Pith. sign in

REVIEW 2 major objections 6 minor 34 references

A quantum belief-update service on IBM Heron can run sequential Tiger POMDP steps without changing the planner’s action.

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 · grok-4.5

2026-07-10 21:53 UTC pith:IEHEKWKI

load-bearing objection Honest Heron case study: all-step FPAA keeps sequential Tiger posteriors planner-safe on shallow oracles, with action agreement and boundary BIQAE calibration. the 2 major comments →

arxiv 2607.06760 v1 pith:IEHEKWKI submitted 2026-07-07 cs.AI quant-ph

QANTIS: Hardware-Calibrated Sequential POMDP Belief Updates on IBM Heron

Bayram Yuksel Eker , Suayb S. Arslan , Ozgur Nazli , Mustafa Serhat Demirgil , Furkan Deligoz This is my paper
classification cs.AI quant-ph
keywords quantum computingPOMDP inferencebelief updatingamplitude amplificationBayesian amplitude estimationNISQ hardwareIBM Heronfixed-point amplitude amplification
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.

Autonomous systems under partial observability act on beliefs, not raw sensor events. This paper treats a quantum processor as a calibrated belief-update service: it takes a prior and an observation model, estimates the rare-event evidence term with amplitude amplification, and returns an ordinary posterior to a classical planner. The question is whether that service can be reused across a sequential Tiger POMDP horizon on present IBM Heron hardware without corrupting the posterior the planner sees. Controlled hardware runs show that all-step fixed-point amplification preserves the Tiger posterior on primary 8- and 12-step trajectories, that longer 20- and 32-step controls stay in the same operating band, and that the hardware posterior and exact Bayes posterior select the same immediate action in every reported decision check. Boundary-aware amplitude estimation keeps the estimator stable near zero and one, and a rare-event sweep maps the logical sample-complexity envelope down to one-in-a-million evidence. The result is an operating envelope for a hardware-calibrated belief-update primitive, not a wall-clock speedup or end-to-end autonomy claim.

Core claim

On present IBM Heron hardware, an all-step fixed-point amplitude amplification loop, paired with boundary-aware Bayesian amplitude estimation, can run sequential Tiger POMDP belief updates and return planner-facing posteriors that stay close to exact Bayes (max Hellinger 0.009 on the 8-step run and 0.021 on the 12-step run) and that select the same immediate action as exact Bayes under the standard Tiger reward rule, with zero measured cumulative value loss on the reported decision checks.

What carries the argument

QANTIS as a calibrated belief-update service: a shallow belief oracle plus all-step fixed-point amplitude amplification (softer phases that avoid overshoot so every listen step can be amplified) and boundary-aware BIQAE (a coarse scan that chooses a near-zero, near-one, or interior prior before fine estimation), feeding an ordinary classical Bayes update whose posterior becomes the next prior.

Load-bearing premise

The claim rests on shallow two-qubit Tiger oracles remaining a faithful planner-facing service; the paper itself treats circuit depth per belief update as the binding limit when encodings get richer.

What would settle it

Rerun the same sequential Tiger trajectory with all-step FPAA and boundary-aware BIQAE on Heron, compute hardware versus exact Bayes posteriors at every step, and check whether any step produces a different immediate action under the stated Tiger reward rule or pushes max Hellinger clearly outside the reported operating band.

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

If this is right

  • A classical planner can treat the quantum step as a drop-in rare-evidence posterior service and still receive ordinary probability distributions.
  • All-step fixed-point amplification removes the need for a skip guard that otherwise decides when amplification is safe.
  • Boundary-aware estimation is required infrastructure for sequential belief tracking, not a cosmetic fix, because loops repeatedly hit amplitudes near zero and one.
  • The useful operating regime is low-probability observations where classical sampling becomes expensive, not a claim that the whole autonomy stack becomes quantum.
  • Longer horizons and Heron transfer runs stay inside the same band only while compiled belief oracles remain shallow.

