{"id":"63d9b363-8d52-47d4-a455-689c541a5cc0","arxiv_id":"1908.06358","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The authors show that maximum-entropy reasoning, together with a scale-invariance argument for log returns, yields the standard Garman-Kohlhagen foreign-exchange option pricing model.","lead":"This paper derives the standard foreign-exchange option pricing model from a maximum-entropy principle, arguing that scale invariance forces one to model the logarithm of the exchange rate. The result is a new derivation route for a known formula, not a new pricing model.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Abstract claims a derived risk-neutral measure, but Section 3.1 imposes Eq. 27; without a no-arbitrage/martingale derivation, the Garman-Kohlhagen formula is standard finance grafted onto the entropic model, making the derivation claim conditional.","rationale":"The paper is best read as an attempt to re-found GBM and Garman-Kohlhagen option pricing on entropic inference. The derivation of the lognormal transition (23) from the maximum-entropy constraints is internally coherent, and the final call/put formulas (37)/(39) and the BSM PDE (46) are the standard results, which is independent support that no algebraic error has crept in. The load-bearing weakness sits at the boundary between the real-measure dynamics and the pricing measure. The abstract promises a derivation of the risk-neutral measure, but Section 3.1 implements it by replacing the drift in Equation (15) with the interest-rate differential (Equation 27) and states 'we impose'. No entropy maximization is performed at this step, no martingale condition is verified, and no uniqueness argument rules out other equivalent measures. Because the option price is set by expectation under the pricing measure, this step carries the entire option-pricing conclusion. This matches the reader's weakest_assumption exactly. A conditional verdict is appropriate rather than rejection: the formulas are standard and correct, and the gap is repairable by an explicit maximum-entropy/martingale derivation, but as published the paper's strongest claim overstates what is actually derived. The concrete check is to perform that derivation from Equation (23) with the martingale constraint; the outcome determines whether the gap is merely an omitted proof or a genuine external postulate.","tokens_in":11120,"tokens_out":11205,"duration_ms":113107,"concrete_test":"Re-derive Section 3.1 without importing Equation (27). Start from the physical entropic density (Equation 23) and find the transition density Q that maximizes S[Q,P] in Equation (6) subject to normalization and the no-arbitrage martingale constraint E_Q[u' e^{-(r_d-r_f)Δt} | u] = u. If the resulting Q differs from Equation (28), then Equation (27) is an external postulate and the central derivation claim fails; if Q equals Equation (28), the gap is closed and the paper needs only an explicit derivation in place of the imposed constraint.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the entropic framework, not external finance, produces the pricing measure. In Section 3.1 the risk-neutral constraint is introduced as Equation (27) with the words 'we impose the first risk-neutral constraint on Equation (15)', after the abstract says 'we derive a risk-neutral measure.' No argument shows that Equation (27) is the minimal-entropy update from the real-measure density (23) consistent with no arbitrage, no Girsanov/Radon-Nikodym computation is given, and no martingale verification for e^{-(r_d-r_f)t}u_t under Equation (28) is supplied. The skipped derivations noted for Equations (20) and (43) are secondary; the risk-neutral step is the load-bearing one. If Equation (27) is an input rather than a consequence, then Equations (37) and (39) are the standard Garman-Kohlhagen result obtained by assuming lognormal FX and risk-neutral valuation, not a consequence of entropic inference alone. The formula itself is correct, so the issue is one of derivation completeness and claimed novelty, not of numerical error.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an entropic-dynamics model of exchange rates. It maximizes relative entropy subject to continuity and drift constraints to obtain a Gaussian transition density for the log exchange rate, which it identifies with geometric Brownian motion. It then introduces a risk-neutral drift and derives the Garman–Kohlhagen European option pricing formula and the Black–Scholes–Merton PDE. The paper claims that this constitutes a derivation of GBM and of risk-neutral option pricing from entropic inference and scale invariance.","tokens_in":11393,"tokens_out":10457,"duration_ms":87363,"significance":"If the claims were fully established, the paper would offer an information-theoretic foundation for a standard result in foreign-exchange option pricing, which is conceptually interesting. The maximum-entropy calculation is explicit, and the structure clearly shows which constraints produce the Gaussian transition density. However, the risk-neutral measure is imposed rather than derived, the uniqueness proof for the logarithmic variable is mathematically incorrect, and a key second-moment relation is assumed without derivation, so the central claims of the abstract are not supported as stated. The final formula, once corrected for a sign error, is the standard Garman–Kohlhagen price, which