Pith. sign in

REVIEW 5 major objections 5 minor 27 references

How Many Shots Does It Take? A Noise-Aware Quantum Resource Allocation Framework

T0 review · 5 major / 5 minor · reviewed 2026-07-31 · grok-4.5

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

Pith's one-line read A closed-form noise-aware formula gives the exact number of measurement shots a quantum algorithm needs and how to split a fixed shot budget across circuit partitions to minimize total error.

desk verdict Useful closed-form packaging of shot count and variance-proportional partition allocation, with a real but contained soft spot in the F_max majority-readout success model. read the letter →

arxiv 2607.24704 v1 pith:YXBI4QVG submitted 2026-07-27 quant-ph cs.ET

Prateek P. Kulkarni , Sumit K. Mandal This is my paper · ORCID
classification quant-phcs.ET
keywords quantumresourceestimationmeasurementshotscircuitpartitioningnoise-awareallocationshotbudgetdecoherenceNISQ
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

Running any algorithm on today's quantum computers means repeating the circuit many times (shots) because results are noisy and probabilistic; those shots are expensive in time, money and energy. This paper derives a closed-form expression that converts a target success probability plus ordinary hardware numbers (readout fidelity, T1, T2, gate error, circuit time) into the exact shot count required. When a circuit is deeper than the device can handle, the same model yields the minimum number of partitions and proves that the fixed shot budget should be shared in proportion to each partition's noise standard deviation, not equally. On real IBM processors the formulas cut required shots by roughly 58 percent, energy by up to 62 percent, and partition estimation error by about 63 percent versus equal allocation. A reader who cares about making near-term quantum algorithms practical therefore has an analytic way to stop over-sampling and to spend every remaining shot where it most reduces error.

What carries the argument

The invertible shot formula (Eqs. 6/8) that maps target success probability, F_max and additive noise variance onto s, together with the optimal allocation rule s*_i = S · σ_i / Σ_j σ_j that equalises marginal error reduction across partitions.

What would settle it

Execute the same benchmark algorithms on the same calibrated IBM devices while logging both majority-vote readout success and true algorithmic success (correct answer); if the shot counts predicted by the formula reach the target algorithmic success only at substantially higher shot numbers than predicted, the central claim fails.

Watch

Extended reading notes

Core claim

The authors obtain a closed-form expression for the number of shots s needed to reach a prescribed success probability P(s) by modelling each shot as an independent Bernoulli trial with success probability equal to the device's maximum readout fidelity and adding a hardware noise variance that depends on decoherence times, gate error and circuit duration; inverting the resulting Gaussian tail (via the inverse error function) yields s directly. From the same relation they extract the maximum executable depth and the minimum number of partitions. Under a fixed total budget they prove, via Lagrange multipliers and Cauchy-Schwarz, that allocating shots proportional to each partition's measuremen

Load-bearing premise

Reliable execution is identified with obtaining a majority of correct binary readouts under the single best qubit's fidelity, rather than with the algorithm returning the correct computational answer after the full noisy circuit has run.

Editorial extensions

If this is right

  • Cloud users can request roughly 58 percent fewer shots for a chosen reliability target, cutting queue time and billing cost.
  • Energy consumed per thousand shots falls by up to 62 percent when shots are variance-weighted across partitions.
  • Deep circuits receive an a-priori lower bound on the number of partitions and an optimal shot split that reduces total estimation error by about 63 percent versus equal allocation.
  • Device calibration data alone suffice to compute the maximum circuit depth supportable at any chosen success probability.

Reading between the lines

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

  • The same variance-proportional rule can be nested inside variational algorithms that already reallocate shots among Hamiltonian terms, giving a single principle for both partition-level and term-level budgeting.
  • When readout fidelity varies strongly across qubits, first mapping high-variance partitions onto higher-fidelity qubits and then applying the shot formula should compound the measured gains.
  • Classical noisy simulators that currently fix large shot counts could adopt the closed-form expression as an automatic stopping criterion, saving classical compute.
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.

