Subsystem relaxation and a calibrated sampling diagnostic for programmable quantum annealers

Luis Lozano

arxiv: 2605.19381 · v1 · pith:EJVOYFZGnew · submitted 2026-05-19 · 🪐 quant-ph · cond-mat.dis-nn

Subsystem relaxation and a calibrated sampling diagnostic for programmable quantum annealers

Luis Lozano This is my paper

Pith reviewed 2026-05-20 06:19 UTC · model grok-4.3

classification 🪐 quant-ph cond-mat.dis-nn

keywords quantum annealingsubsystem relaxationopen quantum systemssampling diagnosticmemory order parameterD-Wave annealerconditional Boltzmann referenceinitial state independence

0 comments

The pith

Six-qubit subsystems on quantum annealers lose memory of their initial state when the environment is large or strongly coupled.

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

The paper examines how a small subsystem embedded in a programmable quantum annealer relaxes away from its preparation when the surrounding environment changes in size, coupling strength, and disorder. It shows that the subsystem reaches initial-state independence under large or strong environments but stays trapped by quenched disorder or atypical environment configurations. Pairing a memory order parameter with distance to a separately calibrated conditional-Boltzmann reference produces a diagnostic that identifies rare wrong-basin trapping even when memory appears lost. This matters for using annealers as open-system samplers because it supplies a concrete check on whether readout statistics reflect proper thermal relaxation rather than preparation artifacts or model mismatch.

Core claim

A six-qubit subsystem becomes initial-state independent when the environment is large or strongly coupled, while quenched disorder and atypical environment states arrest relaxation. Pairing the memory order parameter with the distance to a calibrated conditional-Boltzmann reference yields a diagnostic that flags rare wrong-basin trapping that memory loss alone misses; memory-retaining conditions stay far from the reference with median distance 0.35. Relaxed ferromagnetic readouts are near-deterministic so small distances there serve as consistency checks. In a mixed-frustration benchmark the local-update model mispredicts QPU relaxation roughly sevenfold while non-local classical sampling is

What carries the argument

Subsystem-environment protocol that varies environment size, coupling, disorder, preparation, geometry and QPU generation, combined with a memory order parameter and distance to an independently calibrated conditional-Boltzmann reference distribution.

If this is right

Relaxed ferromagnetic readouts become near-deterministic and serve as a consistency check rather than a thermometric test.
The local-update model used by practitioners mispredicts observed relaxation dynamics by a factor of roughly seven.
Non-local classical sampling reproduces the QPU relaxation behavior in the mixed-frustration benchmark.
The protocol supplies a subsystem-level validation method for assessing sampling quality on quantum annealers.

Where Pith is reading between the lines

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

The diagnostic could be applied to other open-system quantum devices to separate genuine thermal sampling from preparation memory.
Extending the protocol to larger subsystems might reveal scaling limits on when environment size guarantees relaxation.
The mismatch with local-update models suggests that sampling algorithms for annealers should incorporate non-local moves to match hardware behavior.

Load-bearing premise

The conditional-Boltzmann reference distribution can be calibrated independently of the QPU data in a way that supplies an unbiased benchmark without circular dependence on the same measurements.

What would settle it

Direct measurement showing that the six-qubit subsystem remains dependent on its initial state even in large or strongly coupled environments, or that the diagnostic distance stays large for clearly relaxed ferromagnetic cases.

Figures

Figures reproduced from arXiv: 2605.19381 by Luis Lozano.

**Figure 1.** Figure 1: Reverse-anneal protocol and subsystem–environment partition. a, Energy-landscape concept: three distinct initial preparations of the subsystem (colored balls in different potential wells) converge to the same distribution under relaxing conditions after reverse annealing in the presence of a large on-chip environment. b, Reverse-anneal schedule. The system is initialized at s = 1 (classical), ramped to a p… view at source ↗

**Figure 2.** Figure 2: Subsystem relaxation with tunable environment. a, Memory order parameter M versus environment size |E| on both QPUs (sp = 0.4, λ = 0.5). b, Memory versus coupling strength λ (|E| = 50). Filled markers with solid lines: legacy SDK-default submission path. Open markers with dashed lines: regenerated on direct native-qubit submissions with auto scale=False (no chain couplers). The λ = 0 point is a decoupled c… view at source ↗

**Figure 3.** Figure 3: Disorder arrests relaxation. a, Memory order parameter versus disorder strength W on both QPUs (|E| = 50, λ = 0.5, sp = 0.4). Solid curves: legacy SDK-default submission path. Open triangles (dotted, red): a single-seed reproduction on Advantage system6.4 at the uniform torque compensation chain-strength path with auto scale=True, giving M = 0.54 at W = 1.5 versus the legacy 0.50. Memory rises sharply abov… view at source ↗

