{"id":"a56ed88d-6e69-4dae-a1d2-f3857dce0964","arxiv_id":"1908.09053","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Measuring qubits from finite-state controlled sources generically turns simple quantum processes into classical processes whose optimal predictors need an infinite number of states, a divergence traced to measurement-induced nonunifilarity.","lead":"This paper shows that when you measure a stream of qubits emitted by a finite-state controlled source, the resulting classical data can be so complex that predicting it requires an infinite number of internal states, even though the source itself is simple. It identifies measurement as the cause and offers new ways to quantify the randomness and memory of such measured qubit processes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'quantifiable rate' of C_mu divergence is not established: the reported statistical complexity dimensions are only upper bounds (d_mu ≤ d_LCE), and SM IIE states the bound is strict at θ_b=1.634; direct estimates of d_mu are needed.","rationale":"The reader's weakest assumption already identified the unpublished companion Ref. [22] and the possibility that d_LCE is a loose upper bound. My review finds that this concern lands concretely: SM IIE admits a strict inequality d_mu < d_LCE for one of the three displayed measurement angles, so the reported dimension there is only an upper bound. Because the central claim promises a 'quantifiable rate' of divergence, the missing direct estimate of d_mu is load-bearing. This does not overturn the qualitative nonunifilarity mechanism, and the reader's CONDITIONAL verdict already demands exactly the missing support, so no verdict change is needed. I did not identify an internal inconsistency or a counterexample to the main construction; the issue is an unsupported quantitative step. Independent support in the paper includes the concrete three-state example, the explicit nonunifilar HMM in Fig. 2(c), and numerical mixed-state sets, but no code, data, or formal proof is provided for the entropy-rate estimator or the dimension bound.","tokens_in":13161,"tokens_out":3941,"duration_ms":39770,"concrete_test":"For the Fig. 2(b) cCQS at θ_b = 1.634, compute the true statistical complexity dimension d_mu directly: generate a long mixed-state trajectory using Eq. (S6), build a histogram of the Blackwell measure on a sequence of box sizes ε, compute H_ε[R], and estimate the limit -H_ε[R]/log_2 ε with extrapolation and bootstrap error bars; compare with d_LCE computed from Eq. (S9). If the direct d_mu is statistically below d_LCE, then the reported value is only an upper bound and the quantitative central claim is not established; if d_mu ≈ d_LCE for θ_a, θ_b, and θ_c, the bound is tight in the displayed cases.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Central claim: generically a measured cCQS needs infinitely many causal states, so C_mu diverges at a quantifiable rate. The infinite-state part is supported, at least for the example, by the nonunifilarity mechanism and the numeric mixed-state sets. The quantitative rate, however, is not actually computed. The paper reports d_mu via the upper bound d_mu ≤ d_LCE, Eq. (S8), with d_LCE from Eq. (S9). SM IIE explicitly says that when the open set condition fails, 'the relationship becomes an inequality: d_mu < d_LCE', and identifies the Fig. 3 inset at θ_b = 1.634 as such a case. Thus one of the three displayed 'dimensions' is strictly an upper bound, not the divergence rate. In addition, the estimator (6) for h^B_mu, used in d_LCE, is justified only by contractivity and ergodicity results deferred to unpublished Ref. [22]. Without a direct estimate of d_mu, or a proof of equality d_mu = d_LCE for the displayed cases, the 'quantifiable rate' assertion is unsupported. The qualitative claim of divergent C_mu may survive, but the paper's headline quantitative claim requires either proof or an explicit downgrade to an upper-bound statement.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies classically controlled qubit sources (cCQSs): finite-state hidden Markov models that emit pure qubit states, which are then measured projectively to produce binary classical stochastic processes. The central claim is that, even when the controller is finite-state, measurement generically induces nonunifilarity in the resulting hidden Markov model, so that the measured process's minimal predictor (its epsilon-machine) requires an uncountable infinity of causal states and its statistical complexity C_mu diverges. To quantify this divergence the paper introduces an entropy-rate estimator, Eq. (6), based on time-averaging over mixed states, and the statistical complexity dimension d_mu, bounded by a modified Lyapunov dimension d_LCE in Eqs. (S8)-(S9). The ideas are illustrated on a three-state example, sweeping the measurement angle theta and plotting mixed-state sets, entropy rates, and complexity dimensions.","tokens_in":13495,"tokens_out":3788,"duration_ms":42068,"significance":"If the qualitative claim holds, this is an interesting and potentially important observation: a finite-state controlled quantum source can appear to a classical observer to require an infinite number of predictive features, and the measurement choice can both add and remove apparent randomness and structure. The paper's explicit example and the numerical mixed-state sets provide concrete evidence for the nonunifilarity mechanism