limits the novelty to a reformulation rather than a new pricing result.","major_comments":[{"comment":"The paper states in the abstract and in Section 4 that it 'derives a risk-neutral measure,' but Eq. (27) simply replaces the real drift with the domestic-foreign interest-rate difference as a constraint. No derivation is given of why this is the minimal-entropy update from the real-measure density, and no Girsanov or martingale verification is supplied. Consequently, the option price in Eq. (37) follows from the standard assumption of a lognormal FX rate under a risk-neutral measure, not from entropic inference alone. This is the load-bearing step for the paper's central claim and needs either a genuine derivation or an explicit admission that the risk-neutral drift is an input.","section":"Section 3.1, Eq. (27)"},{"comment":"The definition of d1 in Eq. (34) has the wrong sign on the σ^2T/2 term. For a lognormal variable with mean ln u0+(rd-rf-σ^2/2)T, the expected sale value is u0 e^{(rd-rf)T} N(d1) with d1 = [ln(u0/K)+(rd-rf+σ^2/2)T]/(σ√T), not the expression given, which is actually the standard d2. With d2 defined by d2 = d1 - σ√T, the call price in Eq. (37) and the put price in Eq. (39) do not equal the Garman–Kohlhagen prices. This sign error must be corrected for the central result to be valid.","section":"Section 3.1, Eqs. (33)-(39)"},{"comment":"The relation ⟨(ln u'/u)^2⟩_P = ⟨(Δu/u)^2⟩_P = σ^2Δt is stated after a skipped calculation, but it is not a consequence of the maximum-entropy posterior. From Eq. (17), ⟨(ln u'/u)^2⟩_P = (βσ^2Δt)^2 + σ^2Δt, which reduces to σ^2Δt only to leading order; the equality with ⟨(Δu/u)^2⟩ is an additional assumption. This assumption is what fixes the -σ^2/2 correction in the drift and hence the GBM form, so the derivation of GBM is circular: the GBM drift is effectively inserted through Eqs. (15) and (20) and then recovered in Eq. (24).","section":"Section 2.4, Eq. (20)"},{"comment":"The claim that the scale-invariance condition uniquely yields f(u)=ln u is incorrect. The functional equation C(l)+C(l')=C(ll') has the general solution C(l)=a ln l for any real a, leading to f(u)=a ln u+b. In addition, Eq. (2) is not the correct transformation law for probability densities under a change of variables; densities transform with a Jacobian. The logarithmic variable can be motivated by scale invariance, but the uniqueness proof as stated is false and should be replaced or removed.","section":"Section 2, Eqs. (3)-(5)"}],"minor_comments":[{"comment":"The phrase 'To drive the risk-neutral measure' should be 'To derive the risk-neutral measure,' and the introduction to Section 3 contains a typo 'German-Kohlhagen' that should be 'Garman-Kohlhagen.'","section":"Section 3.1, line before Eq. (27)"},{"comment":"The notation k'(u) is introduced as a function that tends to zero, but later k' is used as a constant; the notation should be clarified and made consistent.","section":"Section 2.4, Eq. (12)"},{"comment":"The Lagrange multiplier β is written as β(u) in Eq. (16) but treated as a constant thereafter; the paper should state the assumption that β is independent of u or justify why it may be treated as constant.","section":"Section 2.4, Eqs. (16)-(19)"},{"comment":"The backward Kolmogorov equation is stated with the remark 'the derivation is skipped'; for a paper whose goal is to derive standard results from entropic principles, this is a nontrivial step and should be shown or at least referenced.","section":"Section 3.2, Eq. (43)"}],"recommendation":"major_revision","confidential_remarks":"The reference list contains the published version of this manuscript (Ref. [40], Entropy 2019, 21, 586). If this submission is intended as a new submission rather than a copy of the already-published paper, the editor should check for duplicate publication. The companion paper on stocks (Ref. [41]) may also be relevant. The paper's main claimed novelty—deriving a risk-neutral measure—is not supported; the risk-neutral drift is imposed. The editor may wish to consider whether a revised version that presents the model as a maximum-entropy reformulation rather than a derivation meets the novelty bar."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The best thing in this paper is the scale-invariance argument for why the logarithm of the exchange rate is the right variable. That argument is clear, and the max-entropy machinery is applied correctly: given the continuity and drift constraints, the transition density is Gaussian in log-u, the Fokker-Planck equation follows, and the option prices come out as Garman-Kohlhagen with the expected call-put parity check. As a pedagogical re-derivation of a known result, it is coherent and mostly sound. The skipped calculations in Eqs. (20) and (43) are annoying but fillable; the citation pattern is fine, including the reference to the authors' own stock paper.\n\nThe load-bearing problem is the risk-neutral measure. The abstract and introduction say it is derived. In Section 3.1, the paper actually imposes the risk-neutral drift in Eq. (27), replacing the real drift with rd-rf with no Girsanov/Radon-Nikodym computation, no martingale verification, and no argument that this is the minimal-entropy update consistent with no-arbitrage. Without that, Eqs. (37) and (39) are just the standard Garman-Kohlhagen result obtained by assuming lognormal FX and risk-neutral valuation. This is a real gap between claim and content, not a matter of taste. Tonally, the paper should say \"we impose the standard risk-neutral drift\" rather than \"we derive a risk-neutral measure.