{"id":"cd39f30d-b48b-4e95-9514-6e7e4ef8f478","arxiv_id":"2607.19084","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A countable-state Markov chain derived from an open Pelikan map is shown to support a continuum of quasi-stationary distributions, each with a distinct escape rate.","lead":"This paper constructs a simple infinite-state random walk with killing that has infinitely many different quasi-stationary distributions at once, each decaying at its own rate. It provides a concrete example of a phenomenon previously known only in abstract Markov-chain theory, in the setting of an open Pelikan random map.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed continuum of QSDs rests on unproven positivity of μ_k(α) over an α-interval; Sec. 3.2 only numerically observes it, leaving the central claim without proof.","rationale":"After reading the manuscript, I agree with the reader that the central claim's weak point is the positivity of the μ_k family. The derivation of A(z) (32) and the eigenvector equation are sound; the eigenvalue formula λ(α) = 1−pεα follows directly. What turns these formal solutions into QSDs is the requirement μ_k ≥ 0 for all k. The paper never proves this for a continuum of α; Sec. 3.2 explicitly relies on numerical iteration up to a computed threshold, and the text admits that validity for α ≤ α_u is 'observed numerically.' The tail criterion (48) is a useful reduction, but its constituents (minimal root, coefficient C_1, and k*) are only computed numerically for discrete α values. A small dip of μ_k below zero in an unexamined α-interval would break the continuum into isolated QSDs, so the headline result is not yet demonstrated. The simulations in Sec. 3.3 and Sec. 5 provide supporting evidence but do not fill the proof gap. The QED uniqueness in Sec. 6 is also asserted via a 'technical reasons' footnote rather than proved, but that is secondary to the QSD continuum claim. The appropriate verdict is conditional: the construction is plausible and likely correct, but a rigorous positivity proof or a certified computational verification is needed.","tokens_in":17587,"tokens_out":6525,"duration_ms":62571,"concrete_test":"Take a fixed generic pair, e.g. (p, ε) = (0.8, 0.2), and implement the recurrence (42)-(44) with rigorous interval arithmetic, treating α as an interval variable over [0, α_u]. Compute μ_0(α), …, μ_{k*}(α), where k*(α) is determined from criterion (48) using interval bounds on the dominant root and coefficient C_1. Verify that every interval lies strictly inside (0, ∞), so that by (48) all subsequent μ_k are positive. If this succeeds, the continuum of QSDs is certified for that parameter pair. If any interval contains zero or negative values, locate the first k and α where positivity fails, which would falsify the continuum claim. Repeat the same verification for the R-transient point (p, ε) = (0.4, 0.2) to probe genericity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that for generic (p, ε) the substochastic chain (9) supports a continuum of QSDs parameterized by α—requires that the formal solutions μ_k(α) of recurrence (42)-(44) be non-negative for all k and for all α in an interval, not just at isolated α. The paper's only support is numerical iteration and the observation 'we rather observe numerically that they are valid for α≤α_u and invalid otherwise' (Sec. 3.2, Fig. 3). The finite-time criterion (48) reduces the check to k≤k*, but k* and the coefficients C_i are computed numerically, and the check is done on a grid of α, not continuously. If there is any α-interval below α_u where μ_k dips negative for some k, the infinite family collapses to a discrete set of QSDs, invalidating the paper's main conclusion. The authors themselves note (Sec. 3.2) that positivity is 'very difficult to determine by inspection of the generating function alone' and rely on a threshold α_u. This is not an external consensus dispute; it is an internal gap between the formal eigenfunction solution and the probabilistic requirement μ≥0. The claim of a 'continuous spectrum' is thus not established by the manuscript as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a countable-state substochastic Markov chain, motivated by an open variant of the Pelikan random map, with transition operator given by Eq. (9). The authors derive a one-parameter family of formal eigenmeasures of the transfer operator, whose generating function is given by Eq. (32) and whose eigenvalue is λ(α)=1−pεα. They claim that, for generic parameter values, there