The Quantum Ensemble Variational Optimization Algorithm: Applications to Molecular Inverse Design

Delmar G. A. Cabral; Francesco Calcagno; Ivan Rivalta; Victor S. Batista

REVIEW 3 major objections 4 minor 60 references

The Quantum Ensemble Variational Optimization Algorithm: Applications to Molecular Inverse Design

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read QEVO encodes molecular inverse design as a search over Pauli strings and samples candidates with a target property from a variational superposition, with simulations showing drug-like anticancer molecules on shallow, few-qubit circuits.

desk verdict Packet mismatch: the supplied full text is arXiv:2508.15895, not the QEVO paper; based on the abstract alone, the method is plausible but its central claims cannot be checked. read the letter →

arxiv 2508.15896 v1 pith:DV7WA3ZB submitted 2025-08-21 quant-ph physics.chem-ph

Francesco Calcagno , Delmar G. A. Cabral , Ivan Rivalta , Victor S. Batista This is my paper

classification quant-phphysics.chem-ph

keywords quantumvariationaloptimizationinversemoleculardesignPaulistringsdrugdiscoverynear-termcomputingansatzcombinatorial

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces QEVO, a variational quantum algorithm for inverse molecular design. It encodes molecular structures into an orthonormal basis of Pauli strings, builds a variational ansatz over that basis, and iteratively optimizes the ansatz against an objective that encodes the desired property. Sampling from the optimized superposition then yields molecular candidates. Numerical simulations reported in the abstract show QEVO designing drug-like molecules with anticancer properties using a shallow circuit and a modest number of qubits. If correct, this would give near-term quantum hardware a practical route around the combinatorial explosion that limits classical molecular design.

What carries the argument

The central object is the orthonormal basis of Pauli strings: products of single-qubit Pauli operators that form a complete, orthogonal basis for operators on the qubit space. Molecular structures are represented in this basis, and a variational ansatz is optimized over the basis coefficients so that sampling from the resulting superposition concentrates probability on structures with the target property. The optimization loop is the mechanism that turns a generic quantum state into a property-directed search over molecular candidates.

What would settle it

Run QEVO on a small molecular library where exhaustive enumeration is possible, then take the sampled candidates and test them in an independent assay (or against an external database of known actives and inactives). If the returned molecules are inactive, invalid, or unsynthesizable—or if the superposition simply maximizes the training objective without achieving the target property—the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that molecular inverse design can be recast as a quantum sampling problem: rather than enumerating structures, QEVO maps molecular structures onto an orthonormal basis of Pauli strings, prepares a superposition over that basis with a shallow variational ansatz, and optimizes the ansatz so that the superposition emphasizes molecules with a desired property. The abstract reports numerical simulations in which this procedure yields drug-like molecules with anticancer properties, and it argues that the resource cost—shallow circuits and modest qubit counts—makes the approach compatible with near-term and early fault-tolerant quantum platforms. The method is presented as a ge

Load-bearing premise

The objective used to score molecules must faithfully measure the real target property—anticancer activity—and the Pauli-string encoding must be able to represent chemically valid, drug-like structures; the abstract does not state how either is guaranteed.

Editorial extensions

If this is right

Molecular discovery becomes a sampling problem from a superposed ensemble rather than a sequential enumeration over a combinatorial space.
QEVO's resource requirements, as described, place the approach within reach of near-term quantum hardware and early fault-tolerant devices.
The same Pauli-string encoding and variational optimization can be redirected toward other molecular objectives such as solubility, toxicity, or binding affinity.
For small search spaces, QEVO's sampled candidates can be directly checked against exhaustive classical enumeration to verify that the optimization found the intended structures.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The abstract does not specify how "anticancer properties" are scored; if the scoring is a simulated surrogate, the real test is whether the sampled molecules show activity in an independent biological assay.
A natural testable extension is to run QEVO on a well-studied molecular library and compare its candidates against known actives and inactives, which would separate genuine generalization from overfitting to the training objective.
The Pauli-string representation may favor bit-string-like molecular encodings, and whether sampled structures are chemically valid and synthesizable is a question the abstract does not address; experimental synthesis would settle it.
If the approach generalizes as claimed, the same variational-sampling strategy could be applied to other high-dimensional inverse problems where the target is a property that can be scored, such as materials or catalyst design.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague—quick take: the packet for 2508.15896 is mismatched. The full text provided is Kim et al., 2508.15895, on measurement-induced phase transitions, not the QEVO paper. So I can't assess the actual manuscript. What I have is the abstract.

