Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

A neural network trained on measurement records alone can estimate measurement-induced entanglement, but only up to a critical circuit depth; beyond it, polynomial-resource learning fails.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 19:15 UTC pith:Q5ANGJUJ

load-bearing objection A useful empirical study of a transformer-based MIE estimator with an over-claimed abstract; the learnability transition is real for this model family, but the fundamental and 'coincides' statements outrun the evidence. the 4 major comments →

arxiv 2512.01317 v3 pith:Q5ANGJUJ submitted 2025-12-01 quant-ph cond-mat.dis-nncs.AI

Data-Driven Learnability Transition of Measurement-Induced Entanglement

Dongheng Qian , Jing Wang This is my paper
classification quant-ph cond-mat.dis-nncs.AI
keywords measurement-induced entanglementlearnability transitionclassical shadow tomographytransformer neural networkrandom quantum circuitspostselection-free estimationcomputational phase transitionnoisy quantum devices
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper asks whether measurement-induced entanglement (MIE) — normally hard to measure because it requires post-selection over exponentially many outcomes — can be extracted from measurement data alone without knowing how the state was prepared. It trains a transformer on binary measurement records to output the two-qubit post-measurement density matrix, and uses a shadow-based uncertainty metric Δ that bounds the gap between upper and lower estimates of the entanglement. For random circuits (1D all-to-all and 2D nearest-neighbor), the authors find a sharp learnability transition: at shallow depth, Δ decreases as measurement data and model parameters grow (polynomial resource scaling), while beyond a critical depth Δ saturates near 2 ln 2 even as resources grow, implying exponential cost. They locate the transition at finite depth via an L=20 vs L=24 crossing and identify it with the teleportation/computational phase boundary where efficient classical simulation breaks down. The same signature survives on noisy quantum hardware, suggesting the limit is practical, not just theoretical.

Core claim

The central claim is that MIE becomes learnable or unlearnable depending on circuit depth, and the boundary is a computational phase transition. Concretely, for a pure L-qubit state with qubits A and B held out and all others measured, the true post-measurement entanglement S_A,m is bracketed by upper and lower bounds whose width is the quantum-classical entropy S^QC_AB,m; averaging over measurement outcomes and classical shadow snapshots gives Δ = E_m[S^QC_AB,m]. A transformer with a density-matrix head trained on measurement records (loss L = -E[2Tr(ρ σ^s) - Tr(ρ²)]) produces an estimator ρ_AB,m; Δ then measures how uncertain the MIE estimate remains. The result: for shallow random circuit

What carries the argument

The key object is the uncertainty metric Δ = E_m E_s[S^SC_AB,m] = E_m[S^QC_AB,m], which upper-bounds the interval width containing the true post-measurement entanglement for each outcome. It is made experimentally accessible by measuring A and B in random Pauli bases (classical shadow tomography), so the same shots that produce the record m also produce the snapshot σ^s_AB,m. The learning machine is a transformer encoder with a density-matrix head that outputs ρ_AB,m via an AA† construction (trace-normalized, ε-mixed), trained self-supervised with L(θ) = -E_mE_s[2Tr(ρ_AB,m σ^s_AB,m) - Tr(ρ²_AB,m)]. The mechanism is that this loss is minimized to -1 exactly when the estimator reproduces the t

Load-bearing premise

The load-bearing premise is that failure of this particular transformer (with these losses and data sizes) to reduce Δ at large depth proves that no polynomial-resource classical learner can extract MIE; the paper offers no lower bound for this learning problem.