is a continuum of positive normalizable quasi-stationary distributions (QSDs), each with its own escape rate, and they test this claim in simulations. The paper also analyzes the asymptotic shape of these QSDs, studies their stability under random perturbations, and constructs quasi-ergodic distributions (QEDs) for the principal eigenvalue in the R-positive region.","tokens_in":17991,"tokens_out":17981,"duration_ms":154226,"significance":"If the main existence claim were fully proved, this would be a valuable and explicit example of non-unique QSDs and initial-condition-dependent escape rates in a simple random dynamical system, with implications for metastability, stochastic resetting, and open dynamical systems. The paper contains several strengths: a clean generating-function derivation, an explicit recurrence for the coefficients, an independent derivation of the principal escape rate from singularity analysis, and extensive numerical simulations with large ensembles. The shape-conservation argument in Sec. 4.2 is also suggestive. However, the central theorem is not actually proved: the existence of a continuum of QSDs rests on a numerical positivity check, and the claimed uniqueness of QEDs rests on an unproved assertion.","major_comments":[{"comment":"The central claim—that for generic (p, ε) there is a continuum of QSDs parameterized by α∈(0, α_u]—requires proving that the coefficients μ_k(α) defined by the recurrence (44) are non-negative for every k and for every α in an interval. The manuscript verifies this only numerically: Fig. 3 is generated by iterating (44) until either μ_k<0 or k>k* from criterion (48), with the coefficients C_i and threshold k* computed numerically on a grid of α. The statement 'we rather observe numerically that they are valid for α≤α_u and invalid otherwise' is an empirical observation, not a proof. If some subinterval of [0, α_u] had a negative μ_k, the infinite family would collapse to isolated QSDs. An analytic sign analysis of the three exponential contributions in (45), or a rigorous dominance argument for the coefficients, is required to establish the paper's main conclusion.","section":"Sec. 3.2, Eqs. (44)–(48)"},{"comment":"The abstract and conclusion claim that unique QEDs are established. However, the uniqueness assertion is supported only by 'for technical reasons we believe this is in fact the only eigenvalue which produces a normalisable QED.' This is not a proof. To claim uniqueness, the authors must show that for every eigenvalue λ≠λ_u of the transfer operator, the product b_i(λ) μ_i(λ) is not summable, or that no corresponding positive left eigenvector exists. The cancellation argument in footnote 5 is local to λ_u and does not exclude other eigenvalues. Without a rigorous argument, the uniqueness claim should be downgraded to a conjecture.","section":"Sec. 6, Eq. (62) and footnote 5"}],"minor_comments":[{"comment":"In the recurrence (42), the value p_1 = α(2−ε)/2 appears to have the wrong sign. The numerator in (41) is α(1 − z(2−ε)/2), so p_1 should be −α(2−ε)/2. The explicit expression for μ_1 in (43) is consistent with the negative sign. Please correct.","section":"Eq. (42)"},{"comment":"The phrase 'enables us to rigorously determine the non-negativity of μ numerically in finite time' overstates what is done: the coefficients C_i in (48) and the threshold k* are computed numerically, and the α-axis is sampled on a grid. The procedure gives strong numerical evidence, but the word 'rigorously' should be replaced by a more qualified statement.","section":"Sec. 3.2, paragraph after Eq. (48)"},{"comment":"The phrase 'continuous spectrum' is likely to be confused with the spectral-theory notion of continuous spectrum. Since the paper establishes (or aims to establish) a continuum of eigenvalues, 'continuum of QSDs' would be less ambiguous.","section":"Abstract and Sec. 1.2"},{"comment":"The stability conclusions are based on numerical experiments at five parameter points and a few noise models. The paper does state this, but the concluding sentence in Sec. 7 ('attested analytically and in simulations') goes beyond the analytic content, which is limited to the shape-conservation argument of Sec. 4.2. Please calibrate the wording.","section":"Sec. 5.1"}],"recommendation":"major_revision","confidential_remarks":"The main concern is the gap between numerical verification and the claimed theorem. If the authors can supply a rigorous proof of positivity for the recurrence coefficients over an α-interval, and either prove or clearly label as a conjecture the uniqueness assertion for QEDs, the paper would be a strong contribution. As it stands, the central existence claim is not established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is worth reading but not worth believing wholesale. It gives an explicit random-walk-with-resetting chain on the half-line, motivated by the open Pelikan map, and shows that for fixed parameters there is a one-parameter family of quasi-stationary distributions, each with its own escape rate. That is genuinely new. Seneta–Vere-Jones allows multiple QSDs in principle, but an explicit, solvable example is scarce.