**Figure 4.** Figure 4: Effective thermal marginal. a, Fraction of relaxed disorder realizations (M ≤ 0.05) in the 10-seed disorder sweep (N = 12, |S| = 4, λ = 0.5, 3-regular logical graph, Advantage2). Relaxation fraction decreases from 10/10 at W = 0 to 4/10 at W = 1.25. b, Total-variation distance between the pooled subsystem marginal and the calibrated classical conditional Gibbs target at βeff = 7.219 vs disorder strength. O… view at source ↗

**Figure 5.** Figure 5: Classical and small-system controls. a, Comparison across system sizes at fixed sp = 0.4 and representative (λ, W) (see Supplementary Sec. 3 for the full Glauber and SVMC sweeps over system size, coupling strength and disorder). Glauber dynamics at the device temperature (gray diamonds) shows M = 1.0 everywhere. ED at sp = 0.4 (green triangles) shows decreasing memory with system size. The blue-shaded band… view at source ↗

**Figure 6.** Figure 6: Non-local classical sampling reproduces the mixed-frustration relaxation gap. a, Relaxation fraction versus classical temperature T on mixed-frustration instances (pS = 0.5, W = 1.0, N = 12, 20 seeds), Advantage2. Local single-spin-flip Glauber (gray) reaches the QPU rate (blue dashed, 70%) only at T ≈ 4.8 × Tdevice—a roughly sevenfold misprediction by the local-update model. At the device temperature, pa… view at source ↗

read the original abstract

Programmable quantum annealers are used as open-system samplers, but it is unclear when reverse annealing erases preparation memory and what the readout represents. Here we implement a subsystem-environment protocol on two D-Wave quantum annealers, varying environment size, coupling, disorder, preparation, geometry and QPU generation. A six-qubit subsystem becomes initial-state independent when the environment is large or strongly coupled, while quenched disorder and atypical environment states arrest relaxation. Pairing the memory order parameter with the distance to a calibrated conditional-Boltzmann reference yields a diagnostic that flags rare wrong-basin trapping memory loss alone misses; memory-retaining conditions stay far from the reference (median 0.35). Relaxed ferromagnetic readouts are near-deterministic, so small distances there are a consistency check, not a thermometric test. In a mixed-frustration benchmark, the local-update model practitioners assume mispredicts QPU relaxation roughly sevenfold, whereas non-local classical sampling recovers it. We provide a subsystem-level validation protocol for quantum-annealer sampling.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a hardware protocol to test subsystem relaxation on D-Wave annealers and a diagnostic pairing memory order with distance to a calibrated reference, but the calibration step looks vulnerable to circularity.

read the letter

The core contribution is a subsystem-environment setup on actual D-Wave hardware that shows a six-qubit block loses dependence on its initial state once the surrounding qubits are large enough or coupled strongly. Quenched disorder and odd environment states block that relaxation. They add a memory order parameter and measure distance to a calibrated conditional-Boltzmann distribution to spot cases where the system is trapped in the wrong basin even after apparent memory loss. In one benchmark the local-update model that most users assume gets the relaxation wrong by roughly a factor of seven, while non-local classical sampling matches better. Relaxed ferromagnetic cases sit close to the reference, which serves as a consistency check rather than a temperature measurement. That combination of controlled environment variation and the paired diagnostic is new enough to be worth attention for people who actually run sampling workloads on these machines. The observations are concrete and tied to real device runs, which is the main strength. The soft spot is the calibration of the conditional-Boltzmann reference. If those parameters are tuned on the same or closely related QPU readouts, the distance metric stops being an independent benchmark and the claim that it flags trapping becomes partly self-referential. The abstract gives no error bars, no raw counts, and no clear description of how the calibration avoids using the test data itself, so the median distance of 0.35 and the sevenfold mismatch are hard to weigh without more controls. This is aimed at experimentalists and practitioners who use quantum annealers for optimization or generative modeling and need a way to check whether the hardware output has actually thermalized or relaxed. A reader working on validation methods or open-system sampling will get direct value from the protocol even if the numbers need tightening. The work is coherent on its own terms and engages the practical question, so it deserves a serious referee. I would send it to review with the main request being a clear, independent description of how the reference distribution is built and whether it can be reproduced without the QPU samples under test.

Referee Report

3 major / 3 minor

Summary. The manuscript implements a subsystem-environment protocol on two D-Wave quantum annealers, varying environment size, coupling strength, disorder, preparation, geometry, and QPU generation. It reports that a six-qubit subsystem becomes independent of its initial state when the environment is large or strongly coupled, while quenched disorder and atypical environment states arrest relaxation. The central contribution is a diagnostic that pairs a memory order parameter with the distance to a calibrated conditional-Boltzmann reference; this diagnostic is claimed to flag rare wrong-basin trapping events that memory loss alone misses, with memory-retaining conditions staying far from the reference (median distance 0.35). In a mixed-frustration benchmark the local-update model is reported to mispredict QPU relaxation roughly sevenfold, whereas non-local classical sampling recovers the observed behavior. The work concludes by offering a subsystem-level validation protocol for quantum-annealer sampling.