Based on the abstract alone: QEVO is a variational quantum algorithm that maps molecular structures onto a Pauli-string basis, optimizes a shallow ansatz against a property objective, and samples candidate molecules. The idea is a recognizable member of the variational-quantum-optimization family, and applying it to inverse drug design is a useful target. If the numerical simulations are there, this could be a reasonable step toward near-term quantum generative chemistry. The abstract is clear about the hardware context, which is a plus.

The soft spots are visible even in the abstract. No dataset, no baselines, no error bars, no qubit counts or circuit depths. More importantly, the objective: the ansatz is iteratively optimized to identify molecules with the desired property, and the reported success is molecules with anticancer properties. Unless the property is scored by an external assay or an independent surrogate, the method may be optimizing its own cost function—success by construction. The abstract doesn't say which. Also, the mapping of molecules to Pauli strings and the density of chemically valid, drug-like structures in that space is a load-bearing assumption that isn't even sketched.

None of this is a judgment against the authors; the body isn't here. The right move is to request the correct manuscript and then send it to referees if it matches the abstract. As it stands, I can't recommend peer review for a paper I cannot read.

Referee Report

3 major / 4 minor

Summary. The paper under review (arXiv:2508.15896) claims to introduce the Quantum Ensemble Variational Optimization (QEVO) algorithm for molecular inverse design. According to the abstract, QEVO maps molecular structures onto an orthonormal basis of Pauli strings, prepares a superposition state with a variational ansatz, and iteratively optimizes that ansatz so that sampling from the superposition yields molecular candidates with a desired property. The abstract reports numerical simulations showing that QEVO can design drug-like molecules with anticancer properties using a shallow quantum circuit and a modest number of qubits. However, the full text supplied with this review is arXiv:2508.15895, a different manuscript on measurement-induced phase transitions using Quantum Attention Networks; it contains no description of QEVO, no molecular representation, no objective function, and no numerical results relevant to molecular design. The report is therefore based on the abstract alone, with the full text serving only as evidence of a mismatch.

Significance. If the QEVO claim were fully supported, the work could be a useful contribution to quantum-assisted molecular inverse design, particularly if the circuit-depth and qubit requirements are low enough for near-term hardware. The abstract also promises a method that generalizes beyond molecular design to combinatorial problems, which would broaden the significance. However, none of these claims can currently be evaluated. There is no reproducible code, no machine-checked proof, no parameter-free derivation, and no falsifiable numerical benchmark in the supplied material. The strength of the contribution is therefore entirely contingent on missing evidence: the Pauli-string mapping and its inverse, the property-scoring function, the simulation protocol, the data set, the baselines, and the hardware-resource estimates.

major comments (3)

[Full text (arXiv:2508.15895)] The supplied full text is not the manuscript under review. It is a paper entitled 'Learning measurement-induced phase transitions using attention' and contains no mention of QEVO, molecular inverse design, Pauli-string encodings, or anticancer properties. Every load-bearing component of the abstract is therefore unverifiable: the encoding of molecular structures into Pauli strings, the variational ansatz, the optimization objective, the simulation results, and the resource counts. This is not a local presentation issue; it removes the basis for assessing the central claim.
[Abstract] The abstract states that the ansatz is 'iteratively optimized to identify molecular candidates with the desired property,' and then reports anticancer molecules as the outcome. If 'desired property' is encoded as the same scalar cost function used during optimization, the result is circular: any converged optimizer will return candidates that maximize the training objective. The paper must specify whether anticancer activity is measured by an independent computational surrogate, docking score, QSAR model, external assay, or synthesizability filter, and must validate that surrogate against known actives and inactives. Without that, the reported 'anticancer properties' are not established.
[Abstract] The claim 'numerical simulations demonstrate the potential of QEVO' is not supported by any quantitative detail in the abstract or in the supplied full text. There is no qubit count, no circuit depth, no number of ansatz parameters, no molecular data set, no baseline comparison (classical genetic algorithms, reinforcement learning, or existing quantum variational methods), no error bars, and no external validation of the designed molecules. The signature claim of near-term feasibility ('shallow quantum circuit,' 'modest number of qubits') is unquantified and untestable as presented.

