 and 4.1 are readable.","section":"Section 1.2"},{"comment":"The class M_loc(0,T) is used in assumption (B6) without a definition; please add a one-line definition or a precise reference to the convention used in [19].","section":"Section 3, assumption (B6)"},{"comment":"The space S^{2,K}_F(0,T) is introduced in the proof of Theorem 4.3 but is not defined in the notation section; please define it before the theorem statement.","section":"Section 4, proof of Theorem 4.3"}],"recommendation":"major_revision","confidential_remarks":"The global-in-time result in Theorem 4.3 is the part that needs careful repair. The errors in Eq. (4.2) and the terminal-weight issue appear fixable, but as written the contraction proof for the arbitrary-horizon claim is not valid. The short-time result in Theorem 4.2 seems largely sound, and the paper's framing is honest about its reliance on prior work. I recommend major revision rather than rejection, provided the authors can produce a corrected weighted Itô estimate and adjust the constants in condition (C) accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short-time result is a real new combination and I think it holds. The global-in-time theorem, Theorem 4.3, has a proof gap that is not cosmetic: in Eq. (4.2), the Itô expansion should put e^{-Kt}|δY_t|^2 + ∫ e^{-Kr}|δZ_r|^2 dr + K∫ e^{-Kr}|δY_r|^2 dr on the left, and the terminal term should be e^{-KT}|δh|^2. The paper writes (-K) and unweighted |δh|^2. If you follow the written inequalities, (4.5)-(4.7) contain E|δh|^2 without the e^{-KT} factor, but (4.9) only bounds E e^{-KT}|δh|^2. Passing from one to the other multiplies by e^{KT}, and condition (C)(ii) imposes nothing on e^{KT}. So (4.10) is not derived. I checked the rest of the step: restoring the e^{-KT} weight in (4.5)-(4.7) makes (4.9) the right bound and the contraction goes through; so this looks like a fixable typo, but as written the theorem is unproven.\n\nWhat is genuinely new: the combination of mean-field dependence, forward-backward coupling, two optional barriers, L^p data for p in (1,2], and law dependence on Z. Table 1 is honest about the literature, the fixed-point scheme is standard but coherent, and the Dynkin-game reading is appropriately labeled a verification interpretation. Using [19] and [18] as black boxes is fine.\n\nThe softer spots: the 'divide into small intervals' step in Theorems 2.4 and 3.4 is asserted, not proved; gluing local solutions for reflected or mean-field problems requires matching conditions at the interfaces, and this is not automatic. The strong Mokobodzki condition (B6) is a heavy assumption and is not checked in any example, though the authors are upfront about it. Constants are absorbed without tracking, which makes verification tedious.\n\nFor whom: anyone working on mean-field FBSDEs, reflected BSDEs, or Dynkin games with irregular payoffs. It deserves a serious referee, but the referee should insist on fixing Theorem 4.3 and expanding the gluing argument. I would not cite it in its current form; I would reconsider once the global theorem is repaired.","headline":"Solid short-time existence for a genuinely new problem class; the global-time theorem has a fixable but real gap (dropped e^{-KT} weight in the contraction estimate).","tokens_in":20833,"tokens_out":5361,"would_cite":false,"duration_ms":42534,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H10","60H30","91A15","91G80"],"pacs":[],"model":"deepseek-v4-flash","headline":"Doubly reflected mean-field FBSDEs admit unique solutions when barriers may jump.","keywords":["mean-field FBSDE","doubly reflected BSDE","optional barriers","L^p data","Mokobodzki condition","Dynkin games","game options","McKean–Vlasov"],"falsifier":"Construct two optional barriers $L\\leq U$ of class (D), with $L_T\\leq h\\leq U_T$, that are pure-jump processes with jumps so large that no semimartingale of the form local martingale plus finite variation can lie between them and satisfy the integrability condition in $(B6)$; if the corresponding reflected backward equation then has no solution, the strong Mokobodzki assumption is not merely technical and the scope claimed in Theorems 4.2 and 4.3 would fail for such