{"id":"4f78a712-30ba-4fc5-aab4-0a01a0a150d9","arxiv_id":"2511.18636","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"An LQ control framework for heterogeneous mean-field SDEs with common noise, solved through a triangular system of Hilbert-space Riccati BSDEs.","lead":"This paper builds a theory for optimal control of heterogeneous mean-field SDEs with a common noise, deriving a system of infinite-dimensional backward Riccati equations with existence, uniqueness, and a verification theorem. It applies the framework to optimal trading and systemic risk among heterogeneous agents.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fundamental relation Prop. 4.11 is unproven: square-completion identity in (B.4) is algebraically wrong.","rationale":"The reader's formal weakest_assumption is the uniform-in-label bound for K (Eq. 4.16), but their rationale also identifies the algebra error in Proposition 4.11. I judge the algebra error to be the single most load-bearing concern: the fundamental relation is the cornerstone of the verification theorem, and if the square-completion expansion is wrong, the entire characterization of the optimal control is unsupported. The uniform-bound issue is a technical gap that may be repairable under Assumption 3.1 (coefficients are uniformly bounded in u), whereas the algebraic error directly invalidates the proof as written. My recommended verdict remains REJECT, hence UNCHANGED. I agree with the reader's overall rejection but partially disagree on which assumption is weakest. I credit the paper for a plausible and likely repairable approach, but the current manuscript does not establish its central claim.","tokens_in":33051,"tokens_out":17304,"duration_ms":145702,"concrete_test":"Set d=m=1, take a single label (I={1}) with no mean-field integrals, and use a computer algebra system (e.g., SymPy) to compare the left side 2(UX + V\\bar X + Γ)α + Oα² with the printed right side of (B.4) after omitting the quadratic term and the \\bar K/driver terms. The difference should equal O^{-1}UV(\\bar X² − X\\bar X), showing the printed identity is false. If instead the two sides match, the reader's algebra critique is refuted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on Proposition 4.11 (fundamental relation), whose proof in Appendix B completes the square for χ_t(α^u) = 2⟨ζ_t, α_t⟩ + ⟨α_t, O_t α_t⟩ with ζ_t = U_t X_t + ∫_I V_t(u,v)\\bar X_v dv + Γ_t. The standard identity gives χ = ⟨α+O^{-1}ζ, O(α+O^{-1}ζ)⟩ − ⟨ζ, O^{-1}ζ⟩, so the subtracted term must be ⟨U X, O^{-1}U X⟩ + 2⟨U X, O^{-1}V\\bar X⟩ + 2⟨U X, O^{-1}Γ⟩ + ⟨V\\bar X, O^{-1}V\\bar X⟩ + 2⟨V\\bar X, O^{-1}Γ⟩ + ⟨Γ, O^{-1}Γ⟩. Display (B.4) instead contains −⟨U X, O^{-1}V\\bar X⟩ − ⟨V\\bar X, O^{-1}U\\bar X⟩ (the second has \\bar X in place of X) and −⟨V\\bar X, O^{-1}Γ⟩ − ⟨VΓ, O^{-1}X⟩ (again one X/\\bar X swap). Consequently the X–\\bar X cross term is halved and an extraneous −O^{-1}UV\\bar X^2 term appears; in (B.5) the required −(V^*O^{-1}V) term is missing from the \\bar K bracket, so (4.4) does not cancel. The proof cannot be closed as printed, leaving Theorem 4.15 and feedback (4.23) unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an LQ optimal control framework for heterogeneous mean-field SDEs with common noise, Eq. (3.1)-(3.6). The authors introduce a triangular system of backward Riccati equations: per-agent standard BSREs (4.1), an infinite-dimensional backward stochastic Riccati equation (4.4) on L2(I x I), a linear BSDE (4.8), and scalar linear BSDEs (4.11). The main theoretical results are a 'fundamental relation' (Prop. 4.11) decomposing the cost into a value function plus a nonnegative quadratic term, solvability theorems for the Riccati system, and a verification theorem (Thm. 4.15) identifying the optimal feedback control (4.23). Two financial applications are then presented: optimal trading with heterogeneous market participants and a systemic risk model with heterogeneous banks.","tokens_in":33454,"tokens_out":7815,"duration_ms":77646,"significance":"If correct, this would be a valuable first extension of heterogeneous mean-field control to settings with common noise, with a novel infinite-dimensional Riccati BSDE. The paper is clearly structured, transparent about the triangular nature of the Riccati system, and contains a careful treatment of label-measurability of conditional laws. However, the central proof of the fundamental relation contains an algebraic error, and the existence theory relies on an extra uniform-in-label bound that is assumed rather than proved. As written, the main theorem and the optimal feedback characterization are unsupported.","major_comments":[{"comment":"The square-completion identity for chi_t is algebraically wrong. For chi(alpha)=2+ with zeta=UX+int V \\bar X