minor comments (4)

[Abstract] The term 'curse of dimensionality' is used without citation or formal statement; a precise definition of the combinatorial search space would help.
[Abstract] The abstract distinguishes 'near-term and early fault-tolerant quantum computing platforms' but gives no indication of the assumed noise model, error rate, or fault-tolerance overhead. This distinction should be operationalized in the methods.
[Abstract] The phrase 'drug-like molecules with anticancer properties' should be accompanied by the specific molecular representation (e.g., SMILES, molecular graph, or fragment-based) and by one or more concrete examples from the simulations.
[Full text] The supplied full text is a different arXiv paper with its own code repository. If the QEVO manuscript exists separately, the correct full text should be provided; as submitted, the metadata are inconsistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity can be established from the available QEVO text; the supplied full text is an unrelated manuscript, so the derivation chain is not inspectable.

full rationale

The only QEVO text provided is the abstract. It states that the variational ansatz is 'iteratively optimized to identify molecular candidates with the desired property.' Optimizing an objective and then reporting candidates that score well on that objective is the standard operation of a variational optimizer, not a circular derivation: the paper does not claim to derive the objective from the outputs or to predict the training signal. No equations, fitted parameters, or self-citations are available in the abstract that would reduce a 'prediction' to an input. The supplied full text (arXiv:2508.15895, 'Learning measurement-induced phase transitions using attention') is a different paper and contains no QEVO content; this is a serious evidence gap for assessing the QEVO claims, but it is not itself a circularity. Under the hard rule that circularity must be demonstrated by quoting a specific reduction, no such reduction can be exhibited here. Accordingly the circularity score is 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The QEVO method rests on premises that the abstract asserts rather than establishes: the Pauli-string encoding is chemically faithful, the variational landscape is navigable at the claimed resource scale, and the property score used during optimization tracks the real target property. Because the body of the paper is unavailable (the supplied full text is a different arXiv paper), none of these could be verified. The variational ansatz parameters are fitted by training, making them free parameters by definition; the abstract reports no values. No new physical entities such as particles, forces, or dimensions are postulated.

free parameters (2)

Variational ansatz parameters = not reported in abstract
The ansatz is 'iteratively optimized' against the desired-property objective; its parameters are fitted during training, and their values and count are not given in the abstract.
QEVO hyperparameters (qubit count, circuit depth, ensemble size) = not reported in abstract
The abstract claims a 'modest number of qubits' and a 'shallow' circuit but reports no actual values; these are hand-picked choices for the demonstration, not quantities derived from first principles.

assumptions (3)

domain assumption Molecular structures can be faithfully represented as strings in an orthonormal Pauli-string basis
The entire encoding scheme presupposes that the chosen Pauli-string basis preserves chemically meaningful structural variation. Stated in the abstract ('maps molecular structures onto an orthonormal basis of Pauli strings') without proof or demonstration.
domain assumption A shallow variational ansatz with a modest qubit count can be optimized to concentrate the superposition on desirable candidates
The method's tractability presumes the variational landscape is navigable and that the search space defined by the encoding contains valid drug-like molecules. Not substantiated in the abstract; no trainability or landscape analysis is mentioned.
domain assumption The property used to score candidates during optimization is a faithful measure of the target property (anticancer activity)
The abstract's outcome, molecules with anticancer properties, is the same quantity the ansatz is optimized toward; the fidelity of the scoring function is unstated. This assumption is where the circularity risk concentrates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The Quantum Ensemble Variational Optimization Algorithm: Applications to Molecular Inverse Design." pith.science (2026). https://pith.science/paper/DV7WA3ZB

