{"id":"1506a34a-d1ea-4b5c-9c0d-f033d03685c9","arxiv_id":"2601.14379","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Controlled stochastic/bistochastic circuits have vanishing two-point correlations away from the same site, and multipoint correlations vanish unless the two rightmost points coincide.","lead":"This paper proves that a large class of quantum and classical circuits built from 'controlled' gates have almost no long-range correlations: any two measurements at different positions are completely uncorrelated, and only same-site autocorrelations survive. The result identifies a common property shared by random quantum controlled circuits and classical cellular automata, giving theorists a new exactly solvable family of many-body dynamics.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Thm 2's multipoint proof is a one-sentence sketch that does not address arbitrary time orderings of the rightmost operator; the abstract's multipoint claim is therefore unsupported.","rationale":"The reader correctly identifies a missing justification in the proof of Theorem 1's individual cases, but I judge the more load-bearing concern to be Theorem 2. Theorem 2 is part of the paper's headline claim and its proof is only a heuristic paragraph. The reader's concern about the light-cone truncation, while valid, can be addressed by a straightforward induction that uses only Eq. (5) (for x>0) or Eq. (6) (for x<0), so it does not threaten the truth of the two-point result. The multipoint theorem, however, requires a rigorous argument for arbitrary time orderings of the unique maximal position; without it, the paper overclaims. A concrete numerical check on a small nontrivial configuration can test whether the statement itself is true, which is the first step in deciding whether the proof gap is fatal or merely expository. The verdict remains CONDITIONAL: the main two-point result for bistochastic gates is likely correct, but the multipoint claim needs either a full proof or a precise counterexample.","tokens_in":10231,"tokens_out":32753,"duration_ms":285697,"concrete_test":"Exact tensor-network contraction: evaluate the three-point function for L=3, T=2 with CNOT gates and Pauli-Z insertions at (x,t)=(0,0), (2,1), (1,2) (unique maximum at intermediate time 1). If the result is nonzero to machine precision, Theorem 2 is falsified; if zero, repeat for T=3 with an additional even-layer gate to confirm the sketch's assumptions before accepting the proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in the abstract—'for multipoint correlations the two rightmost operators must act on the same site'—rests on Theorem 2, proved in Sec. IV with a single paragraph: 'The same argument can be extended... all states |−⟩ to the right from the second rightmost operator at the position x_{n−1} can propagate upwards and downwards. If they meet the rightmost operator, the correlations vanish.' This is not a proof. In particular, it does not handle the case where the rightmost operator is not at the latest time. If the rightmost operator acts at an intermediate time, the excitation it creates can later interact with operators to its left via gates whose control is non-flat; the flat-state barrier is no longer a simple spacetime region. The sketch gives no diagrammatic contraction for this case, and no induction on n is provided. Because Theorem 2 is explicitly advertised in the abstract and title, this omitted proof is a load-bearing gap. The reader's weaker assumption about the light-cone truncation in Thm 1 is less severe: for part (i) alone, a direct induction shows the right half remains |−⟩ under Eq. (5), and the analogous statement holds for part (ii) under Eq. (6), so the gap is repairable. The multipoint theorem, by contrast, has no such elementary fallback in the paper.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies one-dimensional brickwork circuits built from two-site controlled gates U=Σ_i |i><i|⊗u_i. It introduces two simple conditions: the controlled-stochastic condition Eq. (5), u_i|−⟩=|−⟩, and the controlled-bistochastic condition Eq. (6), ⟨−|u_i=⟨−|. Theorem 1 claims that under Eq. (5) the infinite-temperature two-point correlation C(x,t) vanishes for x>0, under Eq. (6) for x<0, and under both conditions for all x≠0. Theorem 2 claims that, when both conditions hold, any n-point correlation C(x_1,t_1,...,x_n,t_n) vanishes if the set {x_1,...,x_n} has a unique maximal value, i.e. the two rightmost operators must act on the same site. The authors then give examples: one-replica averages of random controlled quantum gates reduce to controlled-bistochastic classical stochastic circuits, and deterministic controlled cellular automata satisfy the conditions. Section VI argues that the autocorrelation is generically hard to compute and shows numerical exponential decay, with exceptions. Appendices give a characterization of gates satisfying Eqs. (5)-(6) and a generalized