dv+Gamma, the correct subtraction is = + 2 + 2 + + 2 + . Display (B.4) instead writes -int - int (with \\bar X in place of X in the second term) and -int <\\bar X,V(v,u)^T O_v^{-1}Gamma_v> - (X and \\bar X swapped). Hence the X-\\bar X cross term is incorrect and an extraneous \\bar X-\\bar X term is created; in (B.5) the required -(V^*O^{-1}V) quadratic term is missing from the \\bar K bracket. The cancellation with F in (4.4) therefore fails, Proposition 4.11 is not proved, and Theorem 4.15 together with the optimal feedback (4.23), both resting on (4.14),","section":"Appendix B, Eq. (B.4)-(B.5)"},{"comment":"The bound sup_t |K^u_t| <= C, du x dP-a.e., is not a consequence of Assumption 3.1. The standard per-label solvability invoked above (4.16) yields constants depending on the u-labeled coefficients A^u, B^u, C^u, D^u, Q^u, R^u, H^u; since the label set I is uncountable, these constants need not be uniformly bounded. The manuscript simply assumes 'that the essential supremum with respect to the label is also finite.' This uniform bound is load-bearing: it is used to place (K,Z^K) in S^2_{F0}(I;R^{d x d}) x H^2_{F0}(I;R^{d x d}) and to justify the Borel representation in Step 2. The whole triangular system (4.4)-(4.11) and Theorem 4.15 depend on it. A proof, or an explicit uniform-in-label controllability-type assumption, must be supplied.","section":"Section 4.2.1, Eq. (4.16)"},{"comment":"The a priori estimate (4.19) and the contraction argument in Proposition 4.13 use the fundamental relation (4.14) as the mechanism for controlling ||T_{\\bar K_t}||. Since (4.14) rests on the invalid square completion identified above, the existence and uniqueness proof for (4.4) does not follow as written. In particular, the lower bound in (4.21) is obtained from (4.14) by 'taking the optimal control setting the quadratic term to be null', and the exact form of that relation is essential. The two defects together leave the solvability of the abstract Riccati equation open.","section":"Section 4.2.2, Props. 4.12-4.13"}],"minor_comments":[{"comment":"The parenthetical remark 'Not sure if I used them after' should be removed or resolved: either the spaces S^2_{\\tilde F0}(I x I;E), H^2_{\\tilde F0}(I x I;E) are used and should be formally introduced, or they are not and the sentence should be deleted.","section":"Section 2"},{"comment":"The coefficient lists for the applications are difficult to verify: in Section 5.1 the expression for M_t(u) runs into the definition of Gamma_t^u without a separator, and the integral in Lambda_t^u is written as \\int_U although U is not defined there; Section 5.2 has similar formatting problems. These displays should be rewritten with clear punctuation.","section":"Sections 5.1-5.2"},{"comment":"The displayed formula for G^H(u,v) is written with unbalanced parentheses and the claimed equivalence with the centered formulation (3.10) is not immediately verifiable. Please correct the display and check the expansion, especially the cross terms.","section":"Remark 3.8"},{"comment":"There are several typos: the abstract has 'in term of'; Remark 4.10(iv) says 'it will clear from the proof'; Appendix A contains mismatched subscripts in sup expressions (e.g., \\bar X^{Y,u}_q versus \\bar X^{Y,u}_s). These are presentation issues only.","section":"Throughout"}],"recommendation":"reject","confidential_remarks":"The paper addresses a timely and relevant problem and would be a good fit for the journal if the central derivation were repaired. However, the algebraic error in Appendix B is not a local typo: it changes the form of the fundamental relation and the cancellation needed from the abstract Riccati equation (4.4), and the uniform-in-label bound in Section 4.2.1 is an unproved assumption on which the entire triangular construction depends. I therefore do not see the main theorem as supported in the submitted version, and I would not invite a revision unless the authors provide a corrected square-completion argument and verify the resulting system."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things worth knowing. First, the paper addresses a real gap: LQ control for heterogeneous mean-field SDEs with common noise. That is a natural next step after [24] and [45], and the idea of a triangular system of Riccati BSDEs on L2(I) and L2(IxI) is reasonable. Second, the central proof currently does not close. I checked Appendix B against the stress-test note, and the note is right: the square completion in (B.4) has a sign/indices issue — the second cross term is instead of — and the quadratic term V*O^{-1}V is missing from the barX–barX bracket in (B.5). With those errors the driver (4.5) does not cancel, so Proposition 4.11 is not proven and Theorem 4.15 has no foundation.