@misc{pith2026250815896,
  author       = {Pith},
  title        = {Pith review of: The Quantum Ensemble Variational Optimization Algorithm: Applications to Molecular Inverse Design},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DV7WA3ZB}},
  note         = {Machine review of arXiv:2508.15896}
}

read the original abstract

Designing molecules with optimized properties remains a fundamental challenge due to the intricate relationship between molecular structure and properties. Traditional computational approaches that address the combinatorial number of possible molecular designs become unfeasible as the molecular size increases, suffering from the so-called `curse of dimensionality' problem. Recent advances in quantum computing hardware present new opportunities to address this problem. Here, we introduce the Quantum Ensemble Variational Optimization (QEVO) method for near-term and early fault-tolerant quantum computing platforms. QEVO efficiently maps molecular structures onto an orthonormal basis of Pauli strings and samples from a superposition state generated by a variational ansatz. The ansatz is iteratively optimized to identify molecular candidates with the desired property. Our numerical simulations demonstrate the potential of QEVO in designing drug-like molecules with anticancer properties, employing a shallow quantum circuit that requires only a modest number of qubits. We envision that QEVO could be applied to a wide range of complex problems, offering practical solutions to problems with combinatorial complexity.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

60 extracted references · 56 canonical work pages

[1]

Maudlin, Topoi 14, 7{15 (1995)

T. Maudlin, Topoi 14, 7{15 (1995)

work page 1995
[2]

Y. Li, X. Chen, and M. P. A. Fisher, Physical Review B 98, 205136 (2018)

2018
[3]

Y. Li, X. Chen, and M. P. A. Fisher, Physical Review B 100, 134306 (2019)

work page 2019
[4]

Skinner, J

B. Skinner, J. Ruhman, and A. Nahum, Physical Review X 9, 031009 (2019)

2019
[5]

A. Chan, R. M. Nandkishore, M. Pretko, and G. Smith, Physical Review B 99, 224307 (2019)

work page 2019
[6]

M. J. Gullans and D. A. Huse, Physical Review X 10, 041020 (2020)

work page 2020
[7]

Zabalo, M

A. Zabalo, M. J. Gullans, J. H. Wilson, S. Gopalakrish- nan, D. A. Huse, and J. H. Pixley, Physical Review B 101, 060301 (2020)

work page 2020
[8]

Y. Bao, S. Choi, and E. Altman, Physical Review B 101, 104301 (2020)

work page 2020

Show all 60 references

[9]

Jian, Y.-Z

C.-M. Jian, Y.-Z. You, R. Vasseur, and A. W. W. Ludwig, Physical Review B 101, 104302 (2020)

2020
[10]

R. Fan, S. Vijay, A. Vishwanath, and Y.-Z. You, Phys. Rev. B 103, 174309 (2021)

2021
[11]

A. C. Potter and R. Vasseur, in Entanglement in Spin Chains: From Theory to Quantum Technology Applica- tions (Springer, 2022) pp. 211{249

2022
[12]

M. P. Fisher, V. Khemani, A. Nahum, and S. Vijay, An- nual Review of Condensed Matter Physics14, 335 (2023)

2023
[13]

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, Nature Physics 18, 760 (2022)

2022
[14]

Google Quantum AI and Collaborators, Nature 622, 481 (2023)

2023
[15]

J. M. Koh, S.-N. Sun, M. Motta, and A. J. Minnich, Nature Physics 19, 1314 (2023)

2023
[16]

Y. Li, Y. Zou, P. Glorioso, E. Altman, and M. P. A. Fisher, Physical Review Letters 130, 220404 (2023)

2023
[17]

S. J. Garratt and E. Altman, PRX Quantum 5, 030311 (2024)

2024
[18]

Yanay, B

Y. Yanay, B. Swingle, and C. Tahan, Phys. Rev. Lett. 133, 070601 (2024)

2024
[19]

Kamakari, J

H. Kamakari, J. Sun, Y. Li, J. J. Thio, T. P. Gujarati, M. P. A. Fisher, M. Motta, and A. J. Minnich, Phys. Rev. Lett. 134, 120401 (2025)

2025
[20]

X. Feng, J. C^ ot e, S. Kourtis, and B. Skinner, arXiv preprint arXiv:2502.01735 (2025)

2025 arXiv
[21]

Barratt, U

F. Barratt, U. Agrawal, A. C. Potter, S. Gopalakrishnan, and R. Vasseur, Physical Review Letters 129, 200602 (2022)

2022
[22]

Dehghani, A

H. Dehghani, A. Lavasani, M. Hafezi, and M. J. Gullans, Nature Communications 14, 2918 (2023)

2023
[23]