and for the divergence of the causal-state count. The framing in terms of epsilon-machines and computational mechanics is appropriate and makes the operational meaning of the quantities clear. However, the paper's quantitative headline—that the divergence rate is 'quantifiable'—is not yet established: the entropy-rate estimator and the d_LCE bound are both deferred to an unpublished companion manuscript (Ref. [22]), and the paper itself states that the bound can be strict. The manuscript would be strengthened by a self-contained proof of the deferred results or, failing that, by an explicit downgrade of the quantitative claims to upper-bound statements.","major_comments":[{"comment":"The central entropy-rate estimator h^B_mu in Eq. (6) is load-bearing for all reported quantitative randomness values, but its validity is not established in this manuscript. SM IID states: 'The development of that expression and the proof that it is correct is given in Ref. [22].' Since Ref. [22] is an unpublished companion by the same authors, the reader cannot verify the contractivity and ergodicity conditions on which the time-average estimator rests. The revision should either include the proof (even in the SM) or cite a publicly available derivation; otherwise the reported h^B_mu values are unsupported.","section":"SM IID and Eq. (6)"},{"comment":"The statistical complexity dimension d_mu is not actually computed; only the upper bound d_mu <= d_LCE is reported, and SM IIE explicitly states that at theta_b = 1.634 the relationship becomes a strict inequality, d_mu < d_LCE. The insets of Fig. 3 present these d_LCE values as the 'statistical complexity dimensions,' but they are not demonstrated rates of divergence. To support the claim of a 'quantifiable rate,' the paper needs either direct estimates of d_mu (with a documented convergence criterion) or a proof that equality d_mu = d_LCE holds for the displayed cases. The box-counting estimates in SM IIE are finite-epsilon approximations and do not close this gap.","section":"SM IIE and Eqs. (S8)-(S9)"},{"comment":"The statement that 'nonunifilarity is generic to cCQSs and even to more general qubit sources' is presented as a first result, but no proof or precise formal statement is given. The paper demonstrates nonunifilarity for one explicit example and shows numerically that its mixed-state set is fractal, yet the genericity claim is stronger. A rigorous statement of the generic property, or at least a precise conjecture with supporting evidence, is needed for this to be a theorem rather than an observation about a single family.","section":"Main text, 'Uncountable Predictive Features'"}],"minor_comments":[{"comment":"There is a typo in the phrase 'should also beefficient'; it should read 'should also be efficient.'","section":"Introduction"},{"comment":"The sentence 'composing the measurement operator with the the qubit controller HMM' contains a duplicated 'the.'","section":"Main text, 'Measured Qubit Processes'"},{"comment":"In the definition of mixed states, the phrase 'the the nonunifilar HMM’s internal states' contains a duplicated 'the.'","section":"SM IIC"},{"comment":"The caption says complexity measures were computed with '𝓁 = 10^6 iterates,' but iterates are not the same as block length; this should be clarified, and the distinction between direct estimates and upper bounds should be stated in the caption.","section":"Figure 3 caption and SM IIE"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the reliance on unpublished companion Ref. [22] for the two quantitative tools. If that manuscript is not made available with the revision, the quantitative claims cannot be independently checked. The qualitative nonunifilarity-and-divergence example is likely salvageable and could form the core of a publishable paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new and useful thing here is the identification of measurement-induced nonunifilarity as the mechanism that makes measured qubit processes from finite-state controlled sources require infinitely many causal states. The concrete example—Fig. 2(b) measured at various angles—is worked through carefully, and the numerical exploration of complexity as a function of measurement angle is a nice touch. The paper also gives a constructive route to entropy-rate estimation via Eq. (6), which is a real step beyond the abstract Blackwell integral. Credit where due: the authors are clear about what is established and what is deferred, and the supplementary materials are honest about the open-set-condition caveat.\n\nThe soft spots are real but not fatal. The central quantitative claim, that C_mu diverges at a 'quantifiable rate' via the statistical complexity dimension, is not actually computed. The paper reports d_mu through the upper bound d_mu ≤ d_LCE, and SM IIE explicitly notes that at θ_b = 1.634 the inequality is strict. So one of the three headline dimensions is an upper bound, not the divergence rate. The stress-test note gets this right. The entropy-rate estimator Eq. (6) also rests on contractivity and ergodicity proofs deferred to unpublished Ref. [22], so a reader cannot independently verify the numbers. The genericity claim—that nonunifilarity is generic to cCQSs—is asserted rather than proven, though the mechanism is plausible and the example supports it.