Referee Report

5 major / 5 minor

Summary. The manuscript proposes a closed-form framework for quantum shot budgeting. Three components: (1) Eq. (6)/(8), a formula for the number of shots s needed to reach target success probability P(s), derived by modeling each shot as i.i.d. Bernoulli(F_max) — F_max being the best-qubit readout fidelity — applying a CLT approximation with an additive hardware-noise variance σ²_noise taken from Ref. [21] (Eq. 7), and defining reliability as a majority-vote event P(S_n ≥ s/2); (2) Eq. (9)/(10), a maximum executable depth d_max and minimum partition count m_min obtained by truncating the noise model to σ²_noise ≈ (n_q/2s)(d·t_g/T2)²; (3) Eq. (12), the optimal allocation of a fixed shot budget across circuit partitions, s*_i = S·σ_i/Σ_j σ_j, proven optimal via Cauchy–Schwarz. Experiments on IBM Marrakesh/Torino/Fez report 98.2% accuracy for the shot model, ~58.6% shot reduction, up to 62% energy savings, and 45–63% estimation-error reduction versus equal allocation.

Significance. If the shot formula can be repaired or honestly re-scoped, the framework would be genuinely useful: it offers (i) a closed-form, hardware-parameter-only estimate of shot requirements, (ii) a provably optimal allocation rule (with a correct Cauchy–Schwarz optimality proof — credit where due), (iii) validation on three real IBM QPUs rather than simulation alone, and (iv) a full-stack energy accounting, which fits the venue. The variance-proportional allocation is not new mathematically (it is Neyman allocation), but its instantiation for partitioned circuits with per-partition noise variances from calibration data is a practical contribution. As it stands, however, the central quantitative claims (58% shot reduction, 62% energy savings) rest on a per-shot success model that is structurally optimistic for the large, deep circuits the paper targets, so the significance is conditional on Major Comment 1 being addressed.

