{"id":"b4d434aa-9e66-49fc-9e4e-82d9c1fa8c54","arxiv_id":"1908.09446","paper_version":5,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proposes a Fisher-information-based stability criterion for master equations, but the central example is undermined by an algebraic sign error.","lead":"This paper links the classical Glansdorff-Prigogine stability criterion for Markov processes to the Fisher metric of information geometry and proposes a new stability criterion based on the rate of change of the Fisher information speed. The proposed criterion is meant to work for nonlinear master equations where the classical criterion is inconclusive, using an autocatalytic reaction as an example.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The autocatalytic example's claimed instability region comes from a sign error in the linearized excess flux (Eq. 45); with the correct coefficient, the Glansdorff-Prigogine criterion predicts stability for all parameter values, so the paper's demonstration of its generalized criterion collapses.","rationale":"Reading the paper in good faith, I acknowledge a correct and useful result for linear master equations: the identity δ²σ = -1/2 d/dt Σ_i (δp_i)^2/p̄_i (Eqs. 15-17) connects the excess entropy production to the Fisher line element and gives a Lyapunov-style reformulation. I do not object to that part. The central new claim, however, is the generalized information-geometric stability criterion for nonlinear master equations, and its only worked demonstration is the autocatalytic reaction. That demonstration contains a concrete sign error in the linearization of the excess flux. Correcting the sign removes the claimed instability for K_+/K_- < 3 and makes the Glansdorff-Prigogine criterion predict stability for all positive rate constants, so the example no longer shows any advantage of the new criterion. The paper's hedge citing Ref. [8]—that the Glansdorff-Prigogine criterion is sufficient but not necessary—does not repair the algebra, because the alleged counterexample was manufactured by the sign error. In addition, the generalized criterion d/dt[(ds/dt)^2] ≤ 0 is asserted as a stability condition rather than proved equivalent to Lyapunov stability for the nonlinear case; the Cramér-Rao inequality (27) bounds a ratio but does not by itself establish stability. However, the sign error alone is decisive for the central demonstration. The reader's REJECT verdict is therefore appropriate, and my stress-test does not change it.","tokens_in":10881,"tokens_out":5950,"duration_ms":54772,"concrete_test":"Recompute the linearized excess flux for the autocatalytic model by differentiating J(p_X) = K_+ p_X(1-p_X) - K_- p_X^2 at p̄_X = K_+/(K_+ + K_-) and evaluate δ²σ = δJ δF for, say, K_+/K_- = 2 and K_+/K_- = 12. If the calculation gives δJ = -K_+ δp_X and δ²σ > 0 in both cases, rather than changing sign at K_+/K_- = 3, the claimed instability in Eq. (47) is confirmed to be a sign-error artifact. A complementary check is to numerically integrate the deterministic rate equation d p_X/dt = K_+ p_X(1-p_X) - K_- p_X^2 from p_X = 0.001 with K_+/K_- = 2 and confirm monotone relaxation to p̄_X, contradicting the paper's predicted instability region.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central demonstration that the Glansdorff-Prigogine criterion 'does not work well' for the nonlinear master equation rests entirely on the autocatalytic example and the claim that δ²σ < 0 for K_+/K_- < 3 (Eq. 47). This claim is generated by an algebraic sign error in Eq. (45). From Eqs. (42)-(43), J_{Y→X} = K_+ p_X(1-p_X) - K_- p_X^2. Linearizing about p̄_X = K_+/(K_+ + K_-) gives δJ_{Y→X} = [K_+(1-2p̄_X) - 2K_-p̄_X]δp_X = -K_+ δp_X, not [K_+(1-2p̄_X) + 2K_-p̄_X]δp_X. The published expression is the derivative of K_+p(1-p) + K_-p^2, i.e. it has the wrong sign for the reverse reaction. Since δF_{Y→X} is negative (Eq. 46), the corrected excess flux gives δ²σ = K_+[1/(1-p̄_X) + 1/p̄_X](δp_X)^2 > 0 for all K_± > 0. Thus the predicted unstable region is an artifact; the example provides no case where the classical criterion fails, and no valid nonlinear illustration of the proposed criterion remains. The linear-master-equation identity (Eqs. 15-17) appears correct, but it does not establish the generalized criterion, whose status as a Lyapunov condition is asserted rather than derived. The paper's own caution that the Glansdorff-Prigogine criterion is only sufficient (Ref. [8]) cannot repair this error.