What would settle it

A definitive falsifier: if a classical learning algorithm using only the same measurement records drives Δ below the saturation value with polynomial N_m and N_p in the claimed unlearnable phase, the fundamental-limit interpretation collapses. Alternatively, a direct measurement of a teleportation order parameter in the same circuits that yields a critical depth different from the Δ crossing would disprove the claimed coincidence with the classical-simulation boundary.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • In the learnable (shallow) regime, MIE can be estimated with polynomial measurement data and model parameters, so data-driven estimators can replace classical simulation-based decoders and provide practical MIE probes on near-term devices.
  • In the unlearnable (deep) regime, substantial MIE may still be present, but it is not detectable from measurement records with polynomial resources; this is framed as evidence for quantum computational advantage.
  • The transition is finite-depth rather than a finite-size artifact, as shown by the crossing of the Δ curves for L=20 and L=24.
  • The same learnable/unlearnable behavior occurs for 1D all-to-all and 2D nearest-neighbor random circuits, and remains visible under realistic noise, so the boundary is robust to connectivity and hardware imperfections.
  • At large depth the model's output approaches a maximally mixed state (Δ saturates at 2 ln 2), and adding more samples or parameters does not help — it can even hurt due to overfitting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The paper's evidence for a fundamental limit is empirical: no information-theoretic lower bound is proven for this specific learning task, so the 'unlearnable' phase may be a property of this transformer family rather than of all polynomial-resource learners.
  • Because Δ is itself an ensemble average over outcomes, it is exactly the type of quantity that existing no-go theorems allow; the truly hard object is the outcome-conditioned map m ↦ σ_AB,m. A sharper test would try to learn single-outcome post-measurement states rather than averages.
  • The claimed coincidence with the teleportation/classical-simulation boundary could be tested directly by measuring a teleportation order parameter in the same circuits and comparing its critical depth with the Δ crossing.
  • Extending the protocol to structured states (e.g., ground states or stabilizer states), or including training time in the resource count, could sharpen or shift the boundary; the paper leaves both to future work.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes a self-supervised, transformer-based method to estimate measurement-induced entanglement (MIE) from raw measurement records, without post-selection or prior knowledge of the state-preparation circuit. The model maps a binary measurement outcome m to a 4x4 density matrix ρ_AB,m approximating the true post-measurement state σ_AB,m, trained by minimizing a Hilbert–Schmidt-distance-based loss built from classical-shadow snapshots. The authors define an uncertainty metric Δ = E_m[S^QC_AB,m], which upper-bounds the width of the interval containing the true MIE, and benchmark the trained model against independent shadow samples. For random 1D all-to-all and 2D nearest-neighbor circuits, they observe that at shallow depth Δ decreases as the number of measurement shots N_m or model parameters N_p grows, while at larger depth Δ saturates near 2 log(2) or grows. They interpret this as a learnability phase transition that coincides with the breakdown of efficient classical simulation, and they report supporting noisy simulations and IBM quantum-device experiments.

Significance. If the central claims hold, the paper would provide both a practical, scalable estimator for MIE in the shallow-depth regime and evidence for a computational learnability transition tied to measurement-induced phases. The core mathematical construction is sound: Eq. (3) shows that the training loss is exactly E_m[||ρ_AB,m − σ_AB,m||_HS^2] up to a constant, so the optimization genuinely targets state reconstruction, and Δ is evaluated on independent shadow samples against the true post-measurement states, avoiding self-referential benchmarking. The self-supervised protocol, using only measurement records, is a useful and nontrivial contribution. However, the paper's headline claims—that the unlearnable phase is fundamental, that required resources grow exponentially, and that the transition coincides with the classical-simulation boundary—are not supported by the evidence presented. The observed saturation of Δ for one transformer family does not rule out other polynomial-resource learners, and no scaling with system size is shown to substantiate 'exponential' resource growth.