Significance. If the central claims hold, the paper supplies a practical, experimentally grounded diagnostic for assessing when programmable quantum annealers function as unbiased samplers versus when they remain trapped in wrong basins. The quantitative mismatch between the local-update model and hardware data, together with the recovery by non-local sampling, directly challenges a modeling assumption widely used by practitioners. The multi-parameter experimental sweep on real QPUs and the concrete numerical observations (six-qubit independence, median distance 0.35, sevenfold discrepancy) constitute useful benchmarks for the community. The provision of an explicit validation protocol is a constructive contribution to the field of quantum annealing as open-system sampling.

major comments (3)

[Calibration procedure and diagnostic definition] Calibration of conditional-Boltzmann reference: the manuscript states that the reference is 'calibrated' but does not demonstrate that the calibration parameters are obtained from an independent classical procedure or from data disjoint from the QPU subsystem readouts used to compute the distance metric. Because the diagnostic's validity rests on the reference serving as an unbiased benchmark, any dependence on the same measurements renders the distance partially self-referential and weakens the claim that the diagnostic reliably distinguishes proper relaxation from trapping.
[Mixed-frustration benchmark results] Model-mismatch quantification: the abstract and results section report that the local-update model 'mispredicts QPU relaxation roughly sevenfold.' No error bars, raw counts, or full statistical controls are supplied for this factor, nor is the precise definition of 'misprediction' (e.g., which observable, which subset of conditions) given in sufficient detail to allow independent verification. This quantitative claim is load-bearing for the paper's critique of common modeling assumptions.
[Subsystem relaxation results] Statistical controls for relaxation claims: the central observation that 'a six-qubit subsystem becomes initial-state independent when the environment is large or strongly coupled' is presented without reported uncertainties, sample sizes per condition, or explicit tests for post-hoc selection of the 'large/strong' regimes. Because this independence underpins both the memory-order-parameter analysis and the diagnostic, the absence of these controls affects the robustness of the reported phenomenology.

minor comments (3)

[Methods] The definition of the memory order parameter and the precise distance metric to the conditional-Boltzmann reference should be stated as explicit equations in the methods section for reproducibility.
[Figures] Figure panels displaying distance distributions would benefit from inclusion of raw histograms or additional panels stratified by environment size and coupling to allow readers to assess the median value of 0.35 directly.
[Experimental methods] A short table summarizing the number of experimental runs, qubit counts, and annealing schedules for each QPU generation would improve clarity of the experimental design.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thorough review and valuable feedback on our manuscript. We have carefully considered each comment and made revisions to improve the clarity and rigor of the presentation. Our point-by-point responses are provided below.

read point-by-point responses

Referee: [Calibration procedure and diagnostic definition] Calibration of conditional-Boltzmann reference: the manuscript states that the reference is 'calibrated' but does not demonstrate that the calibration parameters are obtained from an independent classical procedure or from data disjoint from the QPU subsystem readouts used to compute the distance metric. Because the diagnostic's validity rests on the reference serving as an unbiased benchmark, any dependence on the same measurements renders the distance partially self-referential and weakens the claim that the diagnostic reliably distinguishes proper relaxation from trapping.

Authors: We appreciate this observation. The calibration of the conditional-Boltzmann reference was performed using an independent classical Monte Carlo sampling procedure on the effective subsystem Hamiltonian, drawing on data sources separate from the QPU subsystem readouts. To eliminate any ambiguity regarding self-referentiality, we have added a dedicated subsection in the Methods that explicitly describes the classical calibration protocol, confirms the disjoint nature of the data, and includes a supplementary figure illustrating the calibration workflow. This revision strengthens the claim that the diagnostic serves as an unbiased benchmark. revision: yes
Referee: [Mixed-frustration benchmark results] Model-mismatch quantification: the abstract and results section report that the local-update model 'mispredicts QPU relaxation roughly sevenfold.' No error bars, raw counts, or full statistical controls are supplied for this factor, nor is the precise definition of 'misprediction' (e.g., which observable, which subset of conditions) given in sufficient detail to allow independent verification. This quantitative claim is load-bearing for the paper's critique of common modeling assumptions.