major comments (5)
  1. [§4.1, Eqs. (1)–(6)] Central structural concern. Each shot is modeled as Bernoulli(F_max), with F_max = 1 − p_readout of the *best* qubit (§3). Consequently the binomial mean μ = s·F_max never degrades with qubit count or depth; all circuit noise enters only through the additive variance σ²_noise (Eq. 7), which widens the distribution but leaves the mean pinned. Physically, per-shot success decays roughly multiplicatively: even under readout error alone it is Π_i F_i ≈ F_avg^{n_q} (~0.74 at F_avg=0.99, n_q=30), and gate/decoherence errors over depth ~1200 reduce it further. Eq. 6 therefore systematically under-predicts s in exactly the large, deep regimes targeted (Shor's, 30 qubits, d≈1200 with [[5,1,3]] overhead). A concrete check: measure empirical per-shot success on a depth-matched circuit and compare with F_max.
  2. [§5.1, Fig. 3(a)] The reported 98.2% average accuracy of Eq. 6 hinges on how 'experimentally observed' required shots were determined. If the ground-truth criterion is the paper's own majority-readout definition P(S_n ≥ s/2) with success = correct readout, the validation checks the CLT arithmetic rather than algorithmic reliability, and is partially circular. Please specify the experimental success criterion explicitly; a correct-algorithm-output criterion (or a direct measurement of per-shot success vs. F_max) would substantially strengthen this section.
  3. [§5.2, Fig. 4] Since Eq. 12 reallocates a *fixed* budget S (Eq. 11), it cannot by itself change energy consumption; the savings must come from Eq. 6's shot reduction. Yet Fig. 4 attributes 57–62% energy savings to 'Our Optimal Allocation' vs. 'State-of-the-Art equal shot allocation'. As written the two mechanisms (shot reduction via Eq. 6, error reduction via Eq. 12) are conflated, and the 62% figure is not traceable. Please disentangle which mechanism produces which saving and state the baseline's total shot count.
  4. [§4.2, Eq. (9)] d_max (Eq. 9) is derived by 'suppressing the higher-order terms' of Eq. 7 to get σ²_noise ≈ (n_q/2s)(d·t_g/T2)². This truncation is valid only for t_circ = d·t_g ≪ T1, T2, but d_max is precisely the regime where t_circ/T2 becomes O(1); the approximation is least reliable exactly where the formula is used. Please state the validity range and quantify the truncation error at the estimated d_max values (e.g., 285 for Marrakesh), ideally against the untruncated Eq. 7.
  5. [Abstract / §1 / §5.3 / §6] The headline numbers are mutually inconsistent: the abstract claims error reduction 'up to 73%', the introduction 'about 53%' on average, §5.3 reports 45% average and 63% with AutoCut, and the conclusion states 63% 'on average'. Similarly energy is 62% (abstract/Fig. 4), 59% (intro/conclusion), and shots 58%/58.6%/59% in different places. The abstract's 73% appears nowhere supported in the body (Fig. 6's geomean bar suggests 63%). Please reconcile these so the abstract reflects the reported experiments.
minor comments (5)
  1. [§4.1, Eq. (6)] Eq. (6) implicitly requires F_max > 1/2 (otherwise the erf argument changes sign and majority vote converges to failure). Please state this assumption explicitly and comment on behavior as F_max → 1/2.
  2. [§5.3] 'we apply our optimal allocation strategy (Eq. 6)' should reference Eq. (12), not Eq. (6).
  3. [Fig. 6 vs. §5] Fig. 6 evaluates QFT, QAOA, VQE, GHZ, and Grover, but §5 lists QFT, QST, VQE, QAOA, and Shor's as the benchmark suite. Please explain the discrepancy (where are QST and Shor's in Fig. 6; where do GHZ/Grover come from?).
  4. [§4.3, Eq. (12)] Eq. (12) yields real-valued allocations; in practice s_i must be positive integers. Briefly comment on rounding and the handling of partitions allocated < 1 shot. Also, the allocation is the classical variance-proportional (Neyman) result; citing that lineage would help readers.
  5. [§5.2–5.3, Figs. 1, 5] Figure 5 is referenced in §5.2 before Figure 4 appears; consider reordering. Fig. 5's caption ('identifies optimal shot count') gives no axis information. Fig. 1's 'Equal Partitions' legend is ambiguous. 'upto' → 'up to' throughout. Reference [21] lists arXiv:2501.03194 with year 2026 — please check.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: shot formula and variance-proportional allocation are self-contained inversions/optimizations validated against external IBM runs.

full rationale

The load-bearing derivations do not reduce to their inputs by construction. Section 4.1 defines per-shot outcomes as i.i.d. Bernoulli(F_max), applies the CLT with an additive noise variance, defines P(s) as majority-vote success P(S_n ≥ s/2), and algebraically inverts the resulting erf expression to obtain s (Eqs. 5–6/8). That is definition-plus-inversion, not a fitted quantity renamed as a prediction. The noise variance (Eq. 7) is imported from an external citation [21] (Seksaria & Prabhakar), not from the present authors. The partition depth bound (Eq. 9) and m_min (Eq. 10) are sequential uses of the same model, which is framework composition rather than circular proof. The fixed-budget allocation (Eq. 12) is the standard Lagrange / Cauchy–Schwarz result s_i ∝ σ_i minimizing Σ σ_i²/s_i; optimality is proved in-paper and does not rely on self-citation or on fitting to the claimed error reductions. Empirical claims (Fig. 3 accuracy, ~58% shot cut, error vs equal allocation) compare the closed form to IBM hardware runs and to an equal-shot baseline, using vendor calibration parameters (F_max, T1, T2, p_g) as external inputs. No self-citation chain, uniqueness theorem from the same authors, or ansatz smuggled via prior work by Kulkarni/Mandal appears. Correctness concerns about whether Bernoulli(F_max) adequately models algorithmic success are outside the scope of circularity.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claims rest on a readout-majority Bernoulli model, CLT, a cited noise-variance formula, mean gate-time depth, and classical optimal allocation under additive squared error—plus a chosen target P(s) and IBM calibration inputs. No new physical entities are postulated. The largest conceptual load is equating algorithm reliability with majority-correct readout at F_max.