\n\nThe generating-function analysis is clean and mostly correct. The eigenvalue formula λ(α) = 1 − pεα falls out of the master equation, not from fitting. The principal rate λ_u = 1/R is derived independently from singularity analysis, and the numerics match. The recurrence for μ_k is explicit and reproducible. The QED construction in the R-positive regime is a nice bonus, and the simulations support it.\n\nThe load-bearing gap: the continuum of QSDs requires μ_k(α) ≥ 0 for all k on an interval of α. The paper checks positivity by iterating the recurrence up to a k* determined numerically (criterion (48)) and asserts validity for α ≤ α_u. That is numerical evidence, not a proof. It could be that the interval is genuinely positive, and the numerics are convincing, but the central claim is stated as a demonstrated result. A rigorous positivity proof, or at least a certificate, would close this.\n\nAlso, QED uniqueness in Sec. 6 relies on a footnote saying \"for technical reasons\" — that is hand-waving. The uniqueness of QEDs for countable R-positive chains may be true via [7], but the normalisability cancellation is only shown for λ_u; other λ are dismissed without proof.\n\nMinor: shape conservation in Sec. 4.2 is heuristic, and the stability analysis in Sec. 5 is simulation-based. These are supporting, not central.\n\nFor researchers in random dynamical systems and QSD theory, this is a valuable example, and the numerics are striking. But as it stands, the headline claim is conditional. It deserves a serious referee — the gaps are addressable and the core construction is strong.\n\nSend it to peer review. Ask for a proof or a much sharper numerical certificate for positivity, and for the QED uniqueness argument to be made explicit.","headline":"A concrete countable Markov chain with a plausible continuum of QSDs, but the continuum rests on numerically verified positivity and a hand-waved QED uniqueness proof.","tokens_in":18363,"tokens_out":1685,"would_cite":false,"duration_ms":17547,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J10","37A50","37H12"],"pacs":[],"model":"deepseek-v4-flash","headline":"An open Pelikan-map chain is claimed to have a continuum of quasi-stationary distributions, each with its own escape rate.","keywords":["quasi-stationary distributions","quasi-ergodic distributions","Pelikan random map","countable Markov chains","substochastic matrices","escape rates","R-recurrence","random resetting"],"falsifier":"Choose rational p, ε inside the claimed valid region and α just below α_u; iterate the recurrence (44) exactly, or with rigorous interval arithmetic, until the dominant-root criterion (48) applies. A single negative μ_k before that cutoff would disprove the continuum for that parameter set.","tokens_in":17522,"feed_emoji":"🎲","tokens_out":4086,"duration_ms":43247,"temperature":0.7,"pith_summary":"The paper studies a biased random walk on the nonnegative integers with resets and escape, derived from an open version of the Pelikan random map. It claims that this countable substochastic Markov chain has not one but infinitely many quasi-stationary distributions for generic parameter values, indexed by a free parameter α, with each distribution tied to its own eigenvalue λ = 1 − pεα and therefore its own long-term escape rate. If the claim is correct, the eventual escape rate is not fixed by the system alone but selected by the initial distribution. The paper also shows analytically and numerically that a unique quasi-ergodic distribution exists in the R-positive regime and does not exist in the R-transient regime.","feed_headline":"A simple random walk hosts infinitely many escape rates","feed_subtitle":"In an open biased walk, each starting distribution picks its own quasi-stationary state and decay rate.","key_machinery":"The central object is the generating function A(z) = Σ μ_k z^k for candidate invariant measures. Imposing