barriers.","tokens_in":19754,"feed_emoji":"📈","tokens_out":10102,"duration_ms":82529,"temperature":0.7,"pith_summary":"This paper establishes existence and uniqueness for a coupled system where a forward state process $X$ and a backward value process $Y$ solve a forward-backward SDE while the value is forced to stay between two barriers $L\\leq Y\\leq U$, and every coefficient may depend on the joint law of $(X,Y,Z)$. The barriers are only assumed optional, so they may jump and need not be right-continuous; the data need only have finite $p$-th moments for $1<p\\leq 2$. The intended application is recursive mean-field Dynkin games and game options, where $L$ is the holder's exercise or surrender payoff, $U$ is the counterparty's cancellation payoff, and the law dependence encodes population or market effects. The main theorems give a unique solution on short time horizons under a strong Mokobodzki separation condition, and for $p=2$ on arbitrary horizons under a monotonicity condition with an exponentially weighted norm. If the theorems hold, this class of irregular-barrier game-option valuation problems is well-posed.","feed_headline":"Mean-field game options with jump barriers get unique prices","feed_subtitle":"New theorems prove existence and uniqueness for reflected mean-field FBSDEs with optional barriers and L^p coefficients.","key_machinery":"The mechanism is a two-step fixed point. For a fixed forward path $x$, a map $\\psi$ solves the mean-field doubly reflected backward equation with optional barriers and returns $(Y,Z)$; for a fixed backward pair $(Y,Z)$, a map $\\phi$ solves the mean-field forward SDE and returns $X$. The composition $\\phi\\circ\\psi$ is shown to be a contraction because the stability estimates for both blocks scale with the Wasserstein distance between the joint laws, and for $p\\in(1,2]$ the law-of-$Z$ contribution is controlled by a factor of $T^{1/2}$; the smallness assumptions in Theorem 4.2 make the product constant less than $1$. For the global $p=2$ result, the proof replaces the $L^p$ norm by the exponentially weighted norm $E\\sup_t e^{-Kt}|X_t|^2$, and condition (C) chooses $K,\\alpha,\\beta$ so that both the backward and forward contraction factors stay below $1$ uniformly in $T$. The optional-barrier theory enters as a black box through existence and stability results for doubly reflected BSDEs with regulated trajectories.","core_discovery":"On the paper's own terms, the central claim is that the mean-field doubly reflected forward-backward SDE with optional barriers is well-posed: Theorem 4.2 constructs a unique quadruple $(X,Y,Z,R)$ with $X,Y\\in S^p_F$, $Z\\in H^p_F$, $R\\in V^p_{0,F}$ for every $p\\in(1,2]$ on intervals short enough that three smallness constants $c_pT^{1/2}$, $C_pT^{1/2}$, $MT^{1/2}$ are below $1$; Theorem 4.3 achieves the same for $p=2$ on an arbitrary time interval when condition (C) holds. The reflection is minimal: the finite-variation process $R$ increases only when $Y$ touches the lower barrier and decreases only when $Y$ touches the upper barrier, with jumps accounted for through optional left and right limits. The solution also satisfies the mean-field consistency condition $\\mu_t=\\mathcal{L}(X_t,Y_t,Z_t)$, so it represents the value of a representative agent in the associated Dynkin game.","pith_inferences":["The paper leaves open the $N$-player approximation question: the verification interpretation suggests an $\\varepsilon$-Nash result for finite-player Dynkin games, but that would need saddle-point and propagation-of-chaos arguments beyond the fixed-point proof.","The restriction $p\\leq 2$ appears structural: the paper's control of the law-of-$Z$ term reverses for $p>2$, so a similar short-time contraction would require a different norm or integrability assumption.","The strong Mokobodzki condition is inherited from the optional-barrier theory and is not verified on examples; checking whether natural barriers such as callable payoffs with credit events satisfy it would show how widely the results apply.","If analogous stability estimates hold under common noise or jumps, the same two-block contraction would likely produce well-posedness for mean-field reflected FBSDEs with common-noise and jump barriers."],"forward_implications":["Game options with exercise and