major comments (4)
  1. [§Learnability transitions, Figs. 2–3] The claim of an 'unlearnable phase' is presented as a property of the learning problem itself, but the evidence is failure of a single BERT-style transformer family. The saturation of Δ near 2 log(2) is exactly what a constant estimator outputting the maximally mixed state would produce; it shows this model cannot exploit m-dependence in the deep phase, but it does not establish that no polynomial-resource learner could. Ref. [53]'s no-go is invoked, but Δ is an ensemble average, which that result explicitly permits, while the hard object is the conditional map m ↦ σ_AB,m. To support the strong conclusion, the authors need either a lower bound for the specific estimation problem or a significantly weaker, architecture-dependent phrasing.
  2. [Abstract and §Learnability transitions] The abstract states that 'above it the required resources grow exponentially,' but no resource-scaling analysis with system size is performed. All training experiments fix L = 20 (or 5×5) and vary N_m, N_p over a finite range; saturation at fixed L does not distinguish exponential resource growth from any non-improving ceiling (e.g., a constant information bottleneck). The claim would require measuring the minimal N_m and N_p needed to reach a target Δ as a function of L, or a scaling argument. As written, 'exponentially' is not supported by the data.
  3. [§Learnability transitions, paragraph after Fig. 1(d); §Discussions] The identification of the observed crossover with the finite-depth teleportation transition and with the breakdown of efficient classical simulation is asserted rather than demonstrated. No teleportation order parameter (e.g., decoder mutual information) is measured for the same circuits, and no comparison of critical depths is made. Refs. [62,63,66] establish related phenomena in different settings, but they do not automatically imply the claimed 'coincides with' relationship. A concrete test would be to compute a teleportation order parameter for these exact random circuits and compare its threshold with the saturation depth of Δ; otherwise the claim should be removed or substantially softened.
  4. [Fig. 1(d) and §Learnability transitions] The finite-size evidence for a genuine phase transition is thin: the only system-size comparison is a single crossing between L = 20 and L = 24. The authors acknowledge that accurately determining the critical depth requires larger L, but the central claim of a 'transition' rests on this crossing. To justify the phase-transition language, the paper should include more system sizes and, ideally, a finite-size scaling or data-collapse analysis. Absent that, the observations are more accurately described as a crossover.
minor comments (4)
  1. [Eq. (3) and surrounding text] The sentence 'The first equality saturates if and only if ρ_AB,m = σ_AB,m for all m' is confusing because Eq. (3) is an inequality chain with no displayed first equality. The intended meaning is likely that the bound in Eq. (3) is saturated. Please rephrase.
  2. [Abstract and Fig. 1(b)] The phrase 'one-dimensional all-to-all connectivity' is oxymoronic. If the circuit applies Haar-random two-qubit gates between uniformly chosen pairs (all-to-all), the system is not geometrically one-dimensional. Clarify the connectivity model.
  3. [Fig. 4(c,d)] The experimental data on ibm_marrakesh use M = 5 circuit realizations and Ne = 5×10^3. No error bars or statistical uncertainty are shown for the hardware results; please add them or state why they are negligible.
  4. [Supplemental Material S3] The analysis of the L2 loss and the flat-entanglement-spectrum requirement is interesting but appears only in the supplement. Since it concerns a potential alternative training objective, a brief mention in the main text (with a pointer to the supplement) would improve accessibility.

Circularity Check

0 steps flagged

No significant circularity: the learned estimator is evaluated against true post-measurement states via classical shadows, and the loss is shown to minimize Hilbert–Schmidt distance to the target.

full rationale

The paper's derivation chain is self-contained and benchmarked externally. It defines Δ as the shadow-classical entropy of the learned estimator ρ_AB,m relative to the true post-measurement state σ_AB,m (Eq. 2), and constructs a loss L(θ) (Eq. 4) whose minimizer provably satisfies ρ_AB,m = σ_AB,m, with the equality conditions stated in Eq. (3). Δ is then evaluated on independent shots (Ne) after training, so the reported behavior is not a fitted parameter renamed as a prediction. The 'unlearnable phase' is inferred from saturation or growth of Δ as Nm and Np increase for a specific transformer family, and the supplemental material checks an alternative loss L1(θ), so the central observation is not forced by the definition of Δ alone. The identification of the transition with the breakdown of efficient classical simulation and with the finite-depth teleportation transition is an interpretive comparison supported by external references [62,63,66]; whether that identification is fully justified is a correctness or over-interpretation concern, not a circularity. The paper's self-citations ([37,48,50]) are peripheral and do not carry any load-bearing step, and no uniqueness theorem from the authors' prior work is invoked. The paper even notes the limitations that accurate critical-depth determination requires larger L and that training time is not included in the resource accounting; these are honest caveats rather than circular reasoning. No specific equation or construction reduces to its own inputs, so the circularity score is 0.

Axiom & Free-Parameter Ledger

2 free parameters · 5 axioms · 0 invented entities

The derivation is lean: no continuously fitted parameters enter the central claim; the phase boundary is read from data rather than fitted. The ledger's weight sits in domain assumptions — the teleportation/complexity identification and the simulability-implies-learnability link — plus standard shadow-tomography machinery. No invented entities: the 'learnability transition' is a label for an observed empirical threshold, and the ε-mixing is a numerical regularizer, not a physical mechanism.