Authors: We agree that additional statistical detail is warranted for this quantitative claim. In the revised manuscript we have added error bars (standard error of the mean) to the relevant plots, supplied the raw event counts underlying the sevenfold factor, and provided an explicit definition of misprediction as the ratio of the local-update model's predicted relaxation probability to the observed QPU relaxation probability for the initial-state independence observable, restricted to the mixed-frustration benchmark conditions. We have also included bootstrap-derived uncertainty estimates and clarified the exact subset of parameter points used. revision: yes
Referee: [Subsystem relaxation results] Statistical controls for relaxation claims: the central observation that 'a six-qubit subsystem becomes initial-state independent when the environment is large or strongly coupled' is presented without reported uncertainties, sample sizes per condition, or explicit tests for post-hoc selection of the 'large/strong' regimes. Because this independence underpins both the memory-order-parameter analysis and the diagnostic, the absence of these controls affects the robustness of the reported phenomenology.

Authors: We acknowledge the value of these controls. The revised results section now reports the number of anneals per condition (typically 1000–2000), includes standard-error bars on the memory order parameter, and states that the 'large' and 'strongly coupled' regimes were defined a priori from theoretical scaling arguments rather than post-hoc inspection. We have added a statistical comparison (two-sample t-test) confirming significant loss of initial-state dependence in those regimes, together with a brief discussion of how the regime boundaries were chosen. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The provided abstract and excerpts describe an empirical protocol on D-Wave QPUs, subsystem relaxation under varying environment size/coupling/disorder, and a diagnostic pairing a memory order parameter with distance to a calibrated conditional-Boltzmann reference. No equations, sections, or derivation steps are quoted that reduce a claimed prediction or result to its own inputs by construction (e.g., no fitted parameter renamed as independent prediction, no self-citation load-bearing a uniqueness claim, no ansatz smuggled via prior work). The calibration procedure is mentioned but not detailed in a manner allowing exhibition of circular dependence on the same QPU readouts. Per hard rules, without a specific quote exhibiting Eq. X = Eq. Y or equivalent reduction, circularity cannot be claimed. The central claims rest on experimental variation and comparison to classical sampling, appearing self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the existence of an independently calibratable conditional-Boltzmann reference and on the assumption that observed relaxation behavior generalizes across the tested parameter variations.

free parameters (1)

calibration parameters for conditional-Boltzmann reference
The reference is described as calibrated, implying one or more parameters adjusted to match expected distributions.

axioms (1)

domain assumption The conditional-Boltzmann distribution provides a valid external benchmark for the relaxed subsystem state on the QPU.
Invoked when the diagnostic pairs the memory order parameter with distance to this reference.

pith-pipeline@v0.9.0 · 5706 in / 1287 out tokens · 54876 ms · 2026-05-20T06:19:44.588145+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Pairing the memory order parameter with the distance to a calibrated conditional-Boltzmann reference yields a diagnostic that flags rare wrong-basin trapping
IndisputableMonolith/Foundation/AlphaCoordinateFixation.lean J_uniquely_calibrated_via_higher_derivative unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

β_eff = 1/2h ln(n↓/n↑) ... conditional marginal P_th_S(σ_S|σ_E=1)

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · 1 internal anchor

[1]

Application of quantum annealing to training of deep neural networks

Steven H. Adachi and Maxwell P. Henderson. “Application of quantum annealing to training of deep neural networks”. Preprint athttps://arxiv.org/abs/1510. 06356(2015)

work page 2015
[2]

Quantum Boltzmann machine

Mohammad H. Amin, Evgeny Andriyash, Jason Rolfe, Bohdan Kulchytskyy, and Roger Melko. “Quantum Boltzmann machine”. Physical Review X8, 021050 (2018)

work page 2018
[3]

Estimation of effective temperatures in quantum annealers for sampling ap- plications: A case study with possible applications in deep learning

Marcello Benedetti, John Realpe-G´ omez, Rupak Biswas, and Alejandro Perdomo- Ortiz. “Estimation of effective temperatures in quantum annealers for sampling ap- plications: A case study with possible applications in deep learning”. Physical Review A94, 022308 (2016)

work page 2016
[4]

High-quality thermal Gibbs sampling with quantum annealing hardware

Jon Nelson, Marc Vuffray, Andrey Y. Lokhov, Tameem Albash, and Carleton Coffrin. “High-quality thermal Gibbs sampling with quantum annealing hardware”. Physical Review Applied17, 044046 (2022)

work page 2022
[5]

Programmable quantum annealers as noisy Gibbs samplers

Marc Vuffray, Carleton Coffrin, Yaroslav A. Kharkov, and Andrey Y. Lokhov. “Programmable quantum annealers as noisy Gibbs samplers”. PRX Quantum3, 020317 (2022)

work page 2022
[6]

Quantum annealing and condensed matter physics

Viv Kendon and Nicholas Chancellor. “Quantum annealing and condensed matter physics”. Journal of Physics: Condensed Matter38, 143005 (2026)

work page 2026
[7]

Global warming: Tempera- ture estimation in annealers

Jack Raymond, Sheir Yarkoni, and Evgeny Andriyash. “Global warming: Tempera- ture estimation in annealers”. Frontiers in ICT3, 23 (2016). 16

work page 2016
[8]

