{"id":"b92d9d69-fddb-429e-8d73-73c0f1f5c9a6","arxiv_id":"2601.20670","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In a stochastic autotoxicity model, demographic noise creates a sharp threshold and an excitable-to-persistent transition, with analytic formulas for extinction-pathway probabilities.","lead":"This paper studies how random demographic fluctuations can kill a population right after a rapid boom-and-bust cycle, even when the same model without noise would recover. It introduces 'noise-induced excitability': a probabilistic threshold that decides whether a small population dies immediately, dies after one boom-bust, or persists.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim is secured only for the reduced one-noise model; the toxin-noise reduction is backed by one yeast estimate and no full-model simulation, so the general autotoxic-system claim remains unvalidated.","rationale":"The reader's conditional verdict is appropriate. I agree that the single most load-bearing assumption is the reduction from the two-noise model (Eq. 6) to the one-noise model (Eq. 8), supported only by an order-of-magnitude estimate for yeast. The paper's central phenomenon may well be real for the reduced model, and the analytic probabilities match reduced-model simulations in Fig. 3, but the broader claim about autotoxic systems generally requires that toxin fluctuations be negligible in the extinction-relevant low-population regime. That has not been demonstrated. I considered whether Appendix C's boundary-condition wording—u(1,y)=0 vs. u(x,1)=0—constitutes a fatal internal inconsistency; it appears to be a small-noise asymptotic boundary-layer statement, and the reported agreement with simulations of Eq. 8 suggests the analytic formulas are not grossly wrong. Similarly, the printed minus sign in Eq. 6's x-noise is a clear typo given Eq. 4 and Appendix C. Neither of these displaces the reduction as the key unvalidated step. Since the concern is a missing validation rather than a demonstrated contradiction, I do not recommend rejection; the verdict remains conditional pending a full-model check.","tokens_in":13109,"tokens_out":10663,"duration_ms":119885,"concrete_test":"Simulate the full two-noise SDE Eq. 6 — using the corrected x-noise sqrt{x(1+y)} — with D_x fixed and D_y=αD_x for α in {10^-4, 10^-2, 0.1, 0.3}, ρ in {0.05, 0.1, 0.5}, x0=0.01, y0=0, using the Milstein scheme with Δt=10^-4 and at least 10^4 trajectories. Classify trajectories by winding index and compare p_sl, p_e, p_p against the reduced-model predictions from Eq. 8/Fig. 3. If any α≥0.1 produces deviations exceeding the Monte Carlo error, the toxin-noise reduction is not universally valid and the general autotoxic-system claim must be qualified; if no significant deviation appears, the reduction is robust and the reader's condition is satisfied.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's analytical results (Eqs. 11–12) and all Monte Carlo phase diagrams (Fig. 3) use the reduced model Eq. 8, in which toxin fluctuations are discarded by setting D_y=0. The only quantitative support for this reduction is the yeast estimate in Appendix A, where r~10^13 gives α=sqrt(ρ/r)≲10^-6 for ρ≤4. But the Discussion claims the framework 'can be applied to arbitrary autotoxic systems,' and for many plant or microbial systems r is not astronomically large. The relative importance of D_y near extinction is not governed by the global ratio α alone: as x→0 the x-noise variance vanishes like x, while the y-noise variance remains O(D_y^2 y), so toxin noise can matter in exactly the low-density phase where extirpation is decided. The full two-noise SDE (Eq. 6) and the original individual-based master equation (Eqs. 1–3) are never simulated, so this load-bearing reduction is untested. A separate presentation issue: Eq. 6 prints the x-noise as sqrt{x(1−y)}, which becomes imaginary for y>1; the Fokker–Planck Eq. 4 and Appendix C require sqrt{x(1+y)}, so the typo should be corrected before the full model is simulated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives a mesoscopic stochastic differential equation model for autotoxic population dynamics from an individual-based birth-death process, then reduces it to a one-noise SDE in which toxin fluctuations are neglected. The authors analyze the deterministic skeleton, classify stochastic extinction pathways by a winding-number observable (short-lived, excitable, persistent), and propose parameter-free analytical approximations, Eq. (11), for the probabilities p_sl, p_e, and p_p. They identify a threshold-like 'noise-induced excitability' regime and a sharp