Ippoliti and V

M. Ippoliti and V. Khemani, PRX Quantum 5, 020304 (2024)

2024
[24]

A. A. Akhtar, H.-Y. Hu, and Y.-Z. You, Phys. Rev. B 109, 094209 (2024)

2024
[25]

Agrawal, J

U. Agrawal, J. Lopez-Piqueres, R. Vasseur, S. Gopalakr- ishnan, and A. C. Potter, Physical Review X 14, 041012 (2024)

2024
[26]

McGinley, PRX Quantum 5, 020347 (2024)

M. McGinley, PRX Quantum 5, 020347 (2024)

2024
[27]

Y. Hu, Y. H. Teoh, W. Witczak-Krempa, and R. G. Melko, Neural network enhanced cross entropy bench- mark for monitored circuits (2025), arXiv:2501.13005 [quant-ph]

2025
[28]

S. Choi, Y. Bao, X.-L. Qi, and E. Altman, Physical Re- view Letters 125, 030505 (2020)

2020
[29]

M. J. Gullans and D. A. Huse, Physical Review Letters 125, 070606 (2020)

2020
[30]

H. Kim, Y. Zhou, Y. Xu, K. Varma, A. H. Karamlou, I. T. Rosen, J. C. Hoke, C. Wan, J. P. Zhou, W. D. Oliver, Y. D. Lensky, K. Q. Weinberger, and E.-A. Kim, At- tention to quantum complexity (2024), arXiv:2405.11632 [quant-ph]

2024
[31]

Szyniszewski, A

M. Szyniszewski, A. Romito, and H. Schomerus, Physical Review B 100, 064204 (2019)

2019
[32]

K. Aziz, A. Chakraborty, and J. Pixley, Physical Review B 110, 064301 (2024)

2024
[33]

[25] referred to this quantity as credence

Ref. [25] referred to this quantity as credence
[34]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, Advances in neural information processing systems 30 (2017)

2017
[35]

J. Lee, Y. Lee, J. Kim, A. Kosiorek, S. Choi, and Y. W. Teh, in International conference on machine learning (PMLR, 2019) pp. 3744{3753

2019
[36]

Learning measurement-induced phase transitio ns using attention

J. M. Pino, J. M. Dreiling, C. Figgatt, J. P. Gaebler, S. A. Moses, M. Allman, C. Baldwin, M. Foss-Feig, D. Hayes, K. Mayer, et al. , Nature 592, 209 (2021). Supplementary materials for “Learning measurement-induced phase transitio ns using attention” Hyejin Kim, 1 Abhishek Ku...

2021
[37]

Prob(p) at γ = 0 with ancilla prepared in σy = 1 eigenstate 2
[38]

Optimal Pcorr in the thermodynamic limit 3 C

Comparison between the tails of Prob( p;M ) at γ = 0 and γ = 1 2 B. Optimal Pcorr in the thermodynamic limit 3 C. Data acquisition 5
[39]

General Setup for Monitored Circuit Dynamics 5
[40]

State Distinguishability Setup 5
[41]

Phase Recognition Setup 6
[42]

Baseline for phase recognition 7
[43]

Estimation of Prob( p) from the measurement record 8

Spatiotemporal correlations in the measurement record as ﬁnite-size artifacts 8 D. Estimation of Prob( p) from the measurement record 8
[44]

Model Architecture and Training scheme 10

Time-independence of Born probabilitiy p(⃗ m)t in weak monitoring phase 9 E. Model Architecture and Training scheme 10
[45]

Embedding 10 b

QuAN architecture 10 a. Embedding 10 b. Encoder 10 c. Pooling Attention Block (PAB) Decoder 11
[46]

Extended machine learning results 13

Data preprocessing, training, and testing scheme 12 F. Extended machine learning results 13
[47]

Optimal hyperparameters study for state distinguishing task 13
[48]

Sample complexity study for state distinguishing task 14
[49]

Minimal sample complexity and the system size scaling for phase rec ognition task 14
[50]

Optimal set size study for phase recognition task 16
[51]

System size scaling of the transition point for phase recognition task 16 G. QuAN accessing the Prob( p) distribution through inter-trajectory attention 17 References 19 Appendix A: Theoretical Prob(p) for ﬁnite sample size M In this section, we discuss the distribution of Bor...
[52]