Where Pith is reading between the lines

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

  • If depth, not state-space size, is the first bottleneck, the next useful hardware milestone is shallower problem-specific belief encodings rather than more qubits alone.
  • The same service contract could be stress-tested on other small POMDP unit cells with closed-form Bayes references to separate amplification fidelity from Tiger-specific structure.
  • Once reliable queue-exclusive timing is logged, the natural next comparison is sample cost per accepted posterior against classical rare-event sampling at the same evidence rates.
  • Boundary-aware priors may transfer to other iterative quantum estimators that repeatedly visit near-zero or near-one amplitudes in closed loops.

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

2 major / 6 minor

Summary. The paper presents a controlled IBM Heron hardware case study of QANTIS as a calibrated sequential belief-update service for a two-state Tiger POMDP. The quantum step estimates a rare-event evidence amplitude (via amplification plus boundary-aware BIQAE); an ordinary classical Bayes update then returns a planner-facing posterior. On the same observation trajectory the authors compare no amplification, guarded Grover-AA, and all-step fixed-point AA (FPAA). Headline all-step FPAA reports max Hellinger 0.009 (8-step) and 0.021 (12-step) versus exact Bayes; 20- and 32-step controls stay in a similar band; and under a fixed Tiger immediate-reward rule every reported decision check shows action agreement with zero cumulative value loss. Supporting material includes same-backend BIQAE boundary calibration, a rare-event logical amplification envelope, Heron R3 transfer rows, and explicit non-claims on wall-clock speedup and end-to-end autonomy.

Significance. If the scoped result holds, the paper supplies a rare, auditable hardware operating envelope for sequential POMDP belief updates on present superconducting devices rather than a hardware-advantage claim. Strengths that should be credited include: (i) same-trajectory no-AA / guarded-Grover / all-step-FPAA controls with exact Bayes Hellinger references (Tables III–IV); (ii) planner-facing action and value-loss checks under an external Tiger reward rule (Table VI); (iii) a concrete two-phase boundary-aware BIQAE protocol with same-backend Pittsburgh pairs that collapse near-zero/near-one error (Table VII, Fig. 6); (iv) an explicit claim hierarchy (Table II) and resource accounting that separate primary 8/12-step results from longer-horizon and scaling probes; and (v) public code and artifact manifests. The contribution is narrow—two-qubit Tiger oracles at ISA depth ~15–18—but the systems framing (posterior service, not full autonomy stack) is appropriate and useful for the hybrid quantum–classical planning community.