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims three main results: (i) for linear master equations, the excess entropy production is exactly the negative time derivative of the Fisher-information line element, giving a Lyapunov interpretation of the Glansdorff-Prigogine criterion; (ii) a generalized stability criterion based on the monotonicity of the Fisher information speed, d/dt[(ds/dt)^2] ≤ 0, which is claimed to work for nonlinear master equations where the classical criterion allegedly fails; and (iii) trade-off relations, including a Cramér-Rao-type bound and an Onsager-coefficient interpretation of the Fisher metric. The advertised nonlinear illustration is an autocatalytic reaction X+Y ⇋ 2X, for which the paper claims the Glansdorff-Prigogine criterion predicts instability for K_+/K_- < 3 while numerical simulations show no such transition.","tokens_in":11194,"tokens_out":2443,"duration_ms":23563,"significance":"The linear-master-equation identity in Eqs. (15)-(17), relating the excess entropy production to -d/dt of the weighted squared deviation, is a clean and potentially useful result, and the connection drawn between the Fisher metric and the Onsager coefficient in the near-equilibrium regime is suggestive. If the generalized criterion were rigorously established, it would provide an information-geometric Lyapunov condition for a broader class of stochastic dynamics. However, the central nonlinear claim is not proven and the sole nonlinear example contains an algebraic error that generates the claimed instability region. As a result, the paper's main advertised contribution, the generalization of the Glansdorff-Prigogine criterion to nonlinear master equations, is unsupported.","major_comments":[{"comment":"The proposed criterion d/dt[(ds/dt)^2] ≤ 0 for stability and d/dt[(ds/dt)^2] > 0 for instability is asserted rather than derived. A valid Lyapunov condition requires not only that the candidate function is nonnegative and vanishes at the steady state, but that it decreases along trajectories and that its decrease implies convergence. The Fisher information speed (ds/dt)^2 is a nonnegative quantity that vanishes at the steady state, but no theorem is given showing that its time derivative controls approach to the steady state for nonlinear master equations. The paper's own monotonicity result, Eq. (25), is quoted for time-independent transition rates, which does not cover the nonlinear case in which the transition rates depend on the probabilities themselves. This is the load-bearing step for the claimed generalization.","section":"A generalization of the Glansdorff-Prigogine criterion for stability, Eq. (28)"},{"comment":"The calculation of the excess flux contains a sign error. From Eqs. (42)-(43), J_{Y→X} = K_+ p_X(1-p_X) - K_- p_X^2, and linearizing about p̄_X = K_+/(K_+ + K_-) gives δJ_{Y→X} = [K_+(1-2p̄_X) - 2K_- p̄_X] δp_X = -K_+ δp_X. The published expression [K_+(1-2p̄_X) + 2K_- p̄_X] is the derivative of K_+ p(1-p) + K_- p^2, with the wrong sign for the reverse reaction term. Since δF_{Y→X} is negative, the corrected δ²σ is K_+ [1/(1-p̄_X) + 1/p̄_X] (δp_X)^2 > 0 for all K_± > 0. Thus the predicted instability region K_+/K_- < 3 is an artifact of the sign error, and the claim that the Glansdorff-Prigogine criterion fails for this model is unsupported. This removes the only nonlinear demonstration of the generalized criterion.","section":"Example: Autocatalytic reaction, Eq. (45)"},{"comment":"The linear derivation is internally consistent, but the claimed equivalence between the generalized criterion and the Glansdorff-Prigogine criterion around steady states is not established for the nonlinear case. The text states that Eq. (25) 'can be regarded as' the Lyapunov stability condition and that for the linear master equation the generalized criterion is equivalent to the Glansdorff-Prigogine criterion; however, the monotonicity of (ds/dt)^2 for linear master equations is a known result, while the same monotonicity does not follow for the nonlinear master equation. Without a proof that d/dt[(ds/dt)^2] is a Lyapunov function for the nonlinear dynamics, Eq. (28) remains a plausible but unproven ansatz.","section":"The linear master equation and the Lyapunov stability, Eqs. (15)-(17)"}],"minor_comments":[{"comment":"The identification δ²L = (1/2)ds² relies on treating δp_i as an infinitesimal dp_i; the notation 'around the steady state p ≃ p̄ and δp_i = dp' is ambiguous and should be stated as a first-order identification with a controlled remainder.","section":"The Lyapunov function and information geometry, Eq. (24)"},{"comment":"There are several typographical issues, including 'Glandorff' for 'Glansdorff', 'Form the expression' for 'From the expression', and 'relasionship' in the Supplementary