Thermalization, freeze-out, and noise: Deciphering experimental quantum annealers

Jeffrey Marshall, Eleanor G. Rieffel, and Itay Hen. “Thermalization, freeze-out, and noise: Deciphering experimental quantum annealers”. Physical Review Applied8, 064025 (2017)

work page 2017
[9]

Power of paus- ing: Advancing understanding of thermalization in experimental quantum annealers

Jeffrey Marshall, Davide Venturelli, Itay Hen, and Eleanor G. Rieffel. “Power of paus- ing: Advancing understanding of thermalization in experimental quantum annealers”. Physical Review Applied11, 044083 (2019)

work page 2019
[10]

Classical thermometry of quantum an- nealers.arXiv preprint arXiv:2512.03162, 2025

George Grattan, Pratik Sathe, and Cristiano Nisoli. “Classical thermometry of quan- tum annealers” (2025). arXiv:2512.03162

work page arXiv 2025
[11]

Erasing classical memory with quantum fluctuations: Shannon information entropy of reverse quantum annealing.arXiv preprint arXiv:2509.10927,

Elijah Pelofske and Cristiano Nisoli. “Erasing classical memory with quantum fluc- tuations: Shannon information entropy of reverse quantum annealing”. Preprint at https://arxiv.org/abs/2509.10927(2025)

work page arXiv 2025
[12]

Finite-temperature criticality through quantum annealing.arXiv preprint arXiv:2507.07167, 2025

Gianluca Teza et al. “Finite-temperature criticality through quantum anneal- ing” (2025). arXiv:2507.07167

work page arXiv 2025
[13]

Localization transition induced by programmable disorder

Jaime L. C. da C. Filho, Zoe Gonzalez Izquierdo, Andreia Saguia, Tameem Albash, Itay Hen, and Marcelo S. Sarandy. “Localization transition induced by programmable disorder”. Physical Review B105, 134201 (2022)

work page 2022
[14]

Entanglement and the foundations of statistical mechanics

Sandu Popescu, Anthony J. Short, and Andreas Winter. “Entanglement and the foundations of statistical mechanics”. Nature Physics2, 754–758 (2006)

work page 2006
[15]

Canon- ical typicality

Sheldon Goldstein, Joel L. Lebowitz, Roderich Tumulka, and Nino Zangh` ı. “Canon- ical typicality”. Physical Review Letters96, 050403 (2006)

work page 2006
[16]

From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics

Luca D’Alessio, Yariv Kafri, Anatoli Polkovnikov, and Marcos Rigol. “From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics”. Advances in Physics65, 239–362 (2016)

work page 2016
[17]

Coherent quantum annealing in a programmable 2,000 qubit Ising chain

Andrew D. King, Sei Suzuki, Jack Raymond, Alex Zucca, Trevor Lanting, Fabio Altomare, Andrew J. Berkley, et al. “Coherent quantum annealing in a programmable 2,000 qubit Ising chain”. Nature Physics18, 1324–1328 (2022)

work page 2022
[18]

Quantum critical dynamics in a 5,000-qubit programmable spin glass

Andrew D. King, Jack Raymond, Trevor Lanting, Richard Harris, Alex Zucca, Fabio Altomare, Andrew J. Berkley, et al. “Quantum critical dynamics in a 5,000-qubit programmable spin glass”. Nature617, 61–66 (2023)

work page 2023
[19]

Unraveling reverse annealing: A study of D-Wave quantum annealers

Vrinda Mehta, Hans De Raedt, Kristel Michielsen, and Fengping Jin. “Unraveling reverse annealing: A study of D-Wave quantum annealers”. Physical Review A112, 012414 (2025)

work page 2025
[20]

Understanding the physics of D-Wave annealers: From Schr¨ odinger to Lindblad to Markovian dynamics

Vrinda Mehta et al. “Understanding the physics of D-Wave annealers: From Schr¨ odinger to Lindblad to Markovian dynamics”. Physical Review A112, 032616 (2025)

work page 2025
[21]

Adiabatic quantum computation

Tameem Albash and Daniel A. Lidar. “Adiabatic quantum computation”. Reviews of Modern Physics90, 015002 (2018)

work page 2018
[22]

Thermalization and its mecha- nism for generic isolated quantum systems

Marcos Rigol, Vanja Dunjko, and Maxim Olshanii. “Thermalization and its mecha- nism for generic isolated quantum systems”. Nature452, 854–858 (2008)

work page 2008
[23]

Many-body localization and thermalization in quantum statistical mechanics

Rahul Nandkishore and David A. Huse. “Many-body localization and thermalization in quantum statistical mechanics”. Annual Review of Condensed Matter Physics6, 15–38 (2015)

work page 2015
[24]

