{"id":"9f0e9d27-fe7d-4a3d-aeba-1625f1bd1799","arxiv_id":"2502.01735","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A postselection-free tree-circuit protocol with Haar-random gates and linear-time classical decoding observes the measurement-induced phase transition on Quantinuum H1-1, consistent with exact theory.","lead":"Physicists detected the measurement-induced phase transition on a 20-qubit trapped-ion quantum computer without the usual postselection overhead, using a tree-shaped circuit with random universal gates and a classical decoder that scales linearly with qubit count. The measured order parameter closely matches analytical and simulated theory, demonstrating a scalable route to studying measurement-driven entanglement transitions.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unbiasedness of \\hat Z in Eq. (6) is exact only for the ideal Haar ensemble; device noise can break the rotational invariance it relies on, and no noise-model test is reported.","rationale":"The reader's weakest_assumption points to the independence of magnitude and direction of the Bloch vector. In the ideal ensemble this property is not a finite-size artifact: it follows from Haar rotational invariance for every t, so I would not attack the mathematics of Eq. (6). The load-bearing gap is the unquantified passage from the ideal Born distribution to the noisy device distribution. A noiseless simulation cannot settle this; a calibrated noisy simulation can. The exact critical point θc=2.2142(2) and λ=1 also rely on prior work, but the pool-method curves in Fig. 5 do not depend on λ=1, so that dependence is less central to the experimental comparison. I therefore keep the reader's conditional verdict: the paper is acceptable if the noise robustness check is provided or if the 'precisely described by theory' claim is softened.","tokens_in":22011,"tokens_out":24740,"duration_ms":267783,"concrete_test":"Run Algorithm 1 on the public Zenodo code for the same t=4 circuits and θ values under (i) a noiseless simulation and (ii) a noisy simulation with an error model calibrated to H1-1 (e.g., two-qubit ZZ(φ) error ~1e-2, single-qubit error, SPAM/readout asymmetry), using Nc=834 and Ns=8. For each θ, compare the noisy \\hat Z_{t,Nc}(θ) with the noiseless Z_t(θ). If the noisy estimate differs from the noiseless one by more than the 1.96·E_{t,Nc}(θ) error bars in Fig. 5 for any θ, the hardware result is not robustly described by the ideal theory; if all differences are within those error bars, the noise-induced bias concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The derivation of Eq. (6) is correct in the ideal setting: at any finite t, Haar invariance of the independently sampled single-qubit unitaries makes the unconditional distribution of the probe Bloch vector n_t rotationally invariant, so |n_t| and its direction are independent and the estimator is unbiased. The vulnerability is the step from this ideal average to hardware. Algorithm 1 computes X_t from the observed weak-measurement record Mw and the *ideal* classical prediction n_t(θ,U,Mw), while m0 is generated by the actual noisy device. The identity E_{m0|U,Mw}[(-1)^{m0}] = n_z(θ,U,Mw) holds only if the measurement record is Born-distributed under the ideal circuit with the same U; under T1/T2, gate, or readout errors this equality fails, and the remaining argument in Eq. (6) cannot be invoked. The paper's statement 'assuming the noise in the Quantinuum H1-1 is not too large' (Section V.A) is not quantified by any error-model simulation or systematic-error estimate. The reported agreement in Fig. 5 may be genuine, but the central claim that the experimental estimates are 'precisely described by theory' depends on an unverified rotational-invariance assumption about the noisy ensemble.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs a binary-tree quantum circuit model (expansion and collapse processes) built from Haar-random single-qubit gates and weak measurements of tunable strength, and proposes a postselection-free estimator \\hat Z_{t,N_c}(θ) (Algorithm 1) for the average minimal eigenvalue Z_t(θ) of the probe qubit. Using rotational invariance of the Haar ensemble, Eq. (6) proves the estimator is unbiased, and the classical decoder is linear in the number of qubits. The authors derive an exact critical point θ_c = 2.2142(2) via a Fisher-KPP-type mapping, simulate finite-t order parameters with the pool method, and compare the simulated curves with data from the Quantinuum H1-1 