M[A] = λA reduces the QSD equation to a rational function with free parameter α = μ_0, and the coefficients μ_k are obtained from a linear recurrence. The eigenvalue is pinned by normalization to λ = 1 − pεα. The same generating-function apparatus yields the Koopman eigenvector B(z), whose product with the principal QSD gives the QED.","core_discovery":"The transfer operator of the chain is shown to admit a family of positive eigenvectors μ(α) satisfying Mμ = λ(α)μ, with λ(α) = 1 − pεα. The paper derives these from a generating function A(z) whose coefficients are generated by a three-term recurrence. Positivity of the coefficients is verified numerically up to a tail criterion, and simulations confirm that each such measure is preserved under the conditional dynamics and produces the predicted exponential escape rate. This continuous family of QSDs appears in both R-recurrent and R-transient parameter regions. In the R-positive region, the paper explicitly constructs the unique quasi-ergodic distribution as the product of the principal QSD","pith_inferences":["If the positivity interval for α is confirmed rigorously, similar generating-function families may yield QSD continua in other resetting random walks with geometric reset distributions and biased noise.","The result suggests that other observables, such as Lyapunov exponents or metric entropy, if defined for the open system, may inherit initial-condition dependence; the paper leaves that question open.","A direct experimental probe would be to initialize an ensemble with a prescribed tail shape and compare the measured long-term survival probability to λ(α); a match for non-principal α would provide independent evidence for the continuum.","Because the paper verifies positivity by iterating to a numerical cutoff, a symbolic or interval-arithmetic proof for rational parameter values could settle the existence of the continuum without exhaustive simulation."],"forward_implications":["If the continuum of QSDs is genuine, the long-term escape rate of this system can be tuned continuously by choosing different initial distributions, without changing the parameters p or ε.","Non-uniqueness of quasi-stationary distributions occurs not only in exotic countable matrices but in a simple random-walk model with resetting and escape.","The asymptotic tail shape of a distribution is conserved by the normalized dynamics, partitioning distribution space into uncountably many invariant classes, each containing at most one QSD.","A unique normalizable quasi-ergodic distribution exists exactly in the R-positive regime; in the R-transient regime the time-averaged history of surviving trajectories does not converge.","Non-principal QSDs appear marginally stable to common smooth noise, while finite-sample noise eventually drives the system toward the principal QSD."],"fun_headline_variants":["Infinite quasi-stationary states from a single random map","Each starting measure gets its own escape rate","A random walk with a continuum of decay rates","Infinitely many quasi-stationary distributions found","Unique quasi-ergodic distributions also constructed"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claimed continuum of quasi-stationary distributions exists only if every coefficient μ_k computed from the recurrence is non-negative for all k on an interval of α; the paper verifies this numerically up to a tail criterion but does not prove it.","fun_headline_variants_meta":{"raw":{"variants":["Infinite quasi-stationary states from a single random map","Each starting measure gets its own escape rate","A random walk with a continuum of decay rates","Infinitely many quasi-stationary distributions found","Unique quasi-ergodic distributions also constructed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00021,"raw_usage":{"total_tokens":1181,"prompt_tokens":614,"completion_tokens":567,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":358,"completion_tokens_details":{"reasoning_tokens":509}},"tokens_in":358,"tokens_out":567,"duration_ms":6462,"temperature":1.0,"reasoning_tokens":509,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T13:28:04.781095+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Choose rational p, ε inside the claimed valid region and α just below α_u; iterate the recurrence (44) exactly, or with rigorous interval arithmetic, until the dominant-root criterion (48) applies. A single negative μ_k before that cutoff would disprove the continuum for that parameter set.","supporting_citations":[],"review_version":1}