Col- loquium: Many-body localization, thermalization, and entanglement

Dmitry A. Abanin, Ehud Altman, Immanuel Bloch, and Maksym Serbyn. “Col- loquium: Many-body localization, thermalization, and entanglement”. Reviews of Modern Physics91, 021001 (2019). 17

work page 2019
[25]

D-Wave ocean software documentation: Reverse annealing

D-Wave Quantum Inc. “D-Wave ocean software documentation: Reverse annealing”. https://docs.dwavequantum.com(2024). Accessed 2026

work page 2024
[26]

Quan- tum mechanical evolution towards thermal equilibrium

Noah Linden, Sandu Popescu, Anthony J. Short, and Andreas Winter. “Quan- tum mechanical evolution towards thermal equilibrium”. Physical Review E79, 061103 (2009)

work page 2009
[27]

How “quantum

Seung Woo Shin, Graeme Smith, John A. Smolin, and Umesh Vazirani. “How “quantum” is the D-Wave machine?”. Preprint athttps://arxiv.org/abs/1401. 7087(2014)

work page 2014
[28]

Identifying quantum coherence in quantum annealers.arXiv preprint arXiv:2602.21355, 2026

Connor Aronoff, Travis Howard, David Nicholaeff, Alejandro Lopez-Bezanilla, and Wade DeGottardi. “Identifying quantum coherence in quantum annealers” (2026). arXiv:2602.21355

work page arXiv 2026
[29]

Why and when pausing is beneficial in quantum annealing

Huo Chen and Daniel A. Lidar. “Why and when pausing is beneficial in quantum annealing”. Physical Review Applied14, 014100 (2020)

work page 2020
[30]

Quantum thermalization through entangle- ment in an isolated many-body system

Adam M. Kaufman, M. Eric Tai, Alexander Lukin, Matthew Rispoli, Robert Schittko, Philipp M. Preiss, and Markus Greiner. “Quantum thermalization through entangle- ment in an isolated many-body system”. Science353, 794–800 (2016)

work page 2016
[31]

Many-body localization in a quantum simulator with pro- grammable random disorder

J. Smith, A. Lee, P. Richerme, B. Neyenhuis, P. W. Hess, P. Hauke, M. Heyl, D. A. Huse, and C. Monroe. “Many-body localization in a quantum simulator with pro- grammable random disorder”. Nature Physics12, 907–911 (2016)

work page 2016
[32]

Experimental signatures of hilbert-space ergodicity: Universal bitstring distributions and applications in noise learning

Adam L. Shaw, Daniel K. Mark, Joonhee Choi, Ran Finkelstein, Pascal Scholl, Soon- won Choi, and Manuel Endres. “Experimental signatures of hilbert-space ergodicity: Universal bitstring distributions and applications in noise learning”. Physical Review X15, 031001 (2025)

work page 2025
[33]

D-Wave system documentation: Solver properties, anneal schedules and qpu-specific parameters

D-Wave Quantum Inc. “D-Wave system documentation: Solver properties, anneal schedules and qpu-specific parameters”.https://docs.dwavequantum.com/docs/ latest/doc_physical_properties.html(2024). Accessed 2026

work page 2024
[34]

A practical heuristic for finding graph minors

Jun Cai, William G. Macready, and Aidan Roy. “A practical heuristic for finding graph minors”. Preprint athttps://arxiv.org/abs/1406.2741(2014)

work page internal anchor Pith review Pith/arXiv arXiv 2014
[35]

Perils of embedding for sampling problems

Jeffrey Marshall, Andrea Di Gioacchino, and Eleanor G. Rieffel. “Perils of embedding for sampling problems”. Physical Review Research2, 023020 (2020). 18

work page 2020

[1] [1]

Application of quantum annealing to training of deep neural networks

Steven H. Adachi and Maxwell P. Henderson. “Application of quantum annealing to training of deep neural networks”. Preprint athttps://arxiv.org/abs/1510. 06356(2015)

work page 2015

[2] [2]

Quantum Boltzmann machine

Mohammad H. Amin, Evgeny Andriyash, Jason Rolfe, Bohdan Kulchytskyy, and Roger Melko. “Quantum Boltzmann machine”. Physical Review X8, 021050 (2018)

work page 2018

[3] [3]

Estimation of effective temperatures in quantum annealers for sampling ap- plications: A case study with possible applications in deep learning

Marcello Benedetti, John Realpe-G´ omez, Rupak Biswas, and Alejandro Perdomo- Ortiz. “Estimation of effective temperatures in quantum annealers for sampling ap- plications: A case study with possible applications in deep learning”. Physical Review A94, 022308 (2016)

work page 2016

[4] [4]