\n\nWhat is genuinely new: the problem formulation (3.1)-(3.6) with random coefficients and common noise, the well-posedness theorem for the state equation, and the explicit BSDE system. The literature review is thorough; the difference from [24] and [45] is real. The paper is honest: it flags that (4.4) is not covered by standard theory and then attempts an existence proof via contraction with a priori estimates.\n\nSoft spots, proportional. 1) The algebra error is load-bearing, not cosmetic. 2) In Section 4.2.1 the uniform-in-label bound (4.16) is simply assumed (\"assuming that the essential supremum with respect to the label is also finite\"). That is exactly the regularity needed to make (K,Z^K) jointly measurable and to start the triangular iteration. It may follow from stronger assumptions on the coefficients, but it is not derived. 3) In the trading application, the expression for M_t(u) includes a term -∫ κ Gκ_tilde(v,u) Y_t(v) dv, which belongs to the systemic-risk model; with G^A=0 in trading it should not be there. Minor by itself, but it suggests the applications were not fully re-checked.\n\nBottom line: this is a serious attempt, not a crackpot paper. But as printed it does not establish its theorem, and verifying it would require substantial algebra and regularity work. The right call is to send it to referees who know BSDEs on Hilbert spaces — they will be needed — but with the expectation of major revision. If the algebra is fixed and the uniform bound is made an assumption or proved, it would be a useful paper.\n\nI would not cite it in the next year. If our group works on heterogeneous mean-field control, bring it to a reading group to trace the square completion.","headline":"Good problem, plausible program, but the key square-completion identity in Prop 4.11 is algebraically wrong as printed, so the verification theorem and the main existence results are not established.","tokens_in":33962,"tokens_out":4016,"would_cite":false,"duration_ms":36244,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49N10","93E20","60K35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives a closed-form optimal control for heterogeneous mean-field systems with common noise, via a new infinite-dimensional Riccati BSDE system.","keywords":["mean-field SDE","common noise","heterogeneous agents","linear-quadratic control","backward stochastic Riccati equations","infinite-dimensional Hilbert spaces","optimal trading","systemic risk"],"falsifier":"Solve the scalar per-agent Riccati BSDE explicitly for a family of coefficients satisfying the paper's assumptions and check whether the supremum over labels of the solution is finite. If a family with uniformly bounded coefficients can be found for which this supremum is infinite, the uniform bound fails and the proof of the verification theorem cannot be carried out; alternatively, an independent numerical solution of the full LQ problem for the two applications can be compared with the claimed feedback to confirm the formula.","tokens_in":32911,"feed_emoji":"📈","tokens_out":5779,"duration_ms":54865,"temperature":0.7,"pith_summary":"This paper claims that a linear-quadratic control problem for a continuum of heterogeneous agents coupled through their conditional means and sharing a common Brownian noise can be solved exactly. The solution is a triangular system of four backward stochastic Riccati equations on Hilbert spaces of square-integrable kernels: a label-wise Riccati BSDE, a new infinite-dimensional Riccati BSDE for the interaction kernel, and two linear BSDEs. The paper proves existence and uniqueness for this system and derives the optimal control in explicit feedback form, with the value function given by a quadratic-plus-linear-plus-constant expression. A sympathetic reader would care because this is the first treatment of heterogeneous mean-field control with common noise, and the explicit formulas are demonstrated on optimal trading and systemic risk models.","feed_headline":"Explicit optimal control found for heterogeneous mean-field systems","feed_subtitle":"A four-equation Riccati system with common noise yields feedback controls for trading and systemic risk.","key_machinery":"The central machinery is the triangular system of backward stochastic Riccati equations (BSREs) on the Hilbert spaces L2(I;R^d) and L2(I×I;R^(d×d)), together with the fundamental relation obtained by completing the square. The first equation solves a standard per-agent Riccati BSDE for each label; the second is a new infinite-dimensional Riccati BSDE for the interaction kernel; the last two are linear BSDEs that absorb affine terms. The work they do is to make the cost functional a sum of a control-free term and a manifestly nonnegative quadratic term, so the zero of that term is the optimal