major comments (2)
  1. [§III-D, Table IV, Abstract, Contribution 1] §III-D and Table IV: the primary Hellinger numbers used in the abstract and Contribution 1 (max 0.009 on the 8-step run) come from the 32 768-shot FPAA headline row, roughly 3× the ~10k per-step budget of the No-AA and guarded-Grover controls. The matched-shot FPAA control at 10k shots reports max Hellinger 0.033 / mean 0.019, which is not better than No-AA (0.029 / 0.015). The manuscript already notes that Table IV is not a fully budget-matched accuracy proof, but the abstract and contribution list still lead with 0.009 without the shot context. Because the central claim is that all-step FPAA can be reused without corrupting the planner-facing posterior, the abstract, Contribution 1, and the opening of §III-D should state the shot budget next to each headline Hellinger figure and more sharply separate “stability under all-step amplification (action preservation)” from “accuracy improvem
  2. [§III-C, §III-D, Table V] §III-C–D and Table V: the sequential results are largely single-trajectory (or few-repeat) runs on fixed observation sequences. Table V gives some Fez repeats, but there are no reported run-to-run distributions, confidence intervals, or multi-seed summaries for the primary 8-step Kingston Hellinger series or for the action-agreement checks. For a hardware case study whose product is an operating envelope, at least a short multi-execution summary (e.g., max/mean Hellinger over N independent submissions of the same 8-step path, or bootstrap intervals on the BIQAE estimates) is load-bearing for the claim that the service “preserves” the posterior across a horizon. Without it, the 0.009 / 0.021 figures remain point estimates that are hard to distinguish from favorable calibration windows.
minor comments (6)
  1. [Fig. 4, §III-D1] Fig. 4 / §III-D1: the rare-event “971832× logical amplification” row is useful as a sample-complexity envelope, but the accompanying text already notes that transpilation collapses the circuit to ≤6 single-qubit gates. Consider moving that caveat into the figure caption so the plot cannot be misread as a deep-circuit fidelity result.
  2. [Table III, Fig. 3] Table III vs Fig. 3: the 8-step audit table and the Hellinger plot are complementary; a one-sentence pointer in the table caption to Fig. 3 (and vice versa) would help readers who land on only one of them.
  3. [§IV-B] §IV-B: the noise-aware visibility is described as a session-level plug-in from short calibration circuits. State explicitly whether that visibility is frozen for the whole sequential trajectory or re-estimated per step, and whether the same plug-in is used for the No-AA baseline (which does not need BIQAE).
  4. [§V-A, Table VIII] §V-A / Table VIII: queue-inclusive and queue-exclusive timings are correctly declared out of scope, but the table still lists “resource accounting.” Renaming the table (e.g., “Shot and circuit accounting”) would avoid implying wall-clock comparison.
  5. [Table III] Notation: “listen(0)” / “listen(1)” in Table III are never defined in the main text (left vs right growl). A short footnote or parenthetical would remove ambiguity.
  6. [Related Work] Related work: the positioning against QANTIS v1 is clear; a single sentence on how the present all-step FPAA + boundary BIQAE stack differs from the multi-step OAA distortion analyses of Zecchi et al. and the QBRL simulation of Cunha et al. would further pin the empirical gap.

Circularity Check

0 steps flagged

No significant circularity: hardware posteriors are scored against independent exact classical Bayes and fixed external Tiger decision rules; self-citation to QANTIS v1 is background, not a load-bearing definition of the new claims.

full rationale

The paper is an empirical hardware case study, not a first-principles derivation that could close on its own inputs. The planner-facing product is an ordinary Bayes posterior; the quantum step only estimates the evidence amplitude, which is then plugged into the classical update and compared to the closed-form exact Bayes posterior on the same Tiger trajectory (Tables III–VI, Hellinger and action/value checks). Action agreement uses a fixed external immediate-reward rule (open above 90% / below 10%, listen in between), so agreement is not forced by construction of the estimator. All-step FPAA is taken from Yoder–Low–Chuang; BIQAE from Li et al.; rare-event logical amplification is measured against analytic Grover–Brassard targets, not fitted from the same counts. Visibility is explicitly a session-level plug-in from known-amplitude calibration circuits, with no joint-identifiability or prediction claim. Self-citation to QANTIS v1 supplies the guarded baseline, platform context, and deferred full derivations of standard Bayes/AA primitives; it does not define the new all-step FPAA, boundary-aware BIQAE, sequential Hellinger, or decision-check results. Those are independently measured on Heron against classical oracles and no-AA controls. No self-definitional loop, fitted-input-as-prediction, uniqueness import, or renamed known result is present. Score 0 is the honest finding.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 3 invented entities

The central operating-envelope claim rests on standard POMDP Bayes updates, established amplitude-amplification and fixed-point schedules, the classical Tiger model, and a small set of experimental knobs (shot budgets, boundary threshold, visibility plug-in). No new physical entities are postulated; QANTIS and boundary-aware BIQAE are service/protocol constructions. Free parameters are experimental choices that affect reported fidelity but are disclosed rather than hidden fits that define the result.