transition between excitable and persistent outcomes. The analytical formulas are compared with Monte Carlo simulations of the reduced model and show good agreement in the displayed parameter slices.","tokens_in":13461,"tokens_out":17046,"duration_ms":180398,"significance":"If the central claims hold, the paper provides a useful parameter-free, analytically tractable picture of boom-bust-extirpation in autotoxic populations, with a clean classification of extinction pathways and a sharp excitable-persistent transition. Strengths include the microscopic derivation, the absence of fitted parameters, and the direct Monte Carlo verification of Eq. (11) at the shown slices. However, the paper's general conclusions about 'arbitrary autotoxic systems' rest on a one-noise reduction that is only supported by a single order-of-magnitude estimate, and the Appendix C derivation of the key quantity eta(rho) is not self-consistent as printed. The central idea is promising, but the manuscript needs correction and a broader validity check before the claims can be accepted.","major_comments":[{"comment":"The derivation of Eq. (12) and hence of p_e in Eq. (11) is not internally consistent. The boundary condition stated before Eq. (C1) is u(1,y)=0 for y<1, but the solution (C13) is matched to u(x,1)=0. More seriously, with nu(y) defined by Eq. (C11), nu(1)=0 and nu(y)<0 for y>1, so the claimed solution u=e^{-xi/nu} exceeds 1 and is not a valid absorption probability. Equation (C9) appears to have the wrong sign; reversing it would make nu(y)>0 and the exponential solution bounded. Because Eq. (12) is load-bearing for Eq. (11), the appendix must be corrected. The Monte Carlo agreement in Fig. 3(e-i) suggests the final formula may survive, but the printed derivation does not justify it.","section":"Appendix C, Eqs. (C1)-(C13)"},{"comment":"The reduction from the full two-noise SDE (6) to the reduced model (8) sets D_y=0 on the basis of alpha=sqrt(rho/r)<<1, supported by a single yeast parameter estimate (Appendix A). The full two-noise SDE or the original individual-based master equation (3) is never simulated. This is a load-bearing gap because near x=0, where extirpation is decided, the x-noise amplitude scales as D_x sqrt{x} while the y-noise amplitude is D_y sqrt{x+y}; hence y-noise can matter even when alpha is small at the deterministic fixed point. To support the Discussion's claim that the framework applies to arbitrary autotoxic systems, the authors should either simulate Eq. (6) (after correcting the typo in its x-noise term) or explicitly restrict the claimed domain of validity.","section":"Section II.B and Discussion"}],"minor_comments":[{"comment":"The x-noise is printed as sqrt{x(1-y)}, which becomes imaginary for y>1 and contradicts the Fokker-Planck equation (4) and the reduced model (8), where the term is sqrt{x(1+y)}. Please correct this typo.","section":"Eq. (6)"},{"comment":"The text contains a stray Spanish word, 'absorción', in the definition of theta_W in Eq. (10). Also, in Appendix A, 'to take our selfs an idea' should be 'to get an idea'.","section":"Section III.A"},{"comment":"The definition of y_c as 'the value of y at which the unstable manifold of S_0 cuts for first time x=1 with y>1' is terse; please clarify the geometric construction and state how y_c(rho) is computed in practice.","section":"Eq. (12)"},{"comment":"Panels (e)-(i) are described as 'numerically computed' and 'approximated analytical'; please specify in the caption which panels are Monte Carlo and which are Eq. (11), to avoid ambiguity.","section":"Fig. 3"}],"recommendation":"major_revision","confidential_remarks":"The paper has a genuinely interesting core result and the simulations support the reduced-model predictions. The main concerns are not about novelty but about rigor: the Appendix C derivation is not self-consistent and the reduction to one noise is insufficiently validated. Both are fixable. I would encourage the editor to request a revision rather than reject. I also note that the authors acknowledged generative AI for language editing; the report is based on the scientific content only."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe short version: this paper identifies a real phenomenon—noise-driven threshold behavior in a simple autotoxic population model—and supports it with analytic extinction probabilities that match simulations of the reduced model. It deserves a serious referee, but the general claim is broader than the evidence.