condition.","tokens_in":10523,"tokens_out":16511,"duration_ms":158870,"significance":"The central conceptual claim is attractive and potentially useful: a broad, simply defined class of circuits has correlation functions supported only on the equal-space line, despite generically complex dynamics. The conditions are clean, the diagrammatic method is economical, and the connection to averaged random controlled gates and to East-type cellular automata gives concrete physical content. The paper contains no fitted parameters, and the main statements are falsifiable. The reduction in Sec. V A from an averaged random quantum circuit to an effective classical stochastic circuit is elegant. However, the multipoint theorem is advertised in the abstract and title but is not proved in the submitted text, and the proof of the separate parts of Theorem 1 is displayed only in a form that assumes both conditions. These gaps are repairable in my view, but they must be fixed before the central claims can be considered established.","major_comments":[{"comment":"The proof of Thm. 1(i),(ii) starts from a diagram that the caption says is 'already simplified ... using the bistochasticity of the gates.' That simplification invokes both U|−−⟩=|−−⟩ and ⟨−−|U=⟨−−|. But parts (i) and (ii) are claimed under Eq. (5) alone and Eq. (6) alone, respectively. As written, the displayed proof does not identify which boundary contractions are legitimate in the stochastic-only case. The gap appears repairable — for example, an induction showing that a gate whose right leg is |−⟩ acts as the identity under Eq. (5) supplies the missing step — but the individual statements are not actually proved by the diagram as presented. Please make explicit, for each case, which equations are used at each simplification.","section":"§IV, Eq. (10), Theorem 1"},{"comment":"The proof of Theorem 2 consists only of the paragraph beginning 'The same argument can be extended...'. It provides no induction on n and no treatment of the time orderings in Eq. (9). If the unique rightmost spatial operator acts at an intermediate time, its insertion punctures the flat-state barrier; after that time the evolution can couple the operator to operators on its left through gates with at least one non-flat leg, a situation the paragraph does not analyze. Since the abstract and title advertise the multipoint statement, this missing case is load-bearing. Please give a complete proof, either by strong induction or by explicit diagrams, covering all relative time orderings of the rightmost operator.","section":"§IV, Theorem 2"}],"minor_comments":[{"comment":"The definition of C(x,t) is written with subscripts x and 0 but not with explicit operators O_x and O_0. Because the theorem concerns diagonal traceless observables, please define the observable explicitly and state its expansion in the |−⟩, |(k)⟩ basis.","section":"§III, Eq. (8)"},{"comment":"The statement 'we can reduce the folded space from four to two dimensional' is asserted rather than shown. Please spell out why the averaged evolution superoperator preserves the subspace spanned by |00⟩ and |11⟩ at every site, including the action of the |#⟩⟨#| factor on the target leg.","section":"§V A, after Eq. (16)"},{"comment":"The phrase 'These matrices annihilate the flat state when acting upon it from the left' should presumably read 'from the right,' since B_α|−⟩=0 for α∈S'. In addition, the u'_α/u_α notation in Eqs. (A12)-(A14) is confusing: the first sum in Eq. (A13) uses the stochastic u_α while Eq. (A12) uses u'_α for the same terms. Please rewrite this derivation with clearer notation.","section":"Appendix A, Eqs. (A10)-(A14)"},{"comment":"Please provide axis labels and specify the gate used and the parameters of the numerical simulation. The caption states that the difference between L=27 and L=29 is almost indistinguishable, but the plotted quantity itself is not described.","section":"§VI, Fig. 1"},{"comment":"The phrasing 'stochastic and bistochastic controlled gates lead to ... vanish everywhere except when the two operators act on the same site' could be misread as applying to the stochastic-only class. Theorem 1 shows that the full statement requires both conditions. Please qualify the abstract accordingly.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The two major comments identify real gaps in the submitted proof, but both appear repairable within the manuscript's scope. The central idea is sound, and the examples are convincing. I would not recommend rejection; the authors should be asked to supply a complete proof of Theorem 2 and to make the hypotheses explicit in the proof of Theorem 1."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Honest take: the two-point core is real and the paper earns its keep, but the multipoint theorem advertised in the title and abstract is a sketch, not a proof. If you only remember one thing: read Sec. IV carefully before citing Thm 2.