free parameters (3)
  • Target success probability P(s) = 0.95 (main experiments)
    Chosen by the authors for experiments (primarily 0.95; sweeps 0.75–0.95). Directly sets the erf^{-1} factor and thus predicted shot counts.
  • F_max (maximum qubit readout fidelity) = Device calibration (IBM platform)
    Taken from device calibration as 1 - p_readout on the best qubit and used as the global Bernoulli success probability for the whole algorithm. Device- and time-dependent input that dominates Eq. 6/8.
  • Effective average gate time t_g
    Used in t_circ ≈ d · t_g and in d_max; treated as a single characteristic time for mixed one- and two-qubit gates.
assumptions (6)
  • ad hoc to paper Successive shots are i.i.d. Bernoulli(F_max); algorithm reliability is P(S_n ≥ s/2) (majority vote).
    Section 4.1 motivates this via fault-tolerance majority-vote citations, but applies it uniformly as readout-level reliability for any algorithm—load-bearing modeling choice, not a theorem about algorithmic success.
  • standard math For large s, S_n is approximately normal with variance σ² + σ²_noise (CLT).
    Standard CLT step from Eq. 1 to Eqs. 2–5.
  • domain assumption Hardware noise variance follows the Seksaria–Prabhakar form (Eq. 7) in T1, T2, p_g, t_circ, n_q, s.
    Imported from [21]; entire closed form and d_max inherit its structure and approximations.
  • domain assumption Circuit time is t_circ ≈ d · t_g with a single average gate duration.
    Section 3 and 4.2; common resource-estimation approximation, acknowledged as simplified.
  • domain assumption Total estimation error is E = Σ_i σ_i² / s_i with independent partition variances; minimize under Σ s_i = S.
    Section 4.3; standard additive variance model—ignores reconstruction/cut classical post-processing correlations from circuit cutting.
  • ad hoc to paper Higher-order terms in the noise exponential can be dropped so σ²_noise ≈ (n_q/(2s))(d t_g/T2)² when deriving d_max.
    Section 4.2 explicit simplification before Eq. 9.

how reviews work

0 comments
Cite this review

Pith. "Pith review of How Many Shots Does It Take? A Noise-Aware Quantum Resource Allocation Framework." pith.science (2026). https://pith.science/paper/YXBI4QVG

@misc{pith2026260724704,
  author       = {Pith},
  title        = {Pith review of: How Many Shots Does It Take? A Noise-Aware Quantum Resource Allocation Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YXBI4QVG}},
  note         = {Machine review of arXiv:2607.24704}
}
read the original abstract

Any algorithm execution on quantum computers requires several repeated and costly executions (known as shots) to obtain reliable results. In this work, we propose a closed-form accurate analytical expression to determine optimal number of shots required for reliable execution of any algorithm on a quantum computer. We also present a theoretically grounded technique to distribute fixed shot budget across different partitions in a quantum circuit minimizing the total error. Our proposed analytical model helps to reduce the shots associated with reliable execution of quantum algorithms by about 58\% compared to current practice, in turn reducing the energy consumption by upto 62\%. Furthermore, our proposed optimal shot allocation technique across different partitions reduces total error by up to 73\% compared to conventional approaches.

Figures

Figures reproduced from arXiv: 2607.24704 by the authors.