for t ≤ 4, N_c = 834, and N_s = 8, reporting good agreement without error mitigation.","tokens_in":22219,"tokens_out":7685,"duration_ms":81113,"significance":"If the results hold, the protocol is a significant advance: it bypasses the exponential postselection overhead for monitored circuits with universal (Haar-random) gates on tree geometries, while providing an exactly solvable critical point and critical scaling. The estimator proof in Eq. (6) is clean and self-contained, and the theoretical curves are parameter-free in the sense that no fitting to the experimental data is performed. The public data and code repository is a further strength. Two caveats should be stated fairly: the experiment is limited to t ≤ 4, so it demonstrates finite-size signatures rather than an asymptotic phase transition, and the agreement validates the hardware implementation of the model rather than providing an independent test of the MIPT. These caveats are partially acknowledged in the text and do not by themselves invalidate the protocol claim.","major_comments":[{"comment":"The proof that \\hat Z_{t,N_c}(θ) is unbiased relies on the conditional Born-rule identity E_{m0|U,Mw}[(-1)^{m0}] = n_z(θ,U,Mw), and on the Haar-invariance argument that makes the Bloch-vector magnitude and direction independent. On the real device, m0 is sampled from a noisy circuit while n_t is computed from the ideal dynamics, so this identity is not exact; the phrase \"assuming the noise in the Quantinuum H1-1 is not too large\" is not a substitute for a quantitative statement. Because the central claim of the paper is that the experimental data are \"precisely described by theory,\" I ask for an explicit error-model or systematic-error analysis: for example, simulate Algorithm 1 with depolarizing, dephasing, and readout noise at the device's calibrated parameters and show that the induced bias in \\hat Z is below the reported 1.96 E_{t,N_c} error bars, or provide a hard bound on the bias derived from measured gate and measurement fidelities.","section":"Section V.A and Eq. (6)"},{"comment":"The agreement in Fig. 5 is assessed only visually, and no goodness-of-fit statistic is reported. The t ≤ 3 curves are obtained by truncating the same t = 4 measurement records (Section V.D), so the four datasets are statistically correlated, and the text does not explain how this correlation affects the comparison. A quantitative comparison, e.g. a reduced chi-square or maximum deviation expressed in units of the reported 95% confidence intervals, with a statement about how truncation correlations are handled, would be needed to support the claim that the experimental results are \"precisely described by theory.\"","section":"Section III and Fig. 5"}],"minor_comments":[{"comment":"The quantity sign[n_z(θ,U,Mw)] is undefined when n_z = 0; please state the convention used (e.g., probability-zero event in the ideal Haar ensemble) or specify a numerical cutoff for the experimental analysis.","section":"Algorithm 1, line 9"},{"comment":"The caption should explicitly distinguish the dashed curves for Z_t(θ) from the dot-dashed curve for Z_{t→∞}(θ); the current legend text is easy to misread because both curve styles appear in the same figure.","section":"Fig. 5 caption"},{"comment":"The HQC accounting is confusing: the text says the 994 circuits consumed approximately 13099 HQCs while the allocation was 11000 HQCs, and then states that 834 circuits were run within that allocation; please clarify how the remaining 160 circuits were funded and how the total HQC estimate is reconciled with the allocation.","section":"Section V.D"},{"comment":"The numerical evaluation of the expectation over U and M used to obtain θ_c = 2.2142(2) should be described (e.g., Monte Carlo sample size, quadrature, or pool method), and the source of the uncertainty in the last digit should be stated, since this is presented as an exact theoretical result.","section":"Eq. (17)"},{"comment":"The phrase \"expectation values of any function of the probe qubit density matrix are invariant to any single-qubit rotation\" is imprecise; it should read \"invariant under the Haar average over single-qubit rotations,\" since a fixed rotation changes the density matrix unless the average is taken.","section":"Section IV"},{"comment":"There is a typo in \"efficiently (as as function of N)\"; it should read \"efficiently (as a function of N).