High-quality thermal Gibbs sampling with quantum annealing hardware

Jon Nelson, Marc Vuffray, Andrey Y. Lokhov, Tameem Albash, and Carleton Coffrin. “High-quality thermal Gibbs sampling with quantum annealing hardware”. Physical Review Applied17, 044046 (2022)

work page 2022

[5] [5]

Programmable quantum annealers as noisy Gibbs samplers

Marc Vuffray, Carleton Coffrin, Yaroslav A. Kharkov, and Andrey Y. Lokhov. “Programmable quantum annealers as noisy Gibbs samplers”. PRX Quantum3, 020317 (2022)

work page 2022

[6] [6]

Quantum annealing and condensed matter physics

Viv Kendon and Nicholas Chancellor. “Quantum annealing and condensed matter physics”. Journal of Physics: Condensed Matter38, 143005 (2026)

work page 2026

[7] [7]

Global warming: Tempera- ture estimation in annealers

Jack Raymond, Sheir Yarkoni, and Evgeny Andriyash. “Global warming: Tempera- ture estimation in annealers”. Frontiers in ICT3, 23 (2016). 16

work page 2016

[8] [8]

Thermalization, freeze-out, and noise: Deciphering experimental quantum annealers

Jeffrey Marshall, Eleanor G. Rieffel, and Itay Hen. “Thermalization, freeze-out, and noise: Deciphering experimental quantum annealers”. Physical Review Applied8, 064025 (2017)

work page 2017

[9] [9]

Power of paus- ing: Advancing understanding of thermalization in experimental quantum annealers

Jeffrey Marshall, Davide Venturelli, Itay Hen, and Eleanor G. Rieffel. “Power of paus- ing: Advancing understanding of thermalization in experimental quantum annealers”. Physical Review Applied11, 044083 (2019)

work page 2019

[10] [10]

Classical thermometry of quantum an- nealers.arXiv preprint arXiv:2512.03162, 2025

George Grattan, Pratik Sathe, and Cristiano Nisoli. “Classical thermometry of quan- tum annealers” (2025). arXiv:2512.03162

work page arXiv 2025

[11] [11]

Erasing classical memory with quantum fluctuations: Shannon information entropy of reverse quantum annealing.arXiv preprint arXiv:2509.10927,

Elijah Pelofske and Cristiano Nisoli. “Erasing classical memory with quantum fluc- tuations: Shannon information entropy of reverse quantum annealing”. Preprint at https://arxiv.org/abs/2509.10927(2025)

work page arXiv 2025

[12] [12]

Finite-temperature criticality through quantum annealing.arXiv preprint arXiv:2507.07167, 2025

Gianluca Teza et al. “Finite-temperature criticality through quantum anneal- ing” (2025). arXiv:2507.07167

work page arXiv 2025

[13] [13]

Localization transition induced by programmable disorder

Jaime L. C. da C. Filho, Zoe Gonzalez Izquierdo, Andreia Saguia, Tameem Albash, Itay Hen, and Marcelo S. Sarandy. “Localization transition induced by programmable disorder”. Physical Review B105, 134201 (2022)

work page 2022

[14] [14]

Entanglement and the foundations of statistical mechanics

Sandu Popescu, Anthony J. Short, and Andreas Winter. “Entanglement and the foundations of statistical mechanics”. Nature Physics2, 754–758 (2006)

work page 2006

[15] [15]

Canon- ical typicality

Sheldon Goldstein, Joel L. Lebowitz, Roderich Tumulka, and Nino Zangh` ı. “Canon- ical typicality”. Physical Review Letters96, 050403 (2006)

work page 2006

[16] [16]

From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics

Luca D’Alessio, Yariv Kafri, Anatoli Polkovnikov, and Marcos Rigol. “From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics”. Advances in Physics65, 239–362 (2016)

work page 2016

[17] [17]

Coherent quantum annealing in a programmable 2,000 qubit Ising chain

Andrew D. King, Sei Suzuki, Jack Raymond, Alex Zucca, Trevor Lanting, Fabio Altomare, Andrew J. Berkley, et al. “Coherent quantum annealing in a programmable 2,000 qubit Ising chain”. Nature Physics18, 1324–1328 (2022)

work page 2022

[18] [18]

Quantum critical dynamics in a 5,000-qubit programmable spin glass

Andrew D. King, Jack Raymond, Trevor Lanting, Richard Harris, Alex Zucca, Fabio Altomare, Andrew J. Berkley, et al. “Quantum critical dynamics in a 5,000-qubit programmable spin glass”. Nature617, 61–66 (2023)

work page 2023

[19] [19]

Unraveling reverse annealing: A study of D-Wave quantum annealers

Vrinda Mehta, Hans De Raedt, Kristel Michielsen, and Fengping Jin. “Unraveling reverse annealing: A study of D-Wave quantum annealers”. Physical Review A112, 012414 (2025)

work page 2025

[20] [20]