\"\n\nA second, lesser issue is the uniqueness of the log transformation. The functional equation f(lu)=f(u)+C(l) with C(l)C(l')=C(ll') is satisfied by C(l)=ln l, and that gives f(u)=ln u, but you need a regularity assumption to rule out other solutions. Minor, because any monotone f would do for the rest of the argument.\n\nAlso worth saying plainly: the empirical content is entirely in the constraints. The continuity and drift constraints already encode the GBM drift, so the \"derived\" stochastic process is basically the information put in. That is not a fatal flaw for a foundational exercise, but it limits the significance.\n\nWho should read this? People curious whether entropic inference can reproduce standard finance models, and whoever wants a self-contained example of the max-entropy route to lognormal dynamics. Practitioners will find nothing new. It does deserve a serious referee: the math is mostly right, the goal is meaningful, and the fix is to either supply a genuine derivation of the risk-neutral measure or soften the claims. As it stands, I would reject as is but invite a revision.","headline":"A clean MaxEnt route to Garman-Kohlhagen, but the risk-neutral measure is imposed rather than derived, making the central derivation claim overstated.","tokens_in":11902,"tokens_out":2392,"would_cite":false,"duration_ms":27766,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91G20","62B10","60J65"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives the standard model of foreign-exchange option pricing—geometric Brownian motion for the exchange rate and the Garman–Kohlhagen formula—from maximizing relative entropy under scale, continuity, and drift constraints.","keywords":["Entropic inference","Maximum entropy","Entropic dynamics","Geometric Brownian motion","Fokker–Planck equation","Garman–Kohlhagen model","Black–Scholes–Merton equation","Scale invariance"],"falsifier":"Look at the implied volatility surface of liquid FX options: the model uses a single volatility $\\sigma$ for every strike, so a pronounced smile or skew—systematically different implied volatilities across strikes—would falsify the lognormal transition density at the core of the derivation. Alternatively, high-frequency log-return data with significant excess kurtosis or jumps would violate the continuity constraint.","tokens_in":10939,"feed_emoji":"💱","tokens_out":9275,"duration_ms":91548,"temperature":0.7,"pith_summary":"This paper argues that the standard model of foreign-exchange dynamics—geometric Brownian motion for the exchange rate and the Garman–Kohlhagen formula for European options on it—can be derived, not assumed, from a single inference principle: maximize relative entropy subject to the little information one actually has. Scale invariance of returns forces the logarithm of the exchange rate to be the modeled variable; continuity and a drift constraint fix the transition distribution as lognormal. The same procedure, with the no-arbitrage drift imposed, reproduces the Garman–Kohlhagen option prices and the Black–Scholes–Merton partial differential equation. If correct, the paper offers a derivation of a standard pricing model from information-theoretic first principles rather than from an assumed stochastic process.","feed_headline":"Max-entropy inference reproduces the FX option pricing formula","feed_subtitle":"Scale symmetry selects log rates; no-arbitrage drift yields the standard currency option model.","key_machinery":"The load-bearing object is the maximum-entropy transition density $P(\\ln u'|\\ln u)$: relative entropy is maximized against a Gaussian prior fixed by the continuity constraint $\\langle(\\Delta\\ln u)^2\\rangle=k$, with entropic time $\\alpha=1/(\\sigma^2\\Delta t)$ serving as the clock. A drift constraint $\\langle\\ln(u'/u)\\rangle\\approx(\\mu_d-\\mu_f)\\Delta t-\\frac12\\sigma^2\\Delta t$ fixes the Gaussian's mean, yielding the lognormal transition density; replacing the real drift with the interest-rate differential $(r_d-r_f)\\Delta t$ turns the same density into the risk-neutral measure. The scale-invariance requirement enters by selecting $f(u)=\\ln u$ as the variable whose probability density is a scalar under $u\\mapsto lu$.","core_discovery":"The paper claims that the standard lognormal model of foreign exchange—geometric Brownian motion for the exchange rate and the Garman–Kohlhagen formulas for European calls and puts—is not an independent postulate but follows from maximizing relative entropy under three pieces of information: scale invariance selects the log exchange rate as the variable; continuity constrains the mean-square displacement to be small; and a drift constraint sets the mean log return. With the risk-neutral drift imposed, the same transition density prices European FX options exactly as Garman–Kohlhagen, and the option premium satisfies the Black–Scholes–Merton partial differential equation. The claim is that all of this is derived from entropic inference rather than assumed from stochastic calculus.","pith_inferences":["Because the derivation never uses the level of the exchange rate, only its log, the same argument would carry over to any positive asset whose returns are scale invariant—for example, other currencies or