\"","section":"Section V.B"}],"recommendation":"major_revision","confidential_remarks":"For the editor: this is a solid and significant protocol paper, and the main theoretical derivation is sound. The weakness identified in my major comment is not a fundamental flaw but a missing robustness check for the central experimental claim: without an error-model study, the unbiasedness of the estimator is only proven for the ideal Haar ensemble. This is fixable within the scope of the paper. I would support acceptance after the authors provide a quantitative noise analysis and a clearer statistical comparison with the experimental data. The paper fits the journal's scope and the finite-size language in Section III appropriately qualifies the 'experimental realization' claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nYou should know two things about this paper. First, it does deliver the first small-scale experimental demonstration of a postselection-free MIPT probe with universal (non-Clifford) gates, on Quantinuum H1-1, and the data match the theory curves without error mitigation. Second, the central estimator's unbiasedness is proven for the ideal Haar ensemble, and the step to noisy hardware is the soft spot: the paper asserts the noise is 'not too large' but provides no error-model simulation or systematic error bound.\n\nWhat's genuinely new: the CNOT+four-Haar-single-qubit entangling gate set (previous work used two-qubit Haar gates), the gate-set-specific critical point θc=2.2142(2) from a Fisher-KPP analysis and pool-method numerics, and the t≤4 hardware data. The derivation of Eq. (6) is careful and self-contained, and the recursive classical prediction is linear in qubit number. They also ship data and code on Zenodo, which lets anyone rerun the analysis.\n\nSoft spots, in proportion. The abstract's claim to 'experimentally realize the MIPT' oversells what t≤4 data can show; these are finite-size signatures and the paper itself mostly acknowledges that in Section III. More substantively, the unbiasedness of \\hat Z relies on the Born probabilities of the measurement record under the ideal circuit. Device noise changes those probabilities, so E[(-1)^{m0}] is no longer n_z, and the rotational-invariance step that separates |n_t| from its direction can fail. That doesn't mean the result is wrong — the agreement in Fig. 5 suggests the bias is small at this scale — but a quantified noise test would have turned a plausible assumption into a checked one. The reliance on Refs [16,25] for the linearization and λ=1 is acceptable, since those results are cited and not contended, but it does mean the paper is not fully self-contained.\n\nBottom line: this is a solid proof-of-principle, not a conclusive observation of the transition. It deserves a serious referee and likely acceptance after a revision that softens the abstract and adds a noise-bias estimate.\n\nI'd bring it to reading group and would cite it if I worked on monitored circuits.","headline":"A clean proof-of-principle for postselection-free MIPT on a tree, with honest small-size data and a real but unquantified noise caveat in the estimator's unbiasedness.","tokens_in":22784,"tokens_out":2570,"would_cite":true,"duration_ms":24080,"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":"This paper reports a postselection-free experimental observation of the measurement-induced phase transition in circuits with universal gates, using tree-shaped circuits and a linear-cost decoder.","keywords":["measurement-induced phase transition","dynamical quantum trees","postselection-free protocol","weak measurements","Haar-random circuits","trapped-ion quantum computer","purification transition","quantum circuits"],"falsifier":"Run the same protocol at fixed $t$ and $\\theta$ while injecting known decoherence into the circuits, and compare the empirical distribution of reconstructed Bloch-vector directions conditioned on their magnitudes against the uniform-on-sphere prediction; a significant correlation, or a systematic offset between $\\hat{Z}_{t,N_c}(\\theta)$ and a brute-force classical evaluation of $Z_t(\\theta)$ for the same small $t$, would falsify the unbiasedness premise.","tokens_in":21807,"feed_emoji":"⚛️","tokens_out":9178,"duration_ms":76326,"temperature":0.7,"pith_summary":"This paper claims that the measurement-induced phase transition can be observed experimentally without exponential postselection overhead even when the circuit uses universal, Haar-random gates, provided the circuit is shaped like a binary tree. It further claims that the tree's recursive structure lets a classical decoder