Figure 1
Figure 1. Estimation error for quantum algorithms across all [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the Proposed Framework. The workflow begins with hardware parameters and algorithm specifications, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Experimental validation of proposed analytical model on IBM Marrakesh at [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Energy consumption per 1000 shots for different [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Normalized estimation error using equal allocation [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 1 canonical work pages

  1. [21]

    Manav Seksaria and Anil Prabhakar. 2026. Estimating shots and variance on noisy quantum circuits. arXiv:2501.03194 [quant-ph] https://arxiv.org/abs/2501.03194

    arXiv 2026

  2. [1]

    Dorit Aharonov and Michael Ben-Or. 1997. Fault-tolerant quantum computation with constant error. InProceedings of the twenty-ninth annual ACM symposium on Theory of computing. 176–188

    1997

  3. [2]

    Pablo Andrés-Martínez and Chris Heunen. 2019. Automated distribution of quantum circuits via hypergraph partitioning.Physical Review A100 (2019), 032308. arXiv:1811.10972 doi:10.1103/PhysRevA.100.032308

    arXiv 2019

  4. [3]

    Samantha V Barron, Daniel J Egger, Elijah Pelofske, Andreas Bärtschi, Stephan Eidenbenz, Matthis Lehmkuehler, and Stefan Woerner. 2024. Provable bounds for noise-free expectation values computed from noisy samples.Nature Compu- tational Science4, 11 (2024), 865–875

    2024

  5. [4]

    Radoslaw C Bialczak, Markus Ansmann, Max Hofheinz, Erik Lucero, Matthew Neeley, Aaron D O’Connell, Daniel Sank, Haohua Wang, James Wenner, Matthias Steffen, et al. 2010. Quantum process tomography of a universal entangling gate implemented with Josephson phase qubits.Nature Physics6, 6 (2010), 409–413

    2010

  6. [5]

    Brańczyk, Almudena Carrera Vazquez, Daniel J

    Agata M. Brańczyk, Almudena Carrera Vazquez, Daniel J. Egger, Bryce Fuller, Julien Gacon, James R. Garrison, Jennifer R. Glick, Caleb Johnson, Saasha Joshi, Edwin Pednault, C. D. Pemmaraju, Pedro Rivero, Ibrahim Shehzad, and Stefan Woerner. 2024. Qiskit addon: circuit cutting. https://github.com/Qiskit/qiskit- addon-cutting. doi:10.5281/zenodo.7987997

    doi:10.5281/zenodo.7987997 2024

  7. [6]

    Sebastian Brandhofer, Ilia Polian, and Kevin Mc Kevitt. 2023. Optimal Parti- tioning of Quantum Circuits using Gate Cuts and Wire Cuts.arXiv preprint arXiv:2308.09567(2023)

    arXiv 2023

  8. [7]

    Sergey Bravyi, David Gosset, and Robert König. 2018. Quantum advantage with shallow circuits.Science362, 6412 (2018), 308–311

    2018

Show all 27 references
  1. [8]

    Rafael P. H. Enriquez et al . 2023. Estimating Energy-Efficiency in Quantum Optimization Algorithms. InProceedings of the Cray User Group Conference (CUG)

    2023

  2. [9]

    Craig Gidney and Martin Ekerå. 2021. How to factor 2048 bit RSA integers in 8 hours using 20 million noisy qubits.Quantum5 (2021), 433

    2021

  3. [10]

    Zhibo Hou, Guo-Yong Zhang, Yu Wang, Yaxin Li, Yifei Gu, Han Li, Huangjun Zhu, Guo-Long Li, Xiang Wang, Z Song, et al. 2021. Benchmarking quantum state tomography on a 32-qubit superconducting quantum processor.PRX Quantum2, 4 (2021), 040311

    2021

  4. [11]

    IBM Quantum Platform. 2025. QPU information. https://quantum.cloud.ibm. com/docs/en/guides/qpu-information. [Accessed: 17-11-2025]

    2025

  5. [12]