free parameters (2)
  • ε-mixing parameter = 10^-4
    Hand-chosen constant mixed into the output density matrix (ρ = (1−ε)ρ_norm + εI) to stabilize training near zero eigenvalues (Supplement S1); technically an ad hoc parameter but too small to affect the transition location materially.
  • Transformer hyperparameters (dh, df, Nl; dropout; learning rate) = 20K–540K parameter models; η0=5×10^-4; dropout 0.1
    Chosen by hand and swept as the resource axis N_p; the transition could in principle shift with architecture, which the paper treats as the resource dimension rather than a fitted constant.
axioms (5)
  • standard math Relative-entropy nonnegativity and monotonicity under partial trace justify the two-sided bounds in Eq. (1) for S_A,m.
    Invoked in 'Quantifying learnability' to define Δ as the MIE uncertainty interval width; mathematically sound and uncontroversial.
  • domain assumption Classical shadow tomography yields unbiased estimators σ^s_AB,m of the conditional post-measurement state.
    The statistical machinery from Ref. [58]; standard but load-bearing for the claim that Δ is experimentally accessible.
  • domain assumption 1D all-to-all and 2D nearest-neighbor random circuits exhibit a finite-depth teleportation transition below which states are classically simulable and above which they are not (Refs. [62,63,66]).
    The identification of the observed resource threshold with a true computational phase boundary imports these results; the paper does not measure the teleportation order parameter.
  • domain assumption Polynomial-resource classical-simulation failure implies failure of polynomial-resource machine learning of the post-measurement state.
    The inference 'efficient classical simulation is believed to fail here... Therefore, polynomial-resource machine-learning models are not expected to succeed' is a conjecture chain, not a derived equivalence; Ref. [53]'s no-go does not directly apply because Δ is an ensemble average.
  • domain assumption Ref. [64]'s classical-simulability-implies-efficient-learnability result applies to these random-circuit ensembles in the shallow regime.
    Cited as 'states are expected to admit efficient learning [64]'; the hedge 'expected' acknowledges the theorem's conditions are not verified for these specific ensembles.

pith-pipeline@v1.3.0-alltime-deepseek · 14453 in / 27369 out tokens · 292330 ms · 2026-08-03T19:15:57.609652+00:00 · methodology

0 comments
read the original abstract

Measurement-induced entanglement (MIE) captures how local measurements generate long-range quantum correlations and drive dynamical phase transitions in many-body systems. Yet estimating MIE experimentally remains challenging: direct evaluation requires extensive post-selection over measurement outcomes, raising the question of whether MIE is accessible with only polynomial resources. We address this challenge by reframing MIE detection as a data-driven learning problem that assumes no prior knowledge of state preparation. Using measurement records alone, we train a neural network in a self-supervised manner to predict the uncertainty metric for MIE--the gap between upper and lower bounds of the average post-measurement bipartite entanglement. Applied to random circuits with one-dimensional all-to-all connectivity, our method reveals a learnability transition with increasing circuit depth: below a threshold the MIE can be effectively learned with resources that grow only polynomially with system size, whereas above it the required resources grow exponentially. This computational phase transition coincides with the breakdown of efficient classical simulation of the underlying quantum state. We further observe signatures of this transition on current noisy quantum devices. These results highlight the power of data-driven approaches for learning MIE and delineate the practical limits of its classical learnability.

Figures

Figures reproduced from arXiv: 2512.01317 by Dongheng Qian, Jing Wang.

Figure 1
Figure 1. Figure 1: FIG. 1. Setup and learnability transition. (a) Schematic [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Learnability transition in 1D random all-to-all cir [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Learnability transition in 1D all-to-all circuits with [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Simple broadband signal detection at the fundamental limit

    quant-ph 2026-01 conditional novelty 6.0

    Broadband AC-field detection at the Grover-like limit can be achieved by a single analog experiment using a randomized SSH control Hamiltonian and a GHZ probe, with the lower bound derived from an integrated-quantum-F...

Reference graph

Works this paper leans on

79 extracted references · 7 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Horodecki, P

    R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, Quantum entanglement, Rev. Mod. Phys. 81, 865 (2009)

    2009

  2. [2]

    M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information (Cambridge University Press, Cambridge, 2010)

    2010

  3. [3]

    Wilde, Quantum Information Theory , 2nd ed

    M. Wilde, Quantum Information Theory , 2nd ed. (Cam- bridge University Press, Cambridge, 2017)

    2017

  4. [4]

    C. L. Degen, F. Reinhard, and P. Cappellaro, Quantum sensing, Rev. Mod. Phys. 89, 035002 (2017)

    2017

  5. [5]

    R. R. Allen, F. Machado, I. L. Chuang, H.-Y. Huang, and S. Choi, Quantum computing enhanced sensing, arXiv:2501.07625 (2025)

    Pith/arXiv arXiv 2025

  6. [6]

    C. H. Bennett, G. Brassard, C. Cr´ epeau, R. Jozsa, A. Peres, and W. K. Wootters, Teleporting an unknown quantum state via dual classical and einstein-podolsky- rosen channels, Phys. Rev. Lett. 70, 1895 (1993)

    1993

  7. [7]