free parameters (5)
  • per-step shot budget (headline FPAA) = 32768
    32 768 shots/step for the headline all-step FPAA fidelity run; not matched to the ~10k no-AA/Grover budgets, so fidelity comparison is same-trajectory stability rather than equal-resource accuracy.
  • adaptive v1 shot allocation = ~15% reallocation; 8k–16k/step
    Moves ~15% of budget toward early/ambiguous steps (8k–16k range, ~10k average); shapes the guarded-Grover baseline resource profile.
  • BIQAE coarse-scan boundary threshold = 0.1
    Threshold 0.1 routes near-zero / interior / near-one priors after 200 coarse shots; directly controls which prior family is used at sequential boundaries.
  • noise-aware visibility plug-in = session-level estimate (not a single global number)
    Session-level visibility from short known-amplitude calibration circuits mixed into the BIQAE likelihood; not jointly identified with the unknown amplitude.
  • Tiger observation accuracy = 0.85
    Fixed 85% sensor accuracy in the unit-cell model; standard textbook choice that sets how fast beliefs concentrate and when amplification is risky.
axioms (6)
  • domain assumption POMDP beliefs update by Bayes rule: predict, weight by observation likelihood, normalize by evidence.
    Section II-A; defines the classical service interface the quantum step must not corrupt.
  • standard math Amplitude amplification reduces rare-event sampling cost from ~1/p toward ~1/√p under ideal reflections (Brassard–Høyer–Mosca).
    Cited [1]; motivation for using AA on the evidence term rather than hidden state.
  • standard math Yoder–Low–Chuang fixed-point phase schedules keep amplification monotone and reduce overshoot for concentrated amplitudes.
    Section III-D; justifies removing the skip guard and applying AA at every listen step.
  • domain assumption BIQAE maintains a Bayesian posterior over amplitude angle and updates with sinusoidal (or noise-aware) likelihoods after chosen Grover depths.
    Sections II-B and IV; estimation layer for converting hardware measurements into evidence probability.
  • domain assumption Standard Tiger immediate-reward rule: open right above 90% belief tiger is left, open left below 10%, listen otherwise; listen cheap, wrong door costly.
    Section III-D decision-impact paragraph and Table VI; maps posterior error into action agreement and value loss.
  • ad hoc to paper Hellinger distance is an appropriate bounded metric for planner-facing posterior fidelity in this study.
    Section III-A; chosen reporting metric for hardware vs exact Bayes; not forced by theory but used as the primary accuracy number.
invented entities (3)
  • QANTIS belief-update service contract no independent evidence
    purpose: Frames the quantum processor as prior+observation-model in, ordinary posterior out, with classical planner remaining outside the validated claim.
    Table I and Fig. 1; organizational construct for the case study, not a physical object with independent mass/charge prediction.
  • Boundary-aware two-phase BIQAE calibration protocol no independent evidence
    purpose: Routes near-zero/near-one/interior priors via a shallow coarse scan before fine BIQAE to stabilize sequential boundary amplitudes.
    Section IV and Fig. 5; new systems protocol layer on top of BIQAE; evidence is the same-backend error reductions in Table VII.
  • All-step FPAA sequential belief-tracking policy no independent evidence
    purpose: Apply fixed-point amplification at every listen step without a skip guard while preserving planner-facing posteriors.
    Section III-D; systems policy choice evaluated on Heron; independent of any new particle or force.

pith-pipeline@v1.1.0-grok45 · 19879 in / 4130 out tokens · 53817 ms · 2026-07-10T21:53:01.729430+00:00 · methodology

0 comments
read the original abstract

Autonomous systems under partial observability act on beliefs, not raw sensor events. QANTIS treats the quantum processor as a calibrated belief-update service in that loop: it receives a prior and an observation model, estimates the rare-event evidence term, and returns an ordinary posterior to a classical planner. This paper asks whether that service can be reused across a sequential Tiger POMDP horizon on present IBM Heron hardware without corrupting the planner-facing posterior. We answer with a controlled hardware case study rather than an end-to-end autonomy or wall-clock speedup claim. The study compares no amplification, guarded Grover amplification, and all-step fixed-point amplification on the same trajectory, then checks whether the returned posterior would change the downstream action. All-step FPAA preserves the Tiger posterior across the reported 8-step and 12-step primary runs, and the 20-step and 32-step controls remain inside the same operating band. In every reported decision check, the hardware posterior and the exact Bayes posterior select the same immediate action. Boundary-aware BIQAE stabilizes amplitude estimation near zero and near one, while a rare-event sweep maps the logical sample-complexity envelope for one-in-a-million evidence. The result is an operating envelope for a hardware-calibrated belief-update primitive, not a standalone hardware-advantage claim.