\n\nNone of this undermines the qualitative conclusion: generically, measured qubit processes from finite-state sources will need infinitely many predictive states. That is a physically meaningful insight and worth having on the record. But the quantitative rate of divergence is not established in this preprint. The authors themselves point to the needed fixes—publish the companion derivations, provide direct estimates of d_mu or conditions for equality with d_LCE, and give error bars on the reported dimensions. No code or data are shipped, which is a minor stumble for a numerical paper.\n\nWho is this for? People working on quantum stochastic processes, computational mechanics, and hidden Markov models. It is a serious contribution that deserves referee time. My recommendation: send it out, but with a clear request that the authors either include the supporting proofs or explicitly label the reported dimensions as upper bounds. This is exactly the kind of paper that is improved by a referee asking for the quantitative claims to be matched by evidence.","headline":"A clear, honest paper that identifies measurement-induced nonunifilarity as the mechanism behind infinite predictive complexity in measured qubit streams, but the headline quantitative claim about a quantifiable divergence rate leans on an unpublished companion and an upper bound.","tokens_in":13935,"tokens_out":624,"would_cite":true,"duration_ms":7814,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["89.70.+c","89.70.Cf","03.65.Ta","03.67.-a"],"model":"deepseek-v4-flash","headline":"Measuring qubits turns simple finite-state sources into infinitely complex prediction problems.","keywords":["quantum measurement","controlled qubit source","hidden Markov model","nonunifilarity","causal states","statistical complexity","entropy rate","mixed-state presentation"],"falsifier":"Generate a very long measurement sequence from the Fig. 2(b) cCQS at a generic angle, compute the time-averaged entropy rate of Eq. (6) for many independent initial mixed states, and box-count the recurrent mixed states at decreasing scales; if the average depends on the initial state or fails to converge while the box-counting dimension is finite, the claimed quantification is falsified.","tokens_in":12977,"feed_emoji":"⚛️","tokens_out":9410,"duration_ms":87270,"temperature":0.7,"pith_summary":"The paper studies sources that emit one qubit per time step under finite-state classical control, with each qubit measured projectively as it is produced. It argues that, for almost all such sources and almost all measurement bases, the resulting classical bit process is unboundedly complex: no finite-state predictor can be optimal because the number of causal states—the minimal predictive features—is infinite. The mechanism is nonunifilarity: after measurement, the same output string can be produced by many internal state paths, so the observer's state of knowledge never collapses to a finite set. To quantify the effect, the paper derives efficient estimators for the entropy rate and for the statistical complexity dimension $d_\\mu$, the rate at which prediction memory diverges. The payoff is a concrete picture of how a simple quantum source can appear unboundedly structured to an experimenter, with the choice of measurement basis itself adding or removing randomness and memory.","feed_headline":"Measuring qubits forces infinite predictive memory","feed_subtitle":"Even a finite-state qubit source, once measured, demands infinitely many predictive features.","key_machinery":"Measurement-induced nonunifilarity is the engine: composing the controller's labeled transition matrices with measurement probabilities gives matrices $T^{(x)} = \\sum_{\\rho_j} T_{\\rho_j} \\Pr(x|\\rho_j)$ in which the same symbol can be emitted along several state paths. The workhorse is the mixed-state presentation, built from the conditional distributions $\\eta(w) = \\Pr(S_\\ell = s | w)$ over the hidden controller states after seeing word $w$; the presentation is unifilar even when the original HMM is not. Because recurrent mixed states form the causal states, the uncountability of this set is what makes $C_\\mu$ diverge. The paper controls the divergence with the statistical complexity dimension $d_\\mu$, the scaling of the coarse-grained Shannon entropy of the Blackwell measure on the mixed-state simplex, and bounds it by the Lyapunov dimension $d_{\\mathrm{LCE}}$ of the mixed-state dynamics.","core_discovery":"Even when the qubit generator is a finite-state classically-controlled qubit source (cCQS)—a hidden Markov model whose emitted symbols are pure qubit states—performing a projective measurement on each output qubit generically produces a measured process that is nonunifilar. In such a process, a measurement string does not determine the sequence of hidden controller states; the set of mixed states, each encoding the observer's probability distribution over controller states given the past, forms an uncountable fractal set. The recurrent mixed states are exactly the process's causal states, so the minimal optimal predictor has infinitely many states and the statistical complexity $C_\\mu$ diverges. The paper shows the divergence rate is finite and quantifiable via the statistical complexity dimension $d_\\mu$, estimated from the measured process's entropy rate and the Lyapunov spectrum of the mixed-state dynamics. The central demonstration uses a three-state cCQS emitting $|0\\rangle$ and $|+\\rangle$ qubits: sweeping the measurement angle, the entropy rate and dimension