feedback.","core_discovery":"The paper's central claim is that the conditional cost functional admits a fundamental relation: it equals a function of the initial state alone plus an integral of a nonnegative quadratic form in the control, shifted by a linear function of the state and the conditional mean. This identity is proved by completing the square, using the four Riccati equations. Because the quadratic term vanishes exactly at the stated feedback control, the minimizer is unique; the corresponding value function is the explicit quadratic-linear-constant expression given in the paper. The existence and uniqueness of the infinite-dimensional backward Riccati equations is the technical heart, established by a per-la","pith_inferences":["Editorial inference: the triangular structure suggests a practical numerical route—first solve the decoupled per-label Riccati BSDEs, then the kernel BSDE, then the two linear BSDEs—so the main computational obstacle is the infinite-dimensional kernel equation.","Editorial inference: a natural testable extension is a finite-N particle approximation of the continuum; the feedback control should be approximately optimal with an error governed by the regularity of the graphon kernels, though the paper does not compute such rates.","Editorial inference: the same framework should extend to vector-valued common noise and to controls entering both drift and diffusion, at the price of more complex Riccati coupling, since the argument only uses operator-norm and L2 structure.","Editorial inference: the load-bearing uniform bound on the per-agent Riccati solutions could be replaced by a label-wise bound plus a growth condition on the label, which would widen applicability to unbounded coefficient families."],"forward_implications":["If the central theorem is correct, the optimal control is given in closed loop by the feedback formula, so no numerical solution of the full control problem is required once the four Riccati equations are solved.","The value function separates into per-agent quadratic, kernel quadratic, linear, and constant terms, making the optimal cost computable from the four BSDE solutions.","The existence and uniqueness result covers a Riccati BSDE on an infinite-dimensional Hilbert space that standard theory does not handle, so the same triangular method applies to other LQ mean-field problems with common noise.","In the two financial applications, the theory yields explicit trading and lending/borrowing policies for heterogeneous agents facing common shocks.","The homogeneous and deterministic limits recover previously known Riccati ODE and mean-field results, providing consistency checks."],"fun_headline_variants":["Four Riccati equations solve heterogeneous mean-field control","Common noise optimal control for heterogeneous mean-field systems","Infinite-dimensional Riccati system yields unique optimal control","Trading and systemic risk solved via mean-field optimal control","Heterogeneous control: four Riccati equations for trading and risk"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The argument that the per-agent Riccati BSDE solutions are uniformly bounded in the label is assumed rather than proved; if the essential supremum over labels can be infinite while all other assumptions hold, the infinite-dimensional system and the verification theorem collapse.","fun_headline_variants_meta":{"raw":{"variants":["Four Riccati equations solve heterogeneous mean-field control","Common noise optimal control for heterogeneous mean-field systems","Infinite-dimensional Riccati system yields unique optimal control","Trading and systemic risk solved via mean-field optimal control","Heterogeneous control: four Riccati equations for trading and risk"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001743,"raw_usage":{"total_tokens":6659,"prompt_tokens":615,"completion_tokens":6044,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":359,"completion_tokens_details":{"reasoning_tokens":5963}},"tokens_in":359,"tokens_out":6044,"duration_ms":40878,"temperature":1.0,"reasoning_tokens":5963,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T20:41:28.399450+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve the scalar per-agent Riccati BSDE explicitly for a family of coefficients satisfying the paper's assumptions and check whether the supremum over labels of the solution is finite. If a family with uniformly bounded coefficients can be found for which this supremum is infinite, the uniform bound fails and the proof of the verification theorem cannot be carried out; alternatively, an independent numerical solution of the full LQ problem for the two applications can be compared with the claimed feedback to confirm the formula.","supporting_citations":[],"review_version":1}