cancellation payoffs that jump at irregular times have a unique continuation value and reflection process, so the recursive Dynkin game is well-posed in the mean-field setting.","Coefficients may depend on the joint law of the martingale integrand $Z$, not just on $(X,Y)$; short-time uniqueness covers this wider law dependence.","For $p=2$, the exponentially weighted fixed point gives existence and uniqueness on arbitrary horizons under monotonicity, so long-maturity contracts are within scope.","The constructed solution satisfies the mean-field consistency condition $\\mu_t=\\mathcal{L}(X_t,Y_t,Z_t)$, yielding a representative-agent equilibrium under the paper's verification interpretation."],"supporting_citations":[{"why":"Supplies existence of a unique solution for doubly reflected BSDEs with two optional barriers, the backward block used in the fixed-point map.","marker":"[19, Theorem 3.9]"},{"why":"Supplies the stability estimate for reflected BSDEs with regulated trajectories that Proposition 3.2 converts into the contraction estimate.","marker":"[18, Corollary 5.5]"},{"why":"Supplies existence and uniqueness for the mean-field SDE used as the forward block in the fixed-point map.","marker":"[26, Theorem 3.17]"},{"why":"Supplies the exponentially weighted norm technique that the global $p=2$ result borrows to remove the small-time restriction.","marker":"[2]"},{"why":"Introduces mean-field FBSDEs and fixes the law-dependence structure and Wasserstein setting that the paper extends to reflection.","marker":"[7]"}],"fun_headline_variants":["Mean-field game options with jump barriers: unique prices","Unique prices for mean-field game options with irregular barriers","Well-posed mean-field reflected FBSDEs with optional barriers","Existence and uniqueness for mean-field Dynkin games with optional barriers","Global uniqueness for mean-field reflected FBSDEs under monotonicity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the strong Mokobodzki condition $(B6)$: there must exist a semimartingale $S$, a local martingale plus an integrable finite-variation process, lying between the barriers $L$ and $U$ with enough integrability to make $f(\\cdot,0,S,0,\\mathcal{L}(0,0,0))$ integrable; if no such separating process exists, the reflected backward equation may have no solution and the main theorems do not apply.","fun_headline_variants_meta":{"raw":{"variants":["Mean-field game options with jump barriers: unique prices","Unique prices for mean-field game options with irregular barriers","Well-posed mean-field reflected FBSDEs with optional barriers","Existence and uniqueness for mean-field Dynkin games with optional barriers","Global uniqueness for mean-field reflected FBSDEs under monotonicity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000673,"raw_usage":{"total_tokens":3027,"prompt_tokens":872,"completion_tokens":2155,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":488,"completion_tokens_details":{"reasoning_tokens":2069}},"tokens_in":488,"tokens_out":2155,"duration_ms":13377,"temperature":1.0,"reasoning_tokens":2069,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:37:36.759378+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct two optional barriers $L\\leq U$ of class (D), with $L_T\\leq h\\leq U_T$, that are pure-jump processes with jumps so large that no semimartingale of the form local martingale plus finite variation can lie between them and satisfy the integrability condition in $(B6)$; if the corresponding reflected backward equation then has no solution, the strong Mokobodzki assumption is not merely technical and the scope claimed in Theorems 4.2 and 4.3 would fail for such barriers.","supporting_citations":[{"cited_title":"Solvability of infinite horizon McKean–Vlasov FBSDEs in mean field control problems and games.Appl","cited_arxiv_id":null,"evidence_quote":"Supplies the exponentially weighted norm technique that the global $p=2$ result borrows to remove the small-time restriction."},{"cited_title":"Mean field forward-backward stochastic differential equations","cited_arxiv_id":null,"evidence_quote":"Introduces mean-field FBSDEs and fixes the law-dependence structure and Wasserstein setting that the paper extends to reflection."}],"review_version":2}