Figures

Figures reproduced from arXiv: 2607.06760 by Bayram Yuksel Eker, Furkan Deligoz, Mustafa Serhat Demirgil, Ozgur Nazli, Suayb S. Arslan.

Figure 1
Figure 1. Figure 1: The paper uses a small acronym set repeatedly: amplitude amplification (AA) boosts rare evidence events; fixed-point AA (FPAA) uses softer phases to avoid overshoot; BIQAE estimates the resulting amplitude; and the POMDP is the classical belief￾and-action model. The reason for using AA is simple: in the rare-event regime it changes the inference-side sampling burden from inverse-evidence sampling toward in… view at source ↗
Figure 2
Figure 2. Figure 2: Hardware posterior extraction and per-step error budget. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Per-step Hellinger distance on the shared 8-step decision-step path. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Rare-observation envelope for Grover–Brassard iterates on a one [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Two-phase boundary-aware BIQAE calibration. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Boundary behavior of BIQAE on hardware. Calibration collapses the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

34 extracted references · 34 canonical work pages · 12 internal anchors

  1. [1]

    Quantum amplitude amplification and estimation,

    G. Brassard, P. Høyer, M. Mosca, and A. Tapp, “Quantum amplitude amplification and estimation,” inQuantum Computation and Information, ser. Contemporary Mathematics. American Mathematical Society, 2002, vol. 305, pp. 53–74

    work page 2002

  2. [2]

    QANTIS: A hardware-validated quantum platform for POMDP planning and multi-target data association,

    B. Y . Eker, S. S. Arslan, O. Nazlı, M. S. Demirgil, and F. Delig ¨oz, “QANTIS: A hardware-validated quantum platform for POMDP planning and multi-target data association,” 2026, arXiv:2603.00785

    work page arXiv 2026

  3. [3]

    Quantum POMDPs

    S. Aaronson, “Quantum POMDPs,” 2014, arXiv:1406.2858

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  4. [4]

    Quantum Bayesian Networks Can Speed up Reinforcement Learning in Partially Observable Environments

    G. Cunha, A. Ram ˆoa, A. Sequeira, M. de Oliveira, and L. Barbosa, “Hy- brid quantum-classical algorithm for near-optimal planning in POMDPs,” 2025, arXiv:2507.18606

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Complexity scaling and optimal policy degeneracy in quantum reinforcement learning via analytically solvable unitary-control-then-measure models

    P. Cintio, A. Michelangeli, and A. Tsutskov, “Complexity scaling and optimal policy degeneracy in quantum reinforcement learning via unitary- control-then-measure models,” Apr. 2026, arXiv:2604.13096

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  6. [6]

    Projected Dynamic Programming for Sequential Quantum State Discrimination

    J. Jeong, D. Ji, H. Jang, and K. Jeong, “Projected dynamic pro- gramming for sequential quantum state discrimination,” Apr. 2026, arXiv:2604.15393

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  7. [7]

    Improved amplitude amplification strategies for the quantum simulation of classical transport problems

    A. A. Zecchi, C. Sanavio, S. Perotto, and S. Succi, “Improved amplitude amplification strategies for the quantum simulation of classical transport problems,” 2025, arXiv:2502.18283

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  8. [8]

    Grover’s algorithm is an approximation of imaginary-time evolution,

    Y . Suzuki, M. Gluza, J. Son, B. H. Tiang, N. H. Y . Ng, and Z. Holmes, “Grover’s algorithm is an approximation of imaginary-time evolution,” 2025, arXiv:2507.15065

    work page arXiv 2025

  9. [9]