    M. Popp, F. Verstraete, M. A. Mart´ ın-Delgado, and J. I. Cirac, Localizable entanglement, Phys. Rev. A 71, 042306 (2005)

    2005

  8. [8]

    M. A. Rajabpour, Post-measurement bipartite entangle- ment entropy in conformal field theories, Phys. Rev. B 92, 075108 (2015)

    2015

  9. [9]

    C.-J. Lin, W. Ye, Y. Zou, S. Sang, and T. H. Hsieh, Probing sign structure using measurement-induced en- tanglement, Quantum 7, 910 (2023)

    2023

  10. [10]

    H. J. Briegel, D. E. Browne, W. D¨ ur, R. Raussendorf, and M. Van den Nest, Measurement-based quantum compu- tation, Nat. Phys. 5, 19 (2009)

    2009

  11. [11]

    H. J. Briegel and R. Raussendorf, Persistent entangle- ment in arrays of interacting particles, Phys. Rev. Lett. 86, 910 (2001)

    2001

  12. [12]

    Piroli, G

    L. Piroli, G. Styliaris, and J. I. Cirac, Quantum circuits assisted by local operations and classical communication: Transformations and phases of matter, Phys. Rev. Lett. 127, 220503 (2021)

    2021

  13. [13]

    J. Y. Lee, W. Ji, Z. Bi, and M. Fisher, Decod- ing measurement-prepared quantum phases and transi- tions: From ising model to gauge theory, and beyond, arXiv:2208.11699 (2022)

    Pith/arXiv arXiv 2022

  14. [14]

    G.-Y. Zhu, N. Tantivasadakarn, A. Vishwanath, 6 S. Trebst, and R. Verresen, Nishimori’s cat: Stable long- range entanglement from finite-depth unitaries and weak measurements, Phys. Rev. Lett. 131, 200201 (2023)

    2023

  15. [15]

    Cowsik, M

    A. Cowsik, M. Ippoliti, and X.-L. Qi, Engineering en- tanglement geometry via spacetime-modulated measure- ments, Phys. Rev. D 112, L081904 (2025)

    2025

  16. [16]

    S.-K. Jian, C. Liu, X. Chen, B. Swingle, and P. Zhang, Measurement-induced phase transition in the monitored sachdev-ye-kitaev model, Phys. Rev. Lett. 127, 140601 (2021)

    2021

  17. [17]

    Cheng, R

    Z. Cheng, R. Wen, S. Gopalakrishnan, R. Vasseur, and A. C. Potter, Universal structure of measurement- induced information in many-body ground states, Phys. Rev. B 109, 195128 (2024)

    2024

  18. [18]

    Ippoliti, M

    M. Ippoliti, M. J. Gullans, S. Gopalakrishnan, D. A. Huse, and V. Khemani, Entanglement phase transitions in measurement-only dynamics, Phys. Rev. X 11, 011030 (2021)

    2021

  19. [19]

    Lavasani, Y

    A. Lavasani, Y. Alavirad, and M. Barkeshli, Measurement-induced topological entanglement transi- tions in symmetric random quantum circuits, Nat. Phys. 17, 342 (2021)

    2021

  20. [20]

    Klocke, D

    K. Klocke, D. Simm, G.-Y. Zhu, S. Trebst, and M. Buch- hold, Entanglement dynamics in monitored kitaev cir- cuits: Loop models, symmetry classification, and quan- tum lifshitz scaling, Phys. Rev. B 111, 224301 (2025)

    2025

  21. [21]

    Y. Li, X. Chen, and M. P. A. Fisher, Measurement- driven entanglement transition in hybrid quantum cir- cuits, Phys. Rev. B 100, 134306 (2019)

    2019

  22. [22]

    Skinner, J

    B. Skinner, J. Ruhman, and A. Nahum, Measurement- induced phase transitions in the dynamics of entangle- ment, Phys. Rev. X 9, 031009 (2019)

    2019

  23. [23]

    Szyniszewski, A

    M. Szyniszewski, A. Romito, and H. Schomerus, Entan- glement transition from variable-strength weak measure- ments, Phys. Rev. B 100, 064204 (2019)

    2019

  24. [24]

    Vasseur, A

    R. Vasseur, A. C. Potter, Y.-Z. You, and A. W. W. Lud- wig, Entanglement transitions from holographic random tensor networks, Phys. Rev. B 100, 134203 (2019)

    2019

  25. [25]

    Nahum and B

    A. Nahum and B. Skinner, Entanglement and dynamics of diffusion-annihilation processes with majorana defects, Phys. Rev. Research 2, 023288 (2020)