vary smoothly, with special angles yielding a single-state iid process or finite-complexity processes.","pith_inferences":["A natural extension would test whether the same generic divergence appears when the emitted states are mixed rather than pure; if nonunifilarity is the mechanism, the divergence should persist, though the dimension formula may need modification.","For experimental work, the result implies that model-order selection from measured qubit streams will systematically under-estimate memory: any finite-Markov approximation is biased, and the bias should grow with sequence length as the complexity dimension dictates.","One could formulate measurement-basis optimization: because the paper's angle sweep shows smooth variation in $d_\\mu$, a basis that minimizes or maximizes the complexity dimension can be sought for tasks that need predictable or random outputs."],"forward_implications":["A generic measured qubit process cannot be optimally predicted by any finite-state model, even though the underlying source is finite-state; the minimal predictor is uncountably infinite.","The entropy rate of a nonunifilar hidden Markov model can be estimated from the mixed-state dynamics by averaging over one long realization, rather than by integrating over an abstract invariant measure.","Choosing the measurement basis is a way of tuning the observed process: some bases erase all memory, producing an iid biased coin, while others preserve finite complexity, and generic bases give divergent complexity with a finite dimension.","The estimation algorithms apply to any ergodic nonunifilar HMM, giving a constructive route to entropy rates and statistical complexity dimensions that previously had only formal solutions."],"supporting_citations":[{"why":"introduced mixed states and the Blackwell measure, giving the formal entropy-rate integral and the mixed-state presentation used throughout.","marker":"[20]"},{"why":"established that recurrent mixed states correspond exactly to the causal states of the process.","marker":"[21]"},{"why":"the unpublished companion that supplies the contractivity and ergodicity results behind Eq. (6) and the d_LCE bound.","marker":"[22]"},{"why":"defined the statistical complexity dimension used to track the divergence of C_mu.","marker":"[23]"},{"why":"defines the epsilon-machine and causal states, the predictive model whose state count diverges.","marker":"[19]"},{"why":"provides the Lyapunov dimension formula adapted to bound d_mu.","marker":"[33]"},{"why":"supplies the open set condition that distinguishes when the d_LCE bound is tight versus an inequality.","marker":"[35]"}],"fun_headline_variants":["Measure qubits, get infinite predictive states","Qubit measurement reveals fractal predictive structure","Infinite memory from measured qubit streams","Measurement makes qubit prediction endlessly complex"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative claims stand on the assumption that as measurement records grow, the observer's probability distribution over the hidden states settles into a unique, stable, and ergodic pattern—a property the paper inherits from an unpublished companion work; if that fails, the reported rates and dimensions are not established.","fun_headline_variants_meta":{"raw":{"variants":["Measure qubits, get infinite predictive states","Qubit measurement reveals fractal predictive structure","Infinite memory from measured qubit streams","Measurement makes qubit prediction endlessly complex"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000176,"raw_usage":{"total_tokens":1236,"prompt_tokens":842,"completion_tokens":394,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":458,"completion_tokens_details":{"reasoning_tokens":340}},"tokens_in":458,"tokens_out":394,"duration_ms":4539,"temperature":1.0,"reasoning_tokens":340,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:23:37.648836+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate a very long measurement sequence from the Fig. 2(b) cCQS at a generic angle, compute the time-averaged entropy rate of Eq. (6) for many independent initial mixed states, and box-count the recurrent mixed states at decreasing scales; if the average depends on the initial state or fails to converge while the box-counting dimension is finite, the claimed quantification is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduced mixed states and the Blackwell measure, giving the formal entropy-rate integral and the mixed-state presentation used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"established that recurrent mixed states correspond exactly to the causal states of the process."},{"cited_title":"Blackwell","cited_arxiv_id":null,"evidence_quote":"the unpublished companion that supplies the contractivity and ergodicity results behind Eq. (6) and the d_LCE bound."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"defined the statistical complexity dimension used to track the divergence of C_mu."},{"cited_title":"Galland, Y","cited_arxiv_id":null,"evidence_quote":"defines the epsilon-machine and causal states, the predictive model whose state count diverges."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the Lyapunov dimension formula adapted to bound d_mu."},{"cited_title":"Frederickson, J","cited_arxiv_id":null,"evidence_quote":"supplies the open set condition that distinguishes when the d_LCE bound is tight versus an inequality."}],"review_version":1}