\n\nWhat's new and good: the controlled-stochastic and controlled-bistochastic conditions (Eqs. 5-6) and the proof that they force two-point infinite-temperature correlations to vanish for x≠0. That is a genuinely general structural statement, unifying specific known results for the East model and dual-unitary hierarchies. The diagrammatic proof for the bistochastic case is solid. I also like App. A: the operator-Schmidt characterization of gates satisfying the conditions is clean and useful. And the random-controlled-gate example in Sec. V A shows the conditions arise naturally from averaging, not as an ad hoc constraint.\n\nSoft spots: the proof of Thm 1 is written for the combined case; the diagrams in Eq. (10) are already simplified using both conditions before the individual parts (i) and (ii) are stated. The reader flagged that the light-cone truncation under only Eq. (5) or only Eq. (6) is not spelled out. That's fair, but a direct induction on the right (or left) boundary shows the flat state is preserved there, so the two halves of Thm 1 are repairable in a few lines. Minor.\n\nLess minor: Thm 2. The text says 'the same argument can be extended' and then gives one sentence about |−⟩ states to the right of the second-rightmost operator propagating up and down. That does not handle the case where the rightmost operator sits at an intermediate time, because the control region between it and the second-rightmost operator is no longer flat. The stress-test note is right: the abstract's multipoint claim currently rests on a one-sentence sketch. This needs a real proof—either a diagrammatic contraction for all time orderings or an induction on n.\n\nThe autocorrelation section is also thin numerically: Fig. 1 shows exponential decay but there is no data, code, or error bars, and the claim is presented as a survey. That part should be marked as a heuristic unless the artifacts are released.\n\nNet: Thm 1 and the structural conditions are worth serious refereeing. Thm 2 is plausibly true but not proved as written. The paper would be publishable after the multipoint proof is supplied and the numerics are either released or downgraded. I would send it to peer review, but I would ask the referee to focus on Sec. IV and Thm 2.","headline":"The two-point result is solid and worth publishing; the multipoint theorem needs a real proof before the abstract's claim is backed.","tokens_in":11014,"tokens_out":3550,"would_cite":true,"duration_ms":30645,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that in brickwork circuits whose gates satisfy controlled-stochastic and controlled-bistochastic conditions, two-point correlation functions vanish unless both operators act on the same site, and multipoint functions vanish","keywords":["controlled gates","stochastic circuits","bistochastic gates","correlation functions","quantum circuits","cellular automata","infinite-temperature correlations","light-cone contraction"],"falsifier":"Simulate a small brickwork chain (say L=5, t=2) with q=3 gates U = Σ_i |i⟩⟨i| ⊗ u_i where each u_i is stochastic and fixes the flat state but is not bistochastic (e.g., a row-stochastic matrix with column sums not all equal). Compute the two-point function C(x,t) for x>0 by exact enumeration. If any such gate yields C(x>0,t)≠0, Theorem 1(i) is false; if it is identically zero, the one-sided condition alone is sufficient.","tokens_in":10100,"feed_emoji":"🎲","tokens_out":8015,"duration_ms":82961,"temperature":0.7,"pith_summary":"The paper proves that in a large family of quantum and classical circuits built from controlled gates, the infinite-temperature correlation functions are almost completely empty: a two-point correlation of diagonal traceless observables vanishes unless both operators sit on the same spatial site, and a multipoint correlation vanishes unless the two rightmost operators sit on the same site. The key is a pair of simple identities — the controlled-stochastic and controlled-bistochastic conditions — which say that the flat uniform state is a fixed point of the gates. When these hold, light-cone diagrams can be contracted down to a single line, and any attempt to separate the two operators in space produces a contraction with an orthogonal state that is zero. The same identities arise naturally from averaging over random controlled quantum gates and from deterministic cellular automata, so the result covers both averaged quantum dynamics and classical stochastic dynamics. The autocorrelation is the sole survivor, and the paper argues it typically decays exponentially to a value exponentially small in system size.","feed_headline":"Proven: correlations vanish off the equal-site line","feed_subtitle":"Even chaotic-looking circuits have trivial correlations unless operators meet on the same site.","key_machinery":"The key objects are the flat state |−⟩, the uniform superposition over local basis states, and its orthogonal complement | ⟩. A two-site controlled gate U acts on a target site conditionally on the control site's state, and the paper