Understanding the physics of D-Wave annealers: From Schr¨ odinger to Lindblad to Markovian dynamics

Vrinda Mehta et al. “Understanding the physics of D-Wave annealers: From Schr¨ odinger to Lindblad to Markovian dynamics”. Physical Review A112, 032616 (2025)

work page 2025

[21] [21]

Adiabatic quantum computation

Tameem Albash and Daniel A. Lidar. “Adiabatic quantum computation”. Reviews of Modern Physics90, 015002 (2018)

work page 2018

[22] [22]

Thermalization and its mecha- nism for generic isolated quantum systems

Marcos Rigol, Vanja Dunjko, and Maxim Olshanii. “Thermalization and its mecha- nism for generic isolated quantum systems”. Nature452, 854–858 (2008)

work page 2008

[23] [23]

Many-body localization and thermalization in quantum statistical mechanics

Rahul Nandkishore and David A. Huse. “Many-body localization and thermalization in quantum statistical mechanics”. Annual Review of Condensed Matter Physics6, 15–38 (2015)

work page 2015

[24] [24]

Col- loquium: Many-body localization, thermalization, and entanglement

Dmitry A. Abanin, Ehud Altman, Immanuel Bloch, and Maksym Serbyn. “Col- loquium: Many-body localization, thermalization, and entanglement”. Reviews of Modern Physics91, 021001 (2019). 17

work page 2019

[25] [25]

D-Wave ocean software documentation: Reverse annealing

D-Wave Quantum Inc. “D-Wave ocean software documentation: Reverse annealing”. https://docs.dwavequantum.com(2024). Accessed 2026

work page 2024

[26] [26]

Quan- tum mechanical evolution towards thermal equilibrium

Noah Linden, Sandu Popescu, Anthony J. Short, and Andreas Winter. “Quan- tum mechanical evolution towards thermal equilibrium”. Physical Review E79, 061103 (2009)

work page 2009

[27] [27]

How “quantum

Seung Woo Shin, Graeme Smith, John A. Smolin, and Umesh Vazirani. “How “quantum” is the D-Wave machine?”. Preprint athttps://arxiv.org/abs/1401. 7087(2014)

work page 2014

[28] [28]

Identifying quantum coherence in quantum annealers.arXiv preprint arXiv:2602.21355, 2026

Connor Aronoff, Travis Howard, David Nicholaeff, Alejandro Lopez-Bezanilla, and Wade DeGottardi. “Identifying quantum coherence in quantum annealers” (2026). arXiv:2602.21355

work page arXiv 2026

[29] [29]

Why and when pausing is beneficial in quantum annealing

Huo Chen and Daniel A. Lidar. “Why and when pausing is beneficial in quantum annealing”. Physical Review Applied14, 014100 (2020)

work page 2020

[30] [30]

Quantum thermalization through entangle- ment in an isolated many-body system

Adam M. Kaufman, M. Eric Tai, Alexander Lukin, Matthew Rispoli, Robert Schittko, Philipp M. Preiss, and Markus Greiner. “Quantum thermalization through entangle- ment in an isolated many-body system”. Science353, 794–800 (2016)

work page 2016

[31] [31]

Many-body localization in a quantum simulator with pro- grammable random disorder

J. Smith, A. Lee, P. Richerme, B. Neyenhuis, P. W. Hess, P. Hauke, M. Heyl, D. A. Huse, and C. Monroe. “Many-body localization in a quantum simulator with pro- grammable random disorder”. Nature Physics12, 907–911 (2016)

work page 2016

[32] [32]

Experimental signatures of hilbert-space ergodicity: Universal bitstring distributions and applications in noise learning

Adam L. Shaw, Daniel K. Mark, Joonhee Choi, Ran Finkelstein, Pascal Scholl, Soon- won Choi, and Manuel Endres. “Experimental signatures of hilbert-space ergodicity: Universal bitstring distributions and applications in noise learning”. Physical Review X15, 031001 (2025)

work page 2025

[33] [33]

D-Wave system documentation: Solver properties, anneal schedules and qpu-specific parameters

D-Wave Quantum Inc. “D-Wave system documentation: Solver properties, anneal schedules and qpu-specific parameters”.https://docs.dwavequantum.com/docs/ latest/doc_physical_properties.html(2024). Accessed 2026

work page 2024

[34] [34]

A practical heuristic for finding graph minors

Jun Cai, William G. Macready, and Aidan Roy. “A practical heuristic for finding graph minors”. Preprint athttps://arxiv.org/abs/1406.2741(2014)

work page internal anchor Pith review Pith/arXiv arXiv 2014

[35] [35]

Perils of embedding for sampling problems

Jeffrey Marshall, Andrea Di Gioacchino, and Eleanor G. Rieffel. “Perils of embedding for sampling problems”. Physical Review Research2, 023020 (2020). 18

work page 2020