    Benchmarking quantum reinforcement learning,

    N. Meyer, C. Ufrecht, G. Yammine, G. Kontes, C. Mutschler, and D. D. Scherer, “Benchmarking quantum reinforcement learning,” inProceedings of the 42nd International Conference on Machine Learning (ICML), 2025

    work page 2025

  10. [10]

    Harnessing Bayesian statistics to accelerate iterative quantum amplitude estimation,

    Q. Li, A. Vidwans, Y . Wang, and M. B. Soley, “Harnessing Bayesian statistics to accelerate iterative quantum amplitude estimation,”Quantum, vol. 10, p. 1962, 2026

    work page 1962

  11. [11]

    Bayesian quantum amplitude estimation,

    A. Ram ˆoa and L. P. Santos, “Bayesian quantum amplitude estimation,” Quantum, vol. 9, p. 1856, 2025, arXiv:2412.04394

    work page arXiv 2025

  12. [12]

    On the bias in iterative quantum amplitude estimation,

    K. Miyamoto, “On the bias in iterative quantum amplitude estimation,” EPJ Quantum Technology, vol. 11, no. 1, p. 42, 2024

    work page 2024

  13. [13]

    Quantum annealing applied to de-conflicting optimal trajectories for air traffic management,

    T. Stollenwerk, B. O’Gorman, D. Venturelli, S. Mandra, O. Rodionova, H. K. Ng, B. Sridhar, E. G. Rieffel, and R. Biswas, “Quantum annealing applied to de-conflicting optimal trajectories for air traffic management,” 2021

    work page 2021

  14. [14]

    Enhancing multiple object tracking accuracy via quantum annealing,

    Y . Ihara, “Enhancing multiple object tracking accuracy via quantum annealing,”Scientific Reports, vol. 15, p. 24294, 2025

    work page 2025

  15. [15]

    Quantum annealing vs. QAOA: 127 qubit gate-model IBM quantum computer and D-Wave advantage,

    E. Pelofske, A. B ¨artschi, and S. Eidenbenz, “Quantum annealing vs. QAOA: 127 qubit gate-model IBM quantum computer and D-Wave advantage,”npj Quantum Information, vol. 10, p. 18, 2024

    work page 2024

  16. [16]

    Planning and acting in partially observable stochastic domains,

    L. P. Kaelbling, M. L. Littman, and A. R. Cassandra, “Planning and acting in partially observable stochastic domains,”Artificial Intelligence, vol. 101, no. 1–2, pp. 99–134, 1998

    work page 1998

  17. [17]

    Monte-Carlo planning in large POMDPs,

    D. Silver and J. Veness, “Monte-Carlo planning in large POMDPs,” inAdvances in Neural Information Processing Systems 23 (NeurIPS). Curran Associates, Inc., 2010, pp. 2164–2172

    work page 2010

  18. [18]

    M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information, 10th ed. Cambridge University Press, 2010

    work page 2010

  19. [19]

    The Hungarian method for the assignment problem,

    H. W. Kuhn, “The Hungarian method for the assignment problem,”Naval Research Logistics Quarterly, vol. 2, no. 1–2, pp. 83–97, 1955

    work page 1955

  20. [20]

    Bar-Shalom and X.-R

    Y . Bar-Shalom and X.-R. Li,Multitarget-Multisensor Tracking: Principles and Techniques. Storrs, CT: YBS Publishing, 1995

    work page 1995

  21. [21]

    An algorithm for tracking multiple targets,

    D. B. Reid, “An algorithm for tracking multiple targets,”IEEE Transac- tions on Automatic Control, vol. 24, no. 6, pp. 843–854, 1979

    work page 1979

  22. [22]

    Quantum approximate optimization algorithm with fixed number of parameters,

    S. Saavedra-Pino, R. Quispe-Mendiz´abal, G. Alvarado Barrios, E. Solano, J. C. Retamal, and F. Albarr ´an-Arriagada, “Quantum approximate optimization algorithm with fixed number of parameters,” 2025

    work page 2025

  23. [23]