    2020

  26. [26]

    Zabalo, M

    A. Zabalo, M. J. Gullans, J. H. Wilson, S. Gopalakrish- nan, D. A. Huse, and J. H. Pixley, Critical properties of the measurement-induced transition in random quantum circuits, Phys. Rev. B 101, 060301 (2020)

    2020

  27. [27]

    Jian, Y.-Z

    C.-M. Jian, Y.-Z. You, R. Vasseur, and A. W. W. Ludwig, Measurement-induced criticality in random quantum cir- cuits, Phys. Rev. B 101, 104302 (2020)

    2020

  28. [28]

    Y. Bao, S. Choi, and E. Altman, Theory of the phase transition in random unitary circuits with measurements, Phys. Rev. B 101, 104301 (2020)

    2020

  29. [29]

    S. Choi, Y. Bao, X.-L. Qi, and E. Altman, Quan- tum Error Correction in Scrambling Dynamics and Measurement-Induced Phase Transition, Phys. Rev. Lett. 125, 030505 (2020)

    2020

  30. [30]

    Turkeshi, R

    X. Turkeshi, R. Fazio, and M. Dalmonte, Measurement- induced criticality in (2+1)-dimensional hybrid quantum circuits, Phys. Rev. B 102, 014315 (2020)

    2020

  31. [31]

    Alberton, M

    O. Alberton, M. Buchhold, and S. Diehl, Entanglement Transition in a Monitored Free-Fermion Chain: From Ex- tended Criticality to Area Law, Phys. Rev. Lett. 126, 170602 (2021)

    2021

  32. [32]

    Sharma, X

    S. Sharma, X. Turkeshi, R. Fazio, and M. Dalmonte, Measurement-induced criticality in extended and long- range unitary circuits, SciPost Phys. Core 5, 023 (2022)

    2022

  33. [33]

    Block, Y

    M. Block, Y. Bao, S. Choi, E. Altman, and N. Y. Yao, Measurement-induced transition in long-range in- teracting quantum circuits, Phys. Rev. Lett. 128, 010604 (2022)

    2022

  34. [34]

    Sierant, M

    P. Sierant, M. Schir` o, M. Lewenstein, and X. Turkeshi, Measurement-induced phase transitions in ( d + 1)- dimensional stabilizer circuits, Phys. Rev. B 106, 214316 (2022)

    2022

  35. [35]

    M. P. A. Fisher, V. Khemani, A. Nahum, and S. Vijay, Random Quantum Circuits,Annu. Rev. Condens. Matter Phys. 14, 335 (2023)

    2023

  36. [36]

    Poboiko, I

    I. Poboiko, I. V. Gornyi, and A. D. Mirlin, Measurement- induced phase transition for free fermions above one di- mension, Phys. Rev. Lett. 132, 110403 (2024)

    2024

  37. [37]

    Qian and J

    D. Qian and J. Wang, Protect measurement-induced phase transition from noise, Phys. Rev. Lett. 134, 020403 (2025)

    2025

  38. [38]

    J. M. Koh, S.-N. Sun, M. Motta, and A. J. Minnich, Measurement-induced entanglement phase transition on a superconducting quantum processor with mid-circuit readout, Nat. Phys. 19, 1314 (2023)

    2023

  39. [39]

    M. J. Gullans and D. A. Huse, Dynamical purifica- tion phase transition induced by quantum measurements, Phys. Rev. X 10, 041020 (2020)

    2020

  40. [40]

    M. J. Gullans and D. A. Huse, Scalable probes of measurement-induced criticality, Phys. Rev. Lett. 125, 070606 (2020)

    2020

  41. [41]

    C. Noel, P. Niroula, D. Zhu, A. Risinger, L. Egan, D. Biswas, M. Cetina, A. V. Gorshkov, M. J. Gullans, D. A. Huse, and C. Monroe, Measurement-induced quan- tum phases realized in a trapped-ion quantum computer, Nat. Phys. 18, 760 (2022)

    2022

  42. [42]

    Dehghani, A

    H. Dehghani, A. Lavasani, M. Hafezi, and M. J. Gullans, Neural-network decoders for measurement induced phase transitions, Nature Comm. 14, 2918 (2023)

    2023

  43. [43]

    J. C. Hoke et al. , Measurement-induced entanglement and teleportation on a noisy quantum processor, Nature 622, 481 (2023)

    2023

  44. [44]