Information.","section":"Throughout"},{"comment":"In Eq. (27), η_R = var[R] / (dE[R]/dt)^2 is bounded below by η* = 1/(ds/dt)^2; the notation η_R ≥ η* is dimensionally and logically consistent only if the observables are dimensionless, which is not stated.","section":"A generalization of the Glansdorff-Prigogine criterion for stability, Eq. (27)"},{"comment":"The numerical simulations displayed in Fig. 3 are described qualitatively but no parameters, integration method, or error estimates are given; this makes it difficult to verify the claimed absence of a transition at K_+/K_- = 3.","section":"Example: Autocatalytic reaction, Fig. 3"}],"recommendation":"reject","confidential_remarks":"The central advertised result depends on an algebraic error in the only nonlinear example, and the generalized criterion is not proven. The linear-master-equation identity is a fine result, but it does not support the paper's main claim as written. Given that the error is load-bearing and the example cannot be repaired within the manuscript's current scope, I recommend rejection rather than major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe useful part of this paper is the linear-master-equation identity: for rates independent of p, δ²σ = −½ d/dt Σ (δp_i)²/p̄_i, so the Glansdorff-Prigogine excess entropy production is a Lyapunov derivative. That is correct and cleanly ties the Fisher line element to the Onsager coefficient near equilibrium. If the paper only claimed that, it would be a fine short letter.\n\nBut the paper's main claim—that its information-geometric criterion works for nonlinear master equations where GP \"does not work well\"—rests on the autocatalytic example, and that example contains a sign error. The excess flux in Eq. (45) is written as [K₊(1−2p̄)+2K₋p̄]δp, but the flux J = K₊p(1−p) − K₋p² has derivative K₊(1−2p̄) − 2K₋p̄ = −K₊. The plus sign before 2K₋ is wrong. With the correct sign, δ²σ = δJ·δF is always positive, so the GP criterion is stable for all rate constants; the claimed unstable region 3 > K₊/K₋ is an artifact. The paper even cites Ref. [8] to excuse the discrepancy, but the discrepancy is manufactured by the algebra.\n\nThe generalization itself is asserted rather than derived. The condition d/dt[(ds/dt)²] ≤ 0 is identified with stability, and the Cramér–Rao bound is invoked for interpretation, but no proof links this to a Lyapunov function for the nonlinear steady state. The linear identity does not establish the generalized criterion; it only gives an equivalent viewpoint in the linear case, where the paper claims (without proof) the two criteria coincide around the steady state.\n\nSo: the identity is worth knowing, and the geometric picture is appealing, but the central nonlinear claim has no valid demonstration. The sign error is not a minor typo; it is the load-bearing example. I would not cite the generalized criterion as is. The paper is from a serious researcher and the linear part is accurate, so it deserves a careful referee rather than a desk reject—but that referee should reject the current version unless the example is corrected and the generalization actually proved.\n\nRecommendation: send to peer review with the expectation of major revision, or reject. Not publishable in current form.","headline":"The linear-master-equation identity is correct and elegant, but the autocatalytic example's sign error invalidates the central claim about nonlinear stability.","tokens_in":11738,"tokens_out":2222,"would_cite":false,"duration_ms":22249,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The excess entropy production of a linear master equation equals the time derivative of a Fisher-metric line element, making the Glansdorff-Prigogine stability criterion a Lyapunov condition.","keywords":["information geometry","Glansdorff-Prigogine criterion","excess entropy production","Fisher metric","Cramér-Rao inequality","Lyapunov stability","master equation","autocatalytic reaction"],"falsifier":"Direct numerical integration of a nonlinear master equation with a known unstable steady state, such as the autocatalytic reaction near $K_+/K_-$ values where the classical criterion is ambiguous, should show $d[(ds/dt)^2]/dt>0$ while the system is near the unstable state and $\\le 0$ as it settles into the stable one; a violation of this sign pattern would falsify the generalised criterion.","tokens_in":10608,"feed_emoji":"📉","tokens_out":9646,"duration_ms":86067,"temperature":0.7,"pith_summary":"This paper sets out to show that stability of a steady state in a Markov process is the same thing as the slowing down of the probability distribution's motion in the space of probability