Pith. sign in

REVIEW 3 major objections 1 minor 14 references

In continuous-time Kyle with Brownian privacy noise, price impact stays constant at λ = σ_v / √(σ_u² + σ_ε²) and cumulative subsidy to traders equals σ_v σ_ε² / √(σ_u² + σ_ε²).

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-29 19:51 UTC pith:7RMPZNHZ

load-bearing objection The paper states closed-form continuous-time privacy subsidy formulas and an LVR analogy but leaves the required Markovian linear equilibrium unproven. the 3 major comments →

arxiv 2605.25631 v4 pith:7RMPZNHZ submitted 2026-05-25 cs.GT cs.CRmath.PRq-fin.TR

The Privacy Subsidy in Continuous-Time Kyle: Cumulative Welfare under Noise-Perturbed Order-Flow Observation

Yuki Nakamura This is my paper
classification cs.GT cs.CRmath.PRq-fin.TR
keywords Kyle modelprivacy subsidycontinuous timeautomated market makerloss versus rebalancingorder flowBrownian noiselinear equilibrium
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper extends a closed-form privacy-subsidy calculation from the single-period Kyle model to continuous time. A committed Bayesian market maker sees aggregate order flow that has been perturbed by an independent Brownian privacy channel. Under the Markovian linear equilibrium the price-impact coefficient remains fixed over the interval, and the expected total transfer from the liquidity pool to traders is given by an explicit formula in the model parameters. The same quantity is shown to stand in a structural one-to-one relation with the Loss-Versus-Rebalancing gap once privacy noise is interpreted as an order-flow observation imperfection. This supplies the continuous-time Kyle ingredient needed to calibrate break-even fees for privacy-aware automated market makers.

Core claim

In the continuous-time Kyle model where order flow is observed through an independent Brownian privacy channel of intensity σ_ε, the Markovian linear equilibrium features a constant price-impact coefficient λ = σ_v / √(σ_u² + σ_ε²), and the cumulative expected transfer from the liquidity pool to traders over the unit interval equals |Π_M| = σ_v σ_ε² / √(σ_u² + σ_ε²). This quantity corresponds structurally to the Loss-Versus-Rebalancing gap when viewing privacy noise as the order-flow observation counterpart.

What carries the argument

The Markovian linear equilibrium of the continuous-time Kyle model under independent Brownian order-flow perturbation; it pins down the constant price impact and supplies the closed-form expression for the cumulative privacy subsidy.

Load-bearing premise

A Markovian linear equilibrium exists when the order flow is perturbed by an independent Brownian privacy channel.

What would settle it

Numerical simulation of the continuous-time trading game with chosen values of σ_v, σ_u and σ_ε; check whether the realized price impact remains exactly constant over [0,1] and whether the cumulative transfer matches the stated formula to machine precision.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The privacy subsidy supplies a concrete welfare-transfer figure that can be used to set fees in committed automated market makers.
  • Privacy-noise welfare is formally analogous to the Loss-Versus-Rebalancing gap under an order-flow rather than price observation interpretation.
  • The continuous-time Kyle leg of the break-even-fee program for privacy-aggregated environments is now complete.
  • The constant price-impact coefficient holds for the entire trading interval once the Markovian linear equilibrium is attained.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same subsidy formula may be used to compare welfare costs of different privacy mechanisms across discrete and continuous trading horizons.
  • Results from the LVR literature on fee design could be ported to privacy settings via the established structural correspondence.
  • Non-linear or non-Markovian equilibria, if they exist, would be natural candidates for testing whether the constant-impact property survives.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 1 minor

Summary. The paper extends the closed-form privacy-subsidy result from the single-period Kyle model (Nakamura 2026) to continuous time. Under the maintained assumption of a Markovian linear equilibrium, a committed Bayesian AMM observes aggregate order flow perturbed by an independent Brownian privacy channel of intensity σ_ε; the price-impact coefficient is claimed to be the constant λ = σ_v / √(σ_u² + σ_ε²) and the cumulative expected transfer from the liquidity pool to traders over [0,1] is |Π_M| = σ_v σ_ε² / √(σ_u² + σ_ε²). A structural correspondence is then drawn between this cumulative privacy subsidy and Loss-Versus-Rebalancing (Milionis et al. 2022).