    Kessler, D

    M. Kessler, D. Alonso, and P. Sánchez. 2023. Determination of the number of shots for Grover’s search algorithm.EPJ Quantum Technology10, 1 (2023), 47. doi:10.1140/epjqt/s40507-023-00204-y

    2023 doi

  6. [13]

    Senwei Liang, Linghua Zhu, Xiaolin Liu, Chao Yang, and Xiaosong Li. 2024. Artificial-intelligence-driven shot reduction in quantum measurement.Chemical Physics Reviews5, 4 (2024), 041403. arXiv:2405.02493 doi:10.1063/5.0219663

    2024 arXiv

  7. [14]

    Enrique Martin-López, Anthony Laing, Thomas Lawson, Roberto Alvarez, Xian- Qi Zhou, and Jeremy L O’Brien. 2012. Experimental realization of Shor’s quantum factoring algorithm using qubit recycling.Nature Photonics6, 11 (2012), 773–776

    2012

  8. [15]

    Mezher, J

    R. Mezher, J. Mills, and E. et al. Kashefi. 2025. Error Mitigation of BQP Com- putations using Measurement-Based Verification.Physical Review A111 (2025), 022602. arXiv:2306.04351 doi:10.1103/PhysRevA.111.022602

    2025 arXiv

  9. [16]

    2010.Quantum computation and quantum information

    Michael A Nielsen and Isaac L Chuang. 2010.Quantum computation and quantum information. Cambridge university press

    2010

  10. [17]

    Alberto Peruzzo, Jarrod McClean, Peter Shadbolt, Man-Hong Yung, Xiao-Qi Zhou, Peter J Love, Alán Aspuru-Guzik, and Jeremy L O’Brien. 2014. A variational eigenvalue solver on a photonic quantum processor.Nature Communications5, 1 (2014), 4213

    2014

  11. [18]

    Koustubh Phalak and Swaroop Ghosh. 2023. Shot Optimization in Quantum Machine Learning Architectures to Accelerate Training.IEEE Access11 (2023), 41514–41523. arXiv:2304.12950 doi:10.1109/ACCESS.2023.3270916

    2023 arXiv

  12. [19]

    John Preskill. 2018. Quantum computing in the NISQ era and beyond.Quantum 2 (2018), 79

    2018

  13. [20]

    IBM Quantum Platform. [n. d.]. Compute resources. https://quantum.cloud.ibm. com/computers?order=two_q_error_best&direction=asc. [Accessed 17-11-2025]

    2025

  14. [22]

    Peter W Shor. 1994. Algorithms for quantum computation: discrete logarithms and factoring. InProceedings 35th Annual Symposium on Foundations of Computer Science. IEEE, 124–134

    1994

  15. [23]

    Wei Tang, Teague Tomesh, Martin Suchara, Jeffrey Larson, and Margaret Martonosi. 2021. CutQC: Using Small Quantum Computers for Large Quan- tum Circuit Evaluations. InProceedings of the 26th ACM International Conference on Architectural Support for Programming Languages and Op...

    2021

  16. [24]

    Temme, S

    K. Temme, S. Bravyi, and J. M. Gambetta. 2017. Error Mitigation for Short- Depth Quantum Circuits.Physical Review Letters119 (2017), 180509. doi:10.1103/ PhysRevLett.119.180509

    2017

  17. [25]

    Leo Zhou, Sheng-Tao Wang, Soonwon Choi, Hannes Pichler, and Mikhail D Lukin

  18. [26]

    Linghua Zhu, Senwei Liang, Chao Yang, and Xiaosong Li. 2024. Optimizing Shot Assignment in Variational Quantum Eigensolver Measurement.Journal of Chemical Theory and Computation20, 6 (2024), 2638–2648. arXiv:2307.06504 doi:10.1021/acs.jctc.3c01113

    2024 arXiv

  19. [2020]

    Quantum approximate optimization algorithm: Performance, mechanism, and implementation on near-term devices.Physical Review X10, 2 (2020), 021067

    2020

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.