distributions. For linear master equations, it proves an exact identity: the excess entropy production is $-1/2\\, d/dt \\sum_i (\\delta p_i)^2/\\bar p_i$, so the classical Glansdorff-Prigogine criterion becomes a Lyapunov stability condition. It then generalises the criterion to nonlinear master equations, where the classical criterion gives an elusive answer, using the time derivative of the Fisher-information speed and the Cramér-Rao inequality. The paper matters because it connects a classical thermodynamic stability test to information geometry and claims a simple 'speed-decay' condition diagnoses stability in situations where the old criterion does not.","feed_headline":"Excess entropy production is a Fisher-metric time derivative","feed_subtitle":"For linear master equations, this makes the Glansdorff-Prigogine stability criterion a Lyapunov test.","key_machinery":"The carrying object is the Fisher metric on the probability simplex, whose line element is $ds^2 = \\sum_i (dp_i)^2/p_i$, together with the Lyapunov function $\\delta^2 L = \\frac{1}{2}\\sum_i (\\delta p_i)^2/\\bar p_i$, which around a steady state is half that line element. The mechanism that carries the argument is the identity $d\\delta^2 L/dt = -\\delta^2\\sigma$, converting the classical thermodynamic stability criterion into decay of an information-geometric distance. The generalisation replaces the excess entropy production by the time derivative of the Fisher-information speed $(ds/dt)^2$ and reads stability from the sign of $d[(ds/dt)^2]/dt$, with the Cramér-Rao inequality supplying the physical bound in terms of fluctuations and responses of observables.","core_discovery":"The central discovery is that the excess entropy production of a linear master equation is minus the time derivative of half the Fisher line element: $\\delta^2\\sigma = -\\frac{1}{2}\\frac{d}{dt}\\sum_i (\\delta p_i)^2/\\bar p_i = -\\frac{d}{dt}\\delta^2 L$. Therefore the Glansdorff-Prigogine condition $\\delta^2\\sigma\\ge 0$ is equivalent to the Lyapunov condition $d\\delta^2 L/dt\\le 0$, with the Lyapunov function itself an information-geometric quantity. Around a steady state the paper identifies this with monotone decrease of the Fisher information $(ds/dt)^2$. For nonlinear master equations, where the classical argument breaks down, the paper proposes the generalised criterion $d[(ds/dt)^2]/dt\\le 0$ for stability and $>0$ for instability, interprets it through the Cramér-Rao bound on fluctuation-response ratios, and demonstrates it on an autocatalytic reaction for which the classical criterion is inconclusive.","pith_inferences":["Editorial inference: the Fisher-information speed is defined for any differentiable probability path, so the same sign test could be applied to non-Markovian or explicitly time-dependent processes, a regime the paper does not treat.","Editorial inference: the identity between excess entropy production and the Fisher-metric derivative suggests that, in the linear regime, the Onsager matrix and the Fisher metric may be the same geometric object; if so, transport coefficients could be estimated directly from trajectory-level probability fluctuations.","Editorial inference: the criterion may certify stability of individual steady states rather than global contraction of the whole probability simplex, so a natural extension is to check whether $d[(ds/dt)^2]/dt\\le 0$ holds on entire basins of attraction.","Editorial inference: the paper's framing suggests reading thermodynamic uncertainty relations as stability certificates, where growth of the fluctuation-to-response ratio marks approach to a steady state; this goes beyond the paper's own narrower connection."],"forward_implications":["For linear master equations, the Glansdorff-Prigogine criterion is a theorem about monotone decay of a Fisher-metric distance, not just a heuristic stability test.","Near equilibrium, the Onsager coefficients can be expressed through the Fisher metric in flux coordinates, so transport coefficients acquire an information-geometric reading.","The generalised criterion $d[(ds/dt)^2]/dt\\le 0$ extends stability analysis to nonlinear master equations, where the classical excess-entropy condition is not reliable.","The Cramér-Rao inequality gives a fluctuation-response ratio whose time dependence signals stabilisation or destabilisation, connecting thermodynamic uncertainty relations to stability.","In the autocatalytic reaction, the criterion assigns instability to the zero-concentration steady state and stability to the nonzero steady state, resolving the ambiguous classical prediction."],"supporting_citations":[{"why":"introduces