Significance. If the equilibrium existence and derivations hold, the result supplies the continuous-time leg of the program for quantifying break-even fees in committed-AMM exchanges under privacy-aggregated information, by identifying privacy-noise welfare as the order-flow analog of LVR's price-observation gap.

major comments (3)
  1. [Abstract] Abstract and equilibrium claim: the formulas for constant λ and |Π_M| are presented under the maintained assumption that a Markovian linear equilibrium exists when the observed process is dY = dX + dU + dε with dε an independent Brownian motion; no derivation or verification of this equilibrium (including that the informed trader's best response remains dX_t = β(t)(v − p_t)dt with time-homogeneous λ) is supplied.
  2. [Main Result] Extension from single-period result: the continuous-time claim appears to obtain the same λ by simply replacing σ_u with √(σ_u² + σ_ε²), but the manuscript does not solve the dynamic filtering and stochastic-control problem under the perturbed observation process to confirm that time-homogeneity is preserved.
  3. [Equilibrium Analysis] Filtering step: the price p_t = λ Y_t is asserted to be the equilibrium price, yet the conditional expectation of v given the noisy cumulative order flow is not derived explicitly, leaving the martingale property and the form of the best-response strategy unverified.
minor comments (1)
  1. The notation σ_ε for the privacy-channel intensity should be introduced with an explicit definition of the observation equation dY_t = … + dε_t at first use.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thoughtful and detailed report. The comments correctly identify that the continuous-time extension relies on an unverified assumption of Markovian linear equilibrium. We address each point below and commit to revisions that will include the missing derivations.

read point-by-point responses

  1. Referee: [Abstract] Abstract and equilibrium claim: the formulas for constant λ and |Π_M| are presented under the maintained assumption that a Markovian linear equilibrium exists when the observed process is dY = dX + dU + dε with dε an independent Brownian motion; no derivation or verification of this equilibrium (including that the informed trader's best response remains dX_t = β(t)(v − p_t)dt with time-homogeneous λ) is supplied.

    Authors: The manuscript explicitly states the results hold 'under the Markovian linear equilibrium,' extending the single-period closed form by adjusting the effective noise intensity. We do not provide a full existence proof, as the focus is on the welfare calculation conditional on the equilibrium form. The referee is right that verification is needed for rigor. We will revise by adding an appendix that solves the filtering problem and confirms the strategy form and time-homogeneous λ. revision: yes

  2. Referee: [Main Result] Extension from single-period result: the continuous-time claim appears to obtain the same λ by simply replacing σ_u with √(σ_u² + σ_ε²), but the manuscript does not solve the dynamic filtering and stochastic-control problem under the perturbed observation process to confirm that time-homogeneity is preserved.

    Authors: The replacement follows directly from the additive property of independent Brownian noises in the observation variance, mirroring the static case. However, confirming preservation of time-homogeneity requires solving the dynamic problem, which is not done in the current version. In the revision, we will include the stochastic control formulation and show that the value function leads to constant λ. revision: yes

  3. Referee: [Equilibrium Analysis] Filtering step: the price p_t = λ Y_t is asserted to be the equilibrium price, yet the conditional expectation of v given the noisy cumulative order flow is not derived explicitly, leaving the martingale property and the form of the best-response strategy unverified.

    Authors: We acknowledge that the explicit computation of E[v | Y_{[0,t]}] via the innovation process or Kalman-Bucy filter is omitted. The paper assumes it follows the standard form with adjusted σ. We will add this derivation in the revised manuscript to explicitly verify the martingale property of p_t and the linear best response. revision: yes

Circularity Check

1 steps flagged

Core λ and |Π_M| formulas reduce directly to single-period self-citation via maintained equilibrium assumption

specific steps
  1. self citation load bearing [Abstract]
    "We extend the closed-form privacy-subsidy result of Nakamura~(2026, arXiv:2605.15746) from the single-period Kyle model to continuous-time. ... Under the Markovian linear equilibrium, the price-impact coefficient is λ = σ_v / √(σ_u² + σ_ε²) -- constant in time -- and the cumulative expected transfer from the protocol's liquidity pool to traders over [0,1] is |Π_M| = σ_v σ_ε² / √(σ_u² + σ_ε²)."

    The quoted λ and |Π_M| are identical in form to the single-period result with σ_u replaced by √(σ_u² + σ_ε²). The paper invokes the Markovian linear equilibrium assumption to carry the prior formulas forward without exhibiting an independent derivation of the informed trader's strategy β(t) or the filtering update that would confirm constancy of λ once dε is added to dY.

full rationale

The paper's headline results on constant λ and cumulative |Π_M| are obtained by extending the author's own prior single-period formulas (Nakamura 2026) under the assumption that a Markovian linear equilibrium with the same functional form persists in continuous time when an independent privacy Brownian motion is added to the observation. This reduces the continuous-time claim to the self-cited result by construction, as the time-homogeneous linearity is not re-derived from the stochastic control problem but maintained from the prior work. The LVR correspondence is external and does not affect the core formulas.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The claim rests on the existence of the Markovian linear equilibrium under Brownian privacy noise; all other elements are standard Kyle-model components.

axioms (1)
  • domain assumption Existence of Markovian linear equilibrium in continuous-time Kyle model with independent Brownian privacy channel
    All stated results are conditioned on this equilibrium as described in the abstract.