\n\nWhat’s genuinely new is the classification of extinction pathways by winding index: short-lived, excitable (one boom-bust then absorption), and persistent (survive the first bust). The probabilities in Eq. (11) are parameter-free approximations, and they track the Monte Carlo results in the parameter slices shown. That’s a real step beyond the standard deterministic boom-bust story, and it’s done without invoking an Allee effect. The deterministic analysis is also clean, and the yeast parameter estimate in Appendix A gives the epsilon scaling some empirical grounding.\n\nThe main soft spot is the reduction to a one-noise model. All analytical and numerical results use Eq. (8), where toxin fluctuations are dropped. The justification is alpha = sqrt(rho/r) << 1, which is plausible for the yeast example but is not a general argument. Near extinction, x->0 and the population noise scales like sqrt(x), so toxin noise can become relatively important exactly in the region where extinction is decided. The full two-noise SDE and the individual-based process are never simulated, so the reduction is untested. This should be checked before the “arbitrary autotoxic systems” claim stands.\n\nThe derivation of p_e in Appendix C has a boundary-condition inconsistency: the stated condition u(1,y)=0 for y<1 is replaced by u(x,1)=0 when the final solution is written as e^{-xi/nu}. This makes the derivation a heuristic matched-asymptotics argument rather than a rigorous one. The fact that it matches simulations is encouraging, but the paper should be honest about the status.\n\nMinor issues: Eq. (6) has the x-noise as sqrt{x(1−y)}, which is imaginary for y>1; it should be sqrt{x(1+y)}. The paper also doesn’t ship code or data.\n\nOverall: the core idea is worth taking seriously, the math is mostly sound, and the flaws are fixable. I’d send it to a good stochastic-dynamics referee and ask specifically for a test of the one-noise reduction and a cleaned-up Appendix C.","headline":"A solid stochastic-dynamics paper with a genuinely new extinction-pathway classification, but the central mechanism is only verified for a reduced one-noise model and the analytic p_e derivation has a boundary-condition blemish.","tokens_in":13860,"tokens_out":4220,"would_cite":false,"duration_ms":45154,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J70","92D25","60H10"],"pacs":[],"model":"deepseek-v4-flash","headline":"By adding demographic noise to a minimal autotoxicity model, the paper shows that boom-bust dynamics become a probabilistic threshold phenomenon, with analytic formulas for the probabilities of early extinction, excitable busts, and persist","keywords":["autotoxicity","demographic noise","boom-bust dynamics","extinction pathways","excitable systems","absorbing state","stochastic differential equations","first-passage probabilities"],"falsifier":"Simulate the full two-noise model (Eq. 6) or the individual-based process with realistic parameters, and compare the extinction-pathway probabilities against Eq. (11); any significant deviation in the excitable regime would falsify the noise-reduction. Alternatively, measure extinction-time distributions in a well-mixed autotoxic culture and check for the predicted exponential forms and ρ-scaling.","tokens_in":13016,"feed_emoji":"🦠","tokens_out":7444,"duration_ms":74652,"temperature":0.7,"pith_summary":"The paper proposes that the irreversible collapse of an autotoxic population—one that poisons its own habitat—can be explained without invoking any positive feedback such as an Allee effect. Starting from an individual-based model, the authors derive a two-variable stochastic differential equation for population and toxin densities, then reduce it to one noise source by arguing that toxin fluctuations are negligible because toxin molecules vastly outnumber individuals. In the reduced model they identify three distinct extinction pathways—short-lived, excitable, and persistent—and find sharp transitions between them as the noise strength D and the toxin-population timescale ratio ρ vary. The central quantitative contribution is a set of explicit formulas for the probabilities of each pathway, with a deterministic 'barrier height' η(ρ) that captures how close a deterministic boom-bust trajectory comes to extinction. The authors argue this provides a general route to understanding boom-bust-extirpation in invasive species, microbial fermentation, and other negatively regulated populations.","feed_headline":"Noise alone can doom a boom-bust population","feed_subtitle":"Analytic formulas predict whether a collapsing population goes extinct at the first bust or persists.","key_machinery":"The reduced model (Eq. 8) carries the argument: only population density x carries noise; toxin density y follows