the general evolution criterion that the paper reinterprets as a Lyapunov condition.","marker":"[1]"},{"why":"states the excess-entropy-production stability criterion that the paper shows is equivalent to Fisher-information decay in the linear case.","marker":"[2]"},{"why":"raises the qualms about the range of validity of the Glansdorff-Prigogine criterion that motivate the nonlinear generalisation.","marker":"[5]"},{"why":"analyses what the Glansdorff-Prigogine criterion can and cannot state for mass-action kinetics, the background for the autocatalytic example.","marker":"[8]"},{"why":"supplies the network-theory cycle-flux and cycle-force framework used to define Onsager coefficients in flux coordinates.","marker":"[10]"},{"why":"provides the information-geometry definitions of the Fisher metric and the Cramér-Rao inequality used in the criterion.","marker":"[37]"},{"why":"gives the preceding thermodynamic interpretation of the information-geometric line element that the excess-entropy identity extends.","marker":"[48]"},{"why":"proves the monotone decrease of the Fisher information and connects the Cramér-Rao bound to thermodynamics, the basis for the generalised criterion.","marker":"[49]"}],"fun_headline_variants":["Excess entropy is Fisher-metric derivative, Lyapunov test","New stability criterion for nonlinear master equations","Information geometry yields trade-off relations and stability","Fisher metric explains Glansdorff-Prigogine stability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the time derivative of the Fisher-information speed is a general stability indicator: that $d[(ds/dt)^2]/dt\\le 0$ marks approach to a stable steady state and $>0$ marks departure from it, in nonlinear as well as linear master equations.","fun_headline_variants_meta":{"raw":{"variants":["Excess entropy is Fisher-metric derivative, Lyapunov test","New stability criterion for nonlinear master equations","Information geometry yields trade-off relations and stability","Fisher metric explains Glansdorff-Prigogine stability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000221,"raw_usage":{"total_tokens":1447,"prompt_tokens":938,"completion_tokens":509,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":554,"completion_tokens_details":{"reasoning_tokens":446}},"tokens_in":554,"tokens_out":509,"duration_ms":5752,"temperature":1.0,"reasoning_tokens":446,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:12:19.166862+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Direct numerical integration of a nonlinear master equation with a known unstable steady state, such as the autocatalytic reaction near $K_+/K_-$ values where the classical criterion is ambiguous, should show $d[(ds/dt)^2]/dt>0$ while the system is near the unstable state and $\\le 0$ as it settles into the stable one; a violation of this sign pattern would falsify the generalised criterion.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduces the general evolution criterion that the paper reinterprets as a Lyapunov condition."},{"cited_title":"& Prigogine, I","cited_arxiv_id":null,"evidence_quote":"states the excess-entropy-production stability criterion that the paper shows is equivalent to Fisher-information decay in the linear case."},{"cited_title":"Remarks on Glansdorff and Prigogine’s Theory of Stability","cited_arxiv_id":null,"evidence_quote":"raises the qualms about the range of validity of the Glansdorff-Prigogine criterion that motivate the nonlinear generalisation."},{"cited_title":"The Glansdorff-Prigogine thermodynamic stability criterion in the light of Lyapunovs theory","cited_arxiv_id":null,"evidence_quote":"analyses what the Glansdorff-Prigogine criterion can and cannot state for mass-action kinetics, the background for the autocatalytic example."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the network-theory cycle-flux and cycle-force framework used to define Onsager coefficients in flux coordinates."},{"cited_title":"E., & Vos, P","cited_arxiv_id":null,"evidence_quote":"provides the information-geometry definitions of the Fisher metric and the Cramér-Rao inequality used in the criterion."},{"cited_title":"M., Crooks, G","cited_arxiv_id":null,"evidence_quote":"gives the preceding thermodynamic interpretation of the information-geometric line element that the excess-entropy identity extends."},{"cited_title":"Stochastic thermodynamic interpretation of infor- mation geometry","cited_arxiv_id":null,"evidence_quote":"proves the monotone decrease of the Fisher information and connects the Cramér-Rao bound to thermodynamics, the basis for the generalised criterion."}],"review_version":1}