    A quantum approximate optimization algorithm,

    E. Farhi, J. Goldstone, and S. Gutmann, “A quantum approximate optimization algorithm,” 2014

    work page 2014

  24. [24]

    Fixed-point quantum search with an optimal number of queries,

    T. J. Yoder, G. H. Low, and I. L. Chuang, “Fixed-point quantum search with an optimal number of queries,”Physical Review Letters, vol. 113, p. 210501, 2014

    work page 2014

  25. [25]

    Grover’s quantum searching algorithm is optimal,

    C. Zalka, “Grover’s quantum searching algorithm is optimal,”Physical Review A, vol. 60, no. 4, pp. 2746–2751, 1999

    work page 1999

  26. [26]

    Dequantizing Short-Path Quantum Algorithms

    F. Le Gall and S. Tamaki, “Dequantizing short-path quantum algorithms,” Apr. 2026, arXiv:2604.12131. TABLE X OPERATIONAL READING OF THEQANTISVALIDATION RESULTS. THE TABLE SUMMARIZES WHAT EACH EXPERIMENT ESTABLISHES FOR AN AUTONOMOUS DECISION STACK WHILE KEEPING DEPLOYMENT LIMITS EXPLICIT. Validation component Evidence in this paper Operational reading Her...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  27. [27]

    A Nested Amplitude Amplification Protocol for the Binary Knapsack Problem

    L. Demmler and M. Hess, “A nested amplitude amplification protocol for the binary knapsack problem,” Apr. 2026, arXiv:2604.05776

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  28. [28]

    Asmussen and P

    S. Asmussen and P. W. Glynn,Stochastic Simulation: Algorithms and Analysis, ser. Stochastic Modelling and Applied Probability. New York: Springer, 2007, vol. 57

    work page 2007

  29. [29]

    Confidence Intervals for Rate Estimation with Importance Sampling in Autonomous Vehicle Evaluation

    A. Chen, R. R. Zhou, J. J. Lee, N. Chamandy, and H. Hohnhold, “Confidence intervals for rate estimation with importance sampling in autonomous vehicle evaluation,” Apr. 2026, arXiv:2604.03827

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  30. [30]

    Efficient tensor network simulation of IBM's Eagle kicked Ising experiment

    J. Tindall, M. Fishman, E. M. Stoudenmire, and D. Sels, “Efficient tensor network simulation of IBM’s Eagle kicked Ising experiment,” PRX Quantum, vol. 5, p. 010308, 2024, arXiv:2306.14887

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  31. [31]

    Fast and converged classical simulations of evidence for the utility of quantum computing before fault tolerance,

    T. Beguˇsi´c, J. Gray, and G. K.-L. Chan, “Fast and converged classical simulations of evidence for the utility of quantum computing before fault tolerance,”Science Advances, vol. 10, p. eadk4321, 2024

    work page 2024

  32. [32]

    Challenging the Quantum Advantage Frontier with Large-Scale Classical Simulations of Annealing Dynamics

    L. Mauron and G. Carleo, “Challenging the quantum advantage frontier with large-scale classical simulations of annealing dynamics,” Mar. 2025, arXiv:2503.08247

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  33. [33]

    Belief Propagation and Tensor Network Expansions for Many-Body Quantum Systems: Rigorous Results and Fundamental Limits

    T. Midha, G.-M. Sommers, J. Tindall, and D. Abanin, “Belief propagation and tensor network expansions for many-body quantum systems: Rigorous results and fundamental limits,” Apr. 2026, arXiv:2604.03228

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  34. [34]

    Extreme Quantum Advantage for Rare-Event Sampling

    C. Aghamohammadi, S. P. Loomis, J. R. Mahoney, and J. P. Crutch- field, “Extreme quantum advantage for rare-event sampling,” 2017, arXiv:1707.09553; Phys. Rev. X 8, 011025 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2017