deterministic dynamics ρ(x-y), justified by the diffusion ratio α=√(ρ/r) with r≈10^13 for yeast. Trajectories are classified by winding number N_W around the coexistence point into three pathways. Two asymptotic limits yield analytic probabilities: ρ→0 gives p_sl = e^{-2x0/D^2}; small D gives p_e via an incomplete-gamma barrier height η(ρ). The exponential e^{-2·scale/D^2} converts the competition between noise scale and deterministic distances into sharp transitions.","core_discovery":"Central claim: noise-induced excitability, a regime where demographic noise replaces the fixed threshold of deterministic excitable systems, so identical initial conditions can lead either to a full boom followed by absorption in the first bust, or to a metastable persistent state. The three pathway probabilities are p_sl = exp(-2x0/D^2) (short-lived), p_e = (1-p_sl) exp(-2η(ρ)/D^2) (excitable), p_p = 1-p_sl-p_e (persistent), with η(ρ) a deterministic barrier height derived from the toxin-free flow. The first transition depends on x0/D^2 and not ρ; the second on η(ρ)/D^2 and not the initial condition. For any nonzero noise, small ρ forces the excitable regime.","pith_inferences":["Inference: The one-noise reduction is load-bearing; the paper does not simulate the full two-noise model (Eq. 6) or the individual-based process, so a direct numerical check is the natural next step.","Inference: The winding-number classification gives a practical time-series observable—counting rotations around the coexistence state—that could map empirical population data onto the p_sl/p_e/p_p diagram.","Inference: The exponential structure likely generalizes to other slow-manifold first-passage problems, with η(ρ) replaced by the analogous deterministic closest-approach distance for the relevant feedback.","Inference: The sharp transition suggests a control principle: increasing habitat volume or toxin degradation rate can move a population from the excitable to the persistent regime, informing both conservation and fermentation management."],"forward_implications":["Small populations (large D) are most likely extirpated before the first boom, so system size alone can determine whether an introduced species ever expands.","With slow toxin dynamics (small ρ), any nonzero noise makes the excitable regime dominant, so a population that survives the first boom is very likely to die in the first bust.","The excitable-persistent transition is sharp, so small changes in carrying capacity or toxin decay rate can flip a population's statistical fate from persistence to extinction.","No Allee effect or positive feedback is required; negative feedback plus demographic stochasticity suffice for boom-bust-extirpation.","The same probabilistic structure should transfer to other negative-feedback mechanisms (resource depletion, pathogen load, predation) when timescale and noise scales are matched."],"fun_headline_variants":["Noise flips the switch in boom-bust extinctions","Same start, different fate: noise rules boom-bust","Noise alone determines if a population booms or dies","In autotoxic systems, noise sets the survival threshold","Boom, bust, or gone: noise decides the path"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The reduction to a single-noise model assumes toxin fluctuations are negligible (α≪1); if the diffusion ratio is not small for a given system, the predicted probabilities and the excitable-persistent transition do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Noise flips the switch in boom-bust extinctions","Same start, different fate: noise rules boom-bust","Noise alone determines if a population booms or dies","In autotoxic systems, noise sets the survival threshold","Boom, bust, or gone: noise decides the path"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000191,"raw_usage":{"total_tokens":1179,"prompt_tokens":741,"completion_tokens":438,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":485,"completion_tokens_details":{"reasoning_tokens":356}},"tokens_in":485,"tokens_out":438,"duration_ms":5073,"temperature":1.0,"reasoning_tokens":356,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T07:16:10.593832+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the full two-noise model (Eq. 6) or the individual-based process with realistic parameters, and compare the extinction-pathway probabilities against Eq. (11); any significant deviation in the excitable regime would falsify the noise-reduction. Alternatively, measure extinction-time distributions in a well-mixed autotoxic culture and check for the predicted exponential forms and ρ-scaling.","supporting_citations":[],"review_version":1}