commodity prices—without changing the pricing formula's structure.","The entropy-maximization route suggests a natural diagnostic: if empirical FX returns show excess kurtosis, one should add jump constraints to the entropic model rather than abandon the framework; the scale-invariance requirement then predicts how jump terms must be parameterized.","A complete derivation of the risk-neutral measure from entropy alone would still be missing; the paper imports the standard no-arbitrage drift, so the framework's novel content currently lies in deriving the dynamics, not in deriving the pricing measure."],"forward_implications":["The exchange-rate process is geometric Brownian motion: log returns are Gaussian with mean $(\\mu_d-\\mu_f-\\frac12\\sigma^2)\\Delta t$ and variance $\\sigma^2\\Delta t$, and the transition density is invariant under rescaling the exchange rate.","European call and put prices are exactly the Garman–Kohlhagen formulas, and call–put parity $C-P=e^{-r_f T}u_0-e^{-r_d T}K$ holds as a consequence of the construction.","Option premia satisfy the Black–Scholes–Merton partial differential equation $\\partial_t E+(r_d-r_f)u\\,\\partial_u E+\\frac12\\sigma^2 u^2\\partial_u^2 E-r_d E=0$.","Any extension of the model that includes jumps must preserve scale invariance; otherwise the paper argues an arbitrage opportunity appears."],"supporting_citations":[{"why":"Supplies the axiomatic basis for using maximum relative entropy to update probabilities, which the paper's transition-density assignment relies on.","marker":"[4]"},{"why":"Defines the entropic notion of time used to convert the transition density into a stochastic process with a clock.","marker":"[11]"},{"why":"The original continuous-time model of price dynamics that the paper's GBM derivation extends and rederives.","marker":"[19]"},{"why":"Establishes the geometric Brownian motion description of properly anticipated prices, the target dynamics derived here.","marker":"[20]"},{"why":"The classic option-pricing equations and risk-neutral valuation logic that the paper rederives for FX options.","marker":"[22, 23, 24]"},{"why":"The benchmark FX option premium formula the entropic derivation reproduces.","marker":"[35]"},{"why":"States the no-arbitrage constraints (risk-free drift and discounting) that the paper imposes to obtain the pricing measure.","marker":"[42]"}],"fun_headline_variants":["Entropic inference derives Garman-Kohlhagen FX pricing","Scale symmetry plus entropy yields FX option formula","From entropy to the currency option pricing model","Max-entropy reproduces the FX derivatives formula"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes, rather than derives, that risk-neutral pricing uses the difference between domestic and foreign interest rates as the drift; if that market rule fails, the option formula does not follow from entropy alone.","fun_headline_variants_meta":{"raw":{"variants":["Entropic inference derives Garman-Kohlhagen FX pricing","Scale symmetry plus entropy yields FX option formula","From entropy to the currency option pricing model","Max-entropy reproduces the FX derivatives formula"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000587,"raw_usage":{"total_tokens":2718,"prompt_tokens":867,"completion_tokens":1851,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":483,"completion_tokens_details":{"reasoning_tokens":1791}},"tokens_in":483,"tokens_out":1851,"duration_ms":13097,"temperature":1.0,"reasoning_tokens":1791,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:48:06.307390+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Look at the implied volatility surface of liquid FX options: the model uses a single volatility $\\sigma$ for every strike, so a pronounced smile or skew—systematically different implied volatilities across strikes—would falsify the lognormal transition density at the core of the derivation. Alternatively, high-frequency log-return data with significant excess kurtosis or jumps would violate the continuity constraint.","supporting_citations":[{"cited_title":"Axiomatic derivation of the princ iple of maximum entropy and the principle of minimum cross-entropy","cited_arxiv_id":null,"evidence_quote":"Supplies the axiomatic basis for using maximum relative entropy to update probabilities, which the paper's transition-density assignment relies on."},{"cited_title":"Entropic time","cited_arxiv_id":null,"evidence_quote":"Defines the entropic notion of time used to convert the transition density into a stochastic process with a clock."},{"cited_title":"Th´ eorie de la sp´ eculation","cited_arxiv_id":null,"evidence_quote":"The original continuous-time model of price dynamics that the paper's GBM derivation extends and rederives."},{"cited_title":"Proof that properly anticipated prices fl uctuate randomly","cited_arxiv_id":null,"evidence_quote":"Establishes the geometric Brownian motion description of properly anticipated prices, the target dynamics derived here."},{"cited_title":"Foreign currency options val ues","cited_arxiv_id":null,"evidence_quote":"The benchmark FX option premium formula the entropic derivation reproduces."},{"cited_title":"Options, Futures, and Other Derivatives ; Pearson Education, Inc: India, 2018","cited_arxiv_id":null,"evidence_quote":"States the no-arbitrage constraints (risk-free drift and discounting) that the paper imposes to obtain the pricing measure."}],"review_version":1}