isolates two identities: the controlled-stochastic condition U(1⊗|−⟩)=1⊗|−⟩ and the controlled-bistochastic condition (1⊗⟨−|)U=1⊗⟨−|. These identities make |−⟩ a fixed point of the gate from one or both sides, which lets the light cone of a correlation diagram be contracted inward until the inserted operator is surrounded by flat states. The diagram then reduces to a contraction of a flat state with an orthogonal state, and that contraction vanishes. The same co","core_discovery":"The central discovery is Theorem 1: for a brickwork circuit of two-site gates satisfying the controlled-stochastic condition U(1⊗|−⟩)=1⊗|−⟩, the two-point function C(x,t)=⟨−…−|O_x U(t) O_0|−…−⟩ of diagonal traceless observables vanishes for x>0; if the gates also satisfy the controlled-bistochastic condition (1⊗⟨−|)U=1⊗⟨−|, it vanishes for x<0; with both, it vanishes for every x≠0. Theorem 2 extends to multipoint functions: under both conditions C(x_1,t_1,...,x_n,t_n) vanishes whenever the set {x_1,...,x_n} has a unique maximal value. The proof works by simplifying the tensor-network diagram for the correlation using the two fixed-point identities, reducing the light cone to a triangle and t","pith_inferences":["If the result holds beyond the proof's one-sided caveat, it suggests that these circuits are correlation-trivial at infinite temperature: for the observables considered, operator spreading in the Heisenberg picture must be confined to the spatial origin, which would rule out standard diffusive or ballistic spreading profiles in those correlation functions.","The paper leaves open a multi-replica generalization of the controlled-stochastic and controlled-bistochastic identities; if such identities can be lifted to the replica structure, quantities like entanglement entropy and out-of-time-ordered correlators may become tractable in the same circuits — an extension the paper notes but does not prove.","A testable asymmetry: under Eq. (5) alone the theorem predicts vanishing only to the right. Numerically checking left-moving correlations in a deterministic cellular automaton with controlled gates that are stochastic but not bistochastic would distinguish the one-sided from the two-sided contraction mechanism."],"forward_implications":["For any brickwork circuit satisfying both conditions, the only nonvanishing two-point infinite-temperature correlations are autocorrelations; all information about a local operator is captured by its return probability to its own site.","Multipoint correlations are similarly constrained: any cluster with a unique rightmost probe has zero correlation, so nonzero multipoint signals require two probes on the same rightmost site.","Random circuits built from controlled Haar-averaged gates fall into this class, so their averaged correlation functions are exactly trivial off the equal-space line, not merely approximately.","Deterministic classical cellular automata built from controlled permutations (e.g., identity and CNOT for bits) inherit the same correlation structure.","Autocorrelations escape these theorems; the paper argues they generically decay exponentially to a value exponentially small in system size, and can be exactly zero or constant in special cases."],"fun_headline_variants":["Correlations vanish off the equal-site line","Theorem: correlations only at same site","Correlations die unless on same site","Complex dynamics, simple correlations: proof","Controlled circuits: correlations vanish off-site"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof assumes that the uniform background state is a fixed point of each gate from both sides, which lets the light cone be collapsed before the one-sided condition is used; if that collapse needs both conditions at once, the individual x>0 and x<0 vanishing statements are not established.","fun_headline_variants_meta":{"raw":{"variants":["Correlations vanish off the equal-site line","Theorem: correlations only at same site","Correlations die unless on same site","Complex dynamics, simple correlations: proof","Controlled circuits: correlations vanish off-site"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1542,"prompt_tokens":675,"completion_tokens":867,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":419,"completion_tokens_details":{"reasoning_tokens":804}},"tokens_in":419,"tokens_out":867,"duration_ms":9757,"temperature":1.0,"reasoning_tokens":804,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T09:14:47.457013+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate a small brickwork chain (say L=5, t=2) with q=3 gates U = Σ_i |i⟩⟨i| ⊗ u_i where each u_i is stochastic and fixes the flat state but is not bistochastic (e.g., a row-stochastic matrix with column sums not all equal). Compute the two-point function C(x,t) for x>0 by exact enumeration. If any such gate yields C(x>0,t)≠0, Theorem 1(i) is false; if it is identically zero, the one-sided condition alone is sufficient.","supporting_citations":[],"review_version":1}