pith-pipeline@v0.9.1-grok · 5733 in / 1194 out tokens · 39175 ms · 2026-06-29T19:51:27.913738+00:00 · methodology

0 comments
read the original abstract

We extend the closed-form privacy-subsidy result of Nakamura~(2026, arXiv:2605.15746) from the single-period Kyle model to continuous-time. A committed Bayesian automated market maker observes the aggregate order flow perturbed by an independent Brownian privacy channel of diffusion intensity $\sigma_\varepsilon$. Under the Markovian linear equilibrium, the price-impact coefficient is $\lambda = \sigma_v / \sqrt{\sigma_u^2 + \sigma_\varepsilon^2}$ -- constant in time -- and the cumulative expected transfer from the protocol's liquidity pool to traders over $[0,1]$ is $|\Pi_M| = \sigma_v \sigma_\varepsilon^2 / \sqrt{\sigma_u^2 + \sigma_\varepsilon^2}$. We then establish a structural correspondence between this cumulative privacy subsidy and Loss-Versus-Rebalancing (Milionis et al.~2022), identifying privacy-noise welfare as the order-flow observation analog of LVR's price observation gap. The result completes the continuous-time Kyle leg of the program of quantifying break-even fees for committed-AMM exchanges under privacy-aggregated information environments.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

14 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    arXiv preprint arXiv:1908.08777 (2019)

    Aase, K.K., Øksendal, B.: Strategic insider trading equilibrium with a non-fiduciary market maker. arXiv preprint arXiv:1908.08777 (2019)

    work page arXiv 1908

  2. [2]

    Review of Financial Studies5(3), 387–409 (1992)

    Back, K.: Insider trading in continuous time. Review of Financial Studies5(3), 387–409 (1992)

    1992

  3. [3]

    Econometrica 78(1), 245–283 (2010)

    Caldentey, R., Stacchetti, E.: Insider trading with a random deadline. Econometrica 78(1), 245–283 (2010)

    2010

  4. [4]

    Annals of Applied Probability26(4), 1996–2029 (2016)

    C ¸etin, U., Danilova, A.: Markovian nash equilibrium in financial markets with asymmetric information and related forward-backward systems. Annals of Applied Probability26(4), 1996–2029 (2016)

    1996

  5. [5]

    In: Financial Cryptography and Data Security (FC 2022)

    Chitra, T., Angeris, G., Evans, A.: Differential privacy in constant function market makers. In: Financial Cryptography and Data Security (FC 2022). Lecture Notes in Computer Science, Springer (2022), iACR eprint 2021/1101

    2022

  6. [6]

    Stock Market Insider Trading in Continuous Time with Imperfect Dynamic Information

    Danilova, A.: Stock market insider trading in continuous time with imperfect dynamic information. Stochastics: An International Journal of Probability and Stochastic Processes82(1), 111–131 (2010), arXiv:1607.00035 14 Y. Nakamura

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  7. [7]

    Journal of Finance51(4), 1437–1478 (1996)

    Foster, F.D., Viswanathan, S.: Strategic trading when agents forecast the forecasts of others. Journal of Finance51(4), 1437–1478 (1996)

    1996

  8. [8]

    Journal of Business62(2), 211–235 (1989)

    Glosten, L.R.: Insider trading, liquidity, and the role of the monopolist specialist. Journal of Business62(2), 211–235 (1989)

    1989

  9. [9]

    Journal of Financial Economics14(1), 71–100 (1985)

    Glosten, L.R., Milgrom, P.R.: Bid, ask and transaction prices in a specialist market with heterogeneously informed traders. Journal of Financial Economics14(1), 71–100 (1985)

    1985

  10. [10]

    Econometrica53(6), 1315–1335 (1985)

    Kyle, A.S.: Continuous auctions and insider trading. Econometrica53(6), 1315–1335 (1985)

    1985

  11. [11]

    Automated Market Making and Loss-Versus-Rebalancing

    Milionis, J., Moallemi, C.C., Roughgarden, T., Zhang, A.L.: Automated market making and loss-versus-rebalancing. arXiv preprint arXiv:2208.06046 (2022)

    work page Pith review arXiv 2022

  12. [12]

    Nakamura, Y.: The privacy subsidy in Glosten–Milgrom: Bid-ask spread and welfare under flip-noise direction observation (2026)

    2026

  13. [13]

    Nakamura, Y.: The privacy subsidy: Kyle’s λ under noise-perturbed order-flow observation (2026)

    2026

  14. [14]

    AIMS Mathematics8(10), 25017–25036 (2023)

    Qiu, J., Zhou, Y.: Insider trading with dynamic asset under market makers’ partial observations. AIMS Mathematics8(10), 25017–25036 (2023). https://doi.org/ 10.3934/math.20231277

    work page doi:10.3934/math.20231277 2023