    Agrawal, J

    U. Agrawal, J. Lopez-Piqueres, R. Vasseur, S. Gopalakr- ishnan, and A. C. Potter, Observing quantum measure- ment collapse as a learnability phase transition, Phys. Rev. X 14, 041012 (2024)

    2024

  45. [45]

    Ippoliti and V

    M. Ippoliti and V. Khemani, Learnability transitions in monitored quantum dynamics via eavesdropper’s classi- cal shadows, PRX Quantum 5, 020304 (2024)

    2024

  46. [46]

    Y. Li, Y. Zou, P. Glorioso, E. Altman, and M. P. A. Fisher, Cross entropy benchmark for measurement- induced phase transitions, Phys. Rev. Lett. 130, 220404 (2023)

    2023

  47. [47]

    Kamakari, J

    H. Kamakari, J. Sun, Y. Li, J. J. Thio, T. P. Gujarati, M. P. A. Fisher, M. Motta, and A. J. Minnich, Exper- imental demonstration of scalable cross-entropy bench- marking to detect measurement-induced phase transi- tions on a superconducting quantum processor, Phys. Rev. Lett. 134, 120401 (2025)

    2025

  48. [48]

    Qian and J

    D. Qian and J. Wang, Coherent information phase tran- sition in a noisy quantum circuit, Phys. Rev. B 112, L180301 (2025)

    2025

  49. [49]

    Turkeshi, Measurement-induced criticality as a data- structure transition, Phys

    X. Turkeshi, Measurement-induced criticality as a data- structure transition, Phys. Rev. B 106, 144313 (2022)

    2022

  50. [50]

    Qian and J

    D. Qian and J. Wang, Steering-induced phase transi- tion in measurement-only quantum circuits, Phys. Rev. B 109, 024301 (2024). 7

    2024

  51. [51]

    H. Kim, A. Kumar, Y. Zhou, Y. Xu, R. Vasseur, and E.-A. Kim, Learning measurement-induced phase transi- tions using attention, arXiv:2508.15895 (2025)

    Pith/arXiv arXiv 2025

  52. [52]

    S. J. Garratt and E. Altman, Probing postmeasurement entanglement without postselection, PRX Quantum 5, 030311 (2024)

    2024

  53. [53]

    McGinley, Postselection-free learning of measurement-induced quantum dynamics, PRX Quan- tum 5, 020347 (2024)

    M. McGinley, Postselection-free learning of measurement-induced quantum dynamics, PRX Quan- tum 5, 020347 (2024)

    2024

  54. [54]

    Huang, R

    H.-Y. Huang, R. Kueng, and J. Preskill, Information- theoretic bounds on quantum advantage in machine learning, Phys. Rev. Lett. 126, 190505 (2021)

    2021

  55. [55]

    Huang, R

    H.-Y. Huang, R. Kueng, G. Torlai, V. V. Albert, and J. Preskill, Provably efficient machine learning for quan- tum many-body problems, Science 377, eabk3333 (2022)

    2022

  56. [56]

    Y. Du, Y. Zhu, Y.-H. Zhang, M.-H. Hsieh, P. Rebentrost, W. Gao, Y.-D. Wu, J. Eisert, G. Chiribella, D. Tao,et al., Artificial intelligence for representing and characterizing quantum systems, arXiv:2509.04923 (2025)

    Pith/arXiv arXiv 2025

  57. [58]

    Huang, R

    H.-Y. Huang, R. Kueng, and J. Preskill, Predicting many properties of a quantum system from very few measure- ments, Nat. Phys. 16, 1050 (2020)

    2020

  58. [59]

    See Supplemental Material for more details

  59. [60]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, /suppress L. Kaiser, and I. Polosukhin, Attention is all you need, in Proceedings of the 31st In- ternational Conference on Neural Information Processing Systems , Vol. 30 (2017) pp. 6000–6010

    2017

  60. [61]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, Bert: Pre-training of deep bidirectional transformers for lan- guage understanding, in Proceedings of the 2019 confer- ence of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers) (2019) pp. 4171–4186

    2019

  61. [62]

    Y. Bao, M. Block, and E. Altman, Finite-Time Telepor- tation Phase Transition in Random Quantum Circuits, Phys. Rev. Lett. 132, 030401 (2024)

    2024

  62. [63]

    J. C. Napp, R. L. La Placa, A. M. Dalzell, F. G. S. L. Brand˜ ao, and A. W. Harrow, Efficient classical simula- tion of random shallow 2d quantum circuits, Phys. Rev. X 12, 021021 (2022)

    2022

  63. [64]