reconstruct the needed probe state in time linear in the qubit number, and that this makes the nonlinear order parameter estimable from ordinary measurement records. Using a trapped-ion quantum computer for trees of up to four layers, the paper reports experimental curves that match theory without error mitigation, and an exact critical point $\\theta_c = 2.2142(2)$. If these claims hold, the MIPT shifts from a postselection-limited phenomenon to one that scalable quantum hardware can probe directly.","feed_headline":"Tree circuits reveal the measurement phase transition on real hardware","feed_subtitle":"Tree-shaped circuits with random gates pin the critical point at θc = 2.2142(2), matching the experiment","key_machinery":"The central object is the dynamical quantum tree: a binary-tree tensor network in which each node applies a two-qubit entangling circuit (a CNOT gate flanked by four Haar-random single-qubit unitaries) followed by weak measurements of tunable strength $\\theta$ on both output qubits. The recursion uses the fact that any node's output state is computed from two independent sub-tree outputs, and near criticality only the linear part $Z_t = A_1 Z'_{t-1} + A_2 Z''_{t-1} + O(Z^2)$ matters; the statistics of the coefficients $A_1,A_2$ determine a wavefront speed $v_\\theta$ whose zero defines the critical point via $\\mathbb{E}[A_1+A_2]=1$. A second load-bearing element is the collapse-process decoder: reversing the flow of time and feeding states back through the tree reconstructs the probe Bloch vector with cost linear in the qubit number, which is what makes the postselection-free estimator possible.","core_discovery":"On the paper's own terms, the central discovery is that the MIPT in a dynamical quantum tree can be detected through the order parameter $Z_t(\\theta) = \\mathbb{E}_{U,M_w}[Z_t(\\theta,U,M_w)]$, the ensemble-averaged smallest eigenvalue of the probe qubit's reduced state, rather than through the typical value $Z_t^{\\mathrm{typ}}(\\theta)$ that requires postselection. The key identity is $Z_t(\\theta) = \\mathbb{E}[\\tfrac12 - (-1)^{m_0}/\\operatorname{sign}(n_t^z)]$, so the order parameter can be estimated by comparing one measured root-qubit outcome $m_0$ with a classically reconstructed Bloch vector $\\vec{n}_t$ obtained by running the tree in the collapse direction. Because the tree is recursive, the reconstruction costs $O(N)$ and never requires the same measurement record twice. For this ensemble of weak measurements with Kraus operators $K_m = \\sin(\\theta/2)\\,I + [\\cos(\\theta/2)-\\sin(\\theta/2)]\\,|m\\rangle\\langle m|$, the recursion for the order parameter linearizes near the transition and yields the exact critical point $\\theta_c = 2.2142(2)$ and the critical scaling $\\ln Z_t^{\\mathrm{typ}} \\sim -t^{1/3}$. The experimental data for $t \\le 4$ follow the simulated $Z_t(\\theta)$ curves and approach the $t\\to\\infty$ asymptotic curve, which the paper presents as closing the gap between analytical theory and postselection-free experimental observation of the MIPT with universal gates.","pith_inferences":["A natural extension not performed in the paper is to test the independence of Bloch-vector direction and magnitude directly by comparing the empirical joint distribution of $(|\\vec{n}_t|, \\hat{n}_t)$ from the collapse decoder with the uniform-on-sphere assumption; a detectable correlation would signal a bias in the estimator.","The same recursive decoding scheme should extend to other recursive tensor geometries and to quantum trees with Abelian or non-Abelian symmetries, where sharpening transitions are predicted to accompany the MIPT.","A sharper falsification would come from running the protocol on larger trees (for example $t=5$ or $t=6$) and checking whether the crossing point moves toward $\\theta_c$ at the predicted rate and whether $\\ln Z_t^{\\mathrm{typ}}$ obeys the predicted $-t^{1/3}$ scaling at criticality.","The linear-cost decoder suggests a practical device-noise diagnostic: agreement between hardware results and pool-method simulations of the same circuit ensemble could be monitored as a function of noise levels without any error mitigation."],"forward_implications":["The MIPT order parameter can be measured for circuits with universal, Haar-random gates without exponentially many circuit repetitions or an exponential classical simulation, so experiments can scale with the available number of qubits.","The tree geometry provides an exactly solvable critical point $\\theta_c \\approx 2.2142$ and a predicted universality class, giving a precision benchmark for quantum hardware.","Experimental data on a trapped-ion machine for trees up to four layers match theory within 95% confidence intervals without error mitigation, suggesting the protocol is robust to realistic device noise.","Because the estimator averages linear quantities, a single measurement record per circuit suffices, and data for smaller trees can be recycled by truncating larger measurement records.","The phase transition can be framed as a predictability transition, predicting the root qubit's initial state from the mid-circuit weak-measurement record, so the probe qubit is not actually required in the experiment."],"supporting_citations":[{"why":"Establishes the dynamical quantum tree model and the analytical and numerical framework for its measurement-induced phase transition, including the recursion the paper adapts.","marker":"[16]"},{"why":"Supplies the postselection-free decoding idea and the relation between the averaged order parameter $Z_t(\\theta)$ and the typical value; Algorithm 1 and the critical-point analysis build on it.","marker":"[25]"},{"why":"The modified analytical approach for the MIPT on quantum trees, including symmetry generalizations, which the paper specializes to its entangling circuit.","marker":"[45]"},{"why":"The pool method used to generate the simulated theory curves and the $t\\to 800$ extrapolation shown in the figures.","marker":"[47-54]"},{"why":"The trapped-ion quantum computer on which the experimental circuits were executed.","marker":"[55]"},{"why":"Public release of the measurement data and code, allowing reproduction of the experimental estimate of the order parameter.","marker":"[60]"}],"fun_headline_variants":["Tree circuits reveal MIPT without postselection on ion trap","Exact MIPT critical point found in tree circuits at 2.2142","Weak-measurement trees give postselection-free MIPT observation","Quantum tree circuits measure MIPT, no postselection required"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The estimate is unbiased only if the direction of the reconstructed Bloch vector is statistically independent of its length and uniformly distributed on the sphere, as the paper assumes follows from Haar invariance; if device noise or finite-size correlations break that independence, the experimental order parameter would be a biased estimator of $Z_t(\\theta)$.","fun_headline_variants_meta":{"raw":{"variants":["Tree circuits reveal MIPT without postselection on ion trap","Exact MIPT critical point found in tree circuits at 2.2142","Weak-measurement trees give postselection-free MIPT observation","Quantum tree circuits measure MIPT, no postselection required"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000231,"raw_usage":{"total_tokens":1600,"prompt_tokens":1173,"completion_tokens":427,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":789,"completion_tokens_details":{"reasoning_tokens":354}},"tokens_in":789,"tokens_out":427,"duration_ms":4491,"temperature":1.0,"reasoning_tokens":354,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T14:40:39.579090+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same protocol at fixed $t$ and $\\theta$ while injecting known decoherence into the circuits, and compare the empirical distribution of reconstructed Bloch-vector directions conditioned on their magnitudes against the uniform-on-sphere prediction; a significant correlation, or a systematic offset between $\\hat{Z}_{t,N_c}(\\theta)$ and a brute-force classical evaluation of $Z_t(\\theta)$ for the same small $t$, would falsify the unbiasedness premise.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the postselection-free decoding idea and the relation between the averaged order parameter $Z_t(\\theta)$ and the typical value; Algorithm 1 and the critical-point analysis build on it."},{"cited_title":"Charge and Spin Sharpening Transitions on Dynamical Quantum Trees","cited_arxiv_id":"2405.13894","evidence_quote":"The modified analytical approach for the MIPT on quantum trees, including symmetry generalizations, which the paper specializes to its entangling circuit."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The trapped-ion quantum computer on which the experimental circuits were executed."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Public release of the measurement data and code, allowing reproduction of the experimental estimate of the order parameter."}],"review_version":1}