We consider the setting with ancilla initial state σy = 1 eigenstate, as in the phase recognition task

Prob(p) at γ = 0 with ancilla prepared in σy = 1 eigenstate First, we focus on the behavior of the distribution in the weak monit oring limit of γ = 0. We consider the setting with ancilla initial state σy = 1 eigenstate, as in the phase recognition task. See SFig. 2(b) for th...
[53]

Prob(p) at γ = 1 In the strong monitoring limit of γ = 1, the protocol projectively measures system qubits. For a random s tate of system size L generated by scrambling dynamics, the Born probabilities follow the Beta distribution [2], given by Prob(p) = (D − 1)(1 −p)D2 → 2LeD...
[54]

optimal decoding

Comparison between the tails of Prob(p;M ) at γ = 0 and γ = 1 We aim to compare the tails of two distributions – between γ = 0 for ancilla prepared in σy = 1 eigenstate and γ = 1 – at a ﬁnite sample size M ≪D = 2L. For each case, Born probability is given by Eq. (A2) (binomial...
[55]

For γ = 1 where Prob(p) is given by Eq

(B5) A discretized quantity that deﬁnes success as p(ψ |m) > 1/2 is an alternative metric referred to as the accuracy α [3, 4] α ≡ E[p(ψ |m)> 1/2], (B6) where ψ is again the correct initial state and E[·] denotes the average over all measurement outcomes. For γ = 1 where Prob(...
[56]

The initial state |ψ⟩ is an entangled and scrambled quantum state generated via applying unit ary circuits to a chosen product state

General Setup for Monitored Circuit Dynamics To investigate unitary-monitored dynamics, we consider a (1+1)d circ uit generating hybrid dynamics acting on some initial state |ψ⟩. The initial state |ψ⟩ is an entangled and scrambled quantum state generated via applying unit ary ...
[57]

# ! !" !

State Distinguishability Setup In this subsection, we provide details about parameters and setups s peciﬁc to the state-distinguishing task (see SFig. 2(a)). First, we specify parameters used for initial state preparation. For s tate distinguishability, we consider two initial...
[58]

# ./%0 # $ % &' ' ( !

Phase Recognition Setup The phase-recognition setup hinges on learning the diﬀerence in cru cial characteristics between the mixed and pure phases of MIPT. We made adequate changes in our protocol, which takes into account the distinct nature of the 7 classiﬁcation task compar...
[59]

Baseline for phase recognition To benchmark QuAN’s prediction of the transition point, we estimate th e critical measurement strength through a diﬀerent protocol. We consider a slightly diﬀerent circuit with an extra reference qubit (Q) maximally entangled to one of the system...
[60]

Spatiotemporal correlations in the measurement record a s ﬁnite-size artifacts SFig. 4. Spatio-temporal correlation C(δt =L,δx = 0) as a function of measurement strength γ for varying system size L. The error bar represents the standard error from averaging over diﬀerent spa c...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Maudlin, Topoi 14, 7{15 (1995)

T. Maudlin, Topoi 14, 7{15 (1995)

work page 1995

[2] [2]

Y. Li, X. Chen, and M. P. A. Fisher, Physical Review B 98, 205136 (2018)

2018

[3] [3]

Y. Li, X. Chen, and M. P. A. Fisher, Physical Review B 100, 134306 (2019)

work page 2019

[4] [4]

Skinner, J

B. Skinner, J. Ruhman, and A. Nahum, Physical Review X 9, 031009 (2019)

2019

[5] [5]

A. Chan, R. M. Nandkishore, M. Pretko, and G. Smith, Physical Review B 99, 224307 (2019)

work page 2019

[6] [6]

M. J. Gullans and D. A. Huse, Physical Review X 10, 041020 (2020)

work page 2020

[7] [7]

Zabalo, M

A. Zabalo, M. J. Gullans, J. H. Wilson, S. Gopalakrish- nan, D. A. Huse, and J. H. Pixley, Physical Review B 101, 060301 (2020)

work page 2020

[8] [8]

Y. Bao, S. Choi, and E. Altman, Physical Review B 101, 104301 (2020)

work page 2020