    Yoganathan, A condition under which classi- cal simulability implies efficient state learnability, arXiv:1907.08163 (2019)

    M. Yoganathan, A condition under which classi- cal simulability implies efficient state learnability, arXiv:1907.08163 (2019)

    Pith/arXiv arXiv 1907

  64. [65]

    L. G. Valiant, A theory of the learnable, Commun. ACM 27, 1134–1142 (1984)

    1984

  65. [66]

    Arute, K

    F. Arute, K. Arya, R. Babbush, D. Bacon, J. C. Bardin, R. Barends, R. Biswas, S. Boixo, F. G. Brandao, D. A. Buell, et al., Quantum supremacy using a programmable superconducting processor, Nature 574, 505 (2019)

    2019

  66. [67]

    S.-B. B. Lee, H. R. Choi, D. D. Ohm, and S.-S. B. Lee, Scalable projected entangled-pair state representa- tion of random quantum circuit states, Phys. Rev. Res. 7, 033252 (2025)

    2025

  67. [68]

    Bene Watts, D

    A. Bene Watts, D. Gosset, Y. Liu, and M. Soleimanifar, Quantum advantage from measurement-induced entan- glement in random shallow circuits, PRX Quantum 6, 010356 (2025)

    2025

  68. [69]

    McGinley, W

    M. McGinley, W. W. Ho, and D. Malz, Measurement- induced entanglement and complexity in random constant-depth 2d quantum circuits, Phys. Rev. X 15, 021059 (2025)

    2025

  69. [70]

    Cao and J

    C. Cao and J. Eisert, Measurement-driven quantum ad- vantages in shallow circuits, arXiv:2505.04705 (2025)

    arXiv 2025

  70. [71]

    AbuGhanem, Ibm quantum computers: evolution, performance, and future directions, J

    M. AbuGhanem, Ibm quantum computers: evolution, performance, and future directions, J. Supercomput. 81, 687 (2025)

    2025

  71. [72]

    Torlai and R

    G. Torlai and R. G. Melko, Neural decoder for topological codes, Phys. Rev. Lett. 119, 030501 (2017)

    2017

  72. [73]

    Bausch, A

    J. Bausch, A. W. Senior, F. J. Heras, T. Edlich, A. Davies, M. Newman, C. Jones, K. Satzinger, M. Y. Niu, S. Blackwell, et al., Learning high-accuracy error de- coding for quantum processors, Nature 635, 834 (2024)

    2024

  73. [74]

    B. M. Varbanov, M. Serra-Peralta, D. Byfield, and B. M. Terhal, Neural network decoder for near-term surface- code experiments, Phys. Rev. Res. 7, 013029 (2025)

    2025

  74. [75]

    G. Hu, W. Ouyang, C.-Y. Lu, C. Lin, and H.- S. Zhong, Efficient and universal neural-network de- coder for stabilizer-based quantum error correction, arXiv:2502.19971 (2025)

    Pith/arXiv arXiv 2025

  75. [76]

    Data-Driven Learnability Transition of Measurement-Induced Entanglement

    Y. Zhou, C. Wan, Y. Xu, J. P. Zhou, K. Q. Weinberger, and E.-A. Kim, Learning to decode logical circuits, Nat. Comput. Sci. (2025) . Supplemental Material for “Data-Driven Learnability Transition of Measurement-Induced Entanglement” Dongheng Qian 1, 2 and Jing Wang 1, 2, 3, 4, ∗ 1State Key Laboratory of Surface Physics and Department of Phy sics, Fudan Un...

    2025

  76. [77]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, Bert: Pre-t raining of deep bidirectional transformers for language understanding, in Proceedings of the 2019 conference of the North American cha pter of the association for computational linguistics: human language technologies, volume 1 (long a nd short papers) (2019) pp. 4171–4186

    2019

  77. [78]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N . Gomez, /suppress L. Kaiser, and I. Polosukhin, Attention is all you need, in Proceedings of the 31st International Conference on Neural Information Processing Systems , Vol. 30 (2017) pp. 6000–6010

    2017

  78. [79]

    Radford, J

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, I. Sutskever, et al., Language models are unsupervised multitask learners, OpenAI blog 1, 9 (2019)

    2019

  79. [80]

    W. Hou, S. J. Garratt, N. M. Eassa, E. Rosenberg, P. Roushan, Y .-Z. You, and E. Altman, Machine learning the effects of many quantum measurements, arXiv:2509.08890 (2025)

    Pith/arXiv arXiv 2025