Non-Clifford Benchmarking via Ensemble Feature Selection

Nikolay V. Vitanov; Stancho G. Stanchev

arxiv: 2607.01180 · v1 · pith:YAI77CYZ · submitted 2026-07-01 · quant-ph

Non-Clifford Benchmarking via Ensemble Feature Selection

Stancho G. Stanchev , Nikolay V. Vitanov This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-02 11:34 UTCgrok-4.3pith:YAI77CYZ record.json open to challenge →

classification quant-ph

keywords Ensemble Feature Selectionprocess infidelitynon-Clifford gatesCCZ gaterandomized benchmarkingquantum computingridge regressionIBM quantum device

0 comments

The pith

Ensemble Feature Selection trains a linear estimator on simulated noisy channels to measure process infidelity of non-Clifford gates such as CCZ.

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

The paper introduces Ensemble Feature Selection (EFS) to estimate the process infidelity of involutory multi-qubit gates that are not Clifford gates, such as the CCZ gate. Standard benchmarking methods rely on Clifford gates, so this approach trains a linear model offline on a physically motivated set of noisy channels to pick a small set of measurable circuits and assign weights via ridge regression. The method is tested on an IBM quantum device using related Clifford benchmarks against Interleaved Randomized Benchmarking, showing agreement within about 0.01 precision for infidelities between 0.02 and 0.2. It is then used to measure the CCZ gate directly on the same hardware.

Core claim

The EFS method selects a compact set of circuit measurements from a candidate pool through offline training on an ensemble of noisy channels and combines them into a linear estimator with weights learned by ridge regression, enabling fast estimation of process infidelity for non-Clifford targets.

What carries the argument

Ensemble Feature Selection (EFS), which uses ridge regression on a training ensemble of noisy channels to learn a linear combination of measurement outcomes for estimating gate infidelity.

If this is right

EFS achieves close agreement with IRB on validation benchmarks across a range of infidelities.
The estimator provides precision of approximately 0.01 over process infidelity 0.02-0.2.
EFS can be applied directly to estimate infidelity of the CCZ gate on the device.
The training ensemble can incorporate prior knowledge about hardware noise mechanisms.

Where Pith is reading between the lines

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

If the training ensemble is tuned well, EFS could reduce the number of experiments needed compared to full randomized benchmarking for non-Clifford gates.
This approach might extend to other non-Clifford gates beyond CCZ if similar structurally related validation benchmarks exist.
Validation on multiple devices would test how well the method generalizes beyond the single IBM device tested.

Load-bearing premise

The training ensemble of noisy channels must be representative enough of the real device's dominant noise so that the learned estimator generalizes accurately to actual measurements.

What would settle it

A significant discrepancy between EFS estimates and independent IRB results on the Clifford validation benchmarks would indicate the method does not work as claimed.

Figures

Figures reproduced from arXiv: 2607.01180 by Nikolay V. Vitanov, Stancho G. Stanchev.

**Figure 3.** Figure 3: FIG. 3: EFS estimates vs true process infidelity for CCZ [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: shows the combined EFS–IRB comparison for ICCZ and CZ(0, 2) on ibm_kingston with M = 24. The fitted slopes and intercepts for the two targets are α = 0.94 and α = 0.93, both with intercept 0.012, and the combined data give Pearson correlation r = 0.957 and RMSEcal = 0.0099 over ε ≈ 0.01–0.20. The combined EFS–IRB agreement across both benchmarks provides the primary validation result, with an estimation … view at source ↗

**Figure 5.** Figure 5: FIG. 5: EFS estimates for CCZ versus [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗

**Figure 6.** Figure 6: shows the transpiled implementation of CZ(0, 2). Owing to its lower process infidelity, this circuit extends the study into a lower-infidelity regime while retaining a genuinely three-qubit error structure through the active participation of the routing qubit. Both ICCZ and CZ(0, 2) are employed as Clifford validation benchmarks for assessing the performance of the EFS estimator on the non-Clifford CCZ t… view at source ↗

**Figure 7.** Figure 7: FIG. 7: Transpiled CCZ circuit on a three-qubit topol [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗

**Figure 8.** Figure 8: FIG. 8: EFS vs IRB for CZ [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗

**Figure 9.** Figure 9: FIG. 9: EFS( [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗

read the original abstract

We propose an Ensemble Feature Selection (EFS) method for fast estimation of process infidelity of involutory multi-qubit gates, including non-Clifford targets, for which standard Clifford-based benchmarking does not apply. The method selects a compact set of experimentally executable circuit measurements from a candidate pool through offline training on a physically motivated ensemble of noisy channels, and combines them into a linear estimator with weights learned by ridge regression. The training ensemble is an explicit and tunable component of the protocol, incorporating prior knowledge about dominant hardware noise mechanisms. The estimator is validated on ibm_kingston using two Clifford validation benchmarks structurally related to the transpiled CCZ circuit, against independent Interleaved Randomized Benchmarking (IRB). Both show close EFS-IRB agreement across a wide range of process infidelities, with an estimation precision of approximately 0.01 over a process infidelity range of 0.02-0.2. EFS is subsequently applied directly to CCZ on the same device.

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

EFS trains a linear estimator on simulated noisy channels to pick measurements for non-Clifford infidelity, shows IRB agreement on related Clifford tests, but leaves the key generalization step unverified.

read the letter

The main takeaway is that this paper gives a concrete protocol for estimating process infidelity on non-Clifford gates by training ridge regression on an explicit ensemble of noisy channels, selecting a compact measurement set, and applying the resulting linear estimator to hardware data.

It does a few things cleanly. The training ensemble is treated as a tunable input that can encode known hardware noise, which is a practical move. They run the method on ibm_kingston, get close agreement with independent IRB on two Clifford benchmarks that share structure with the transpiled CCZ, and report roughly 0.01 precision over the 0.02-0.2 infidelity range. That supplies at least some empirical anchor.

The soft spot is exactly where the stress-test note flags it. Validation stays inside Clifford circuits. The actual target is CCZ, and no independent ground-truth measurement is shown for it. The claim that the estimator carries over therefore rests on the untested assumption that the chosen ensemble captures the device's noise behavior for the non-Clifford case. If that assumption is off, the numbers on CCZ are hard to interpret.

This is aimed at experimental groups that need faster infidelity estimates for universal gates and are willing to invest in offline simulation to build the estimator. A reader already working on benchmarking or gate optimization would find the method description and the hardware numbers worth looking at.

It deserves a serious referee. The problem is real, the approach is explicit, and the reported validation is at least a start, even if more direct checks on the non-Clifford target would be needed before the numbers can be taken at face value.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes an Ensemble Feature Selection (EFS) protocol for estimating process infidelity of involutory multi-qubit gates, including non-Clifford targets. A compact set of executable circuit measurements is selected from a candidate pool by offline ridge-regression training on an explicit ensemble of noisy channels that incorporates prior knowledge of dominant hardware noise; the resulting linear estimator is validated on ibm_kingston against independent IRB using two Clifford benchmarks structurally related to a transpiled CCZ, reported to agree within ~0.01 precision over infidelity range 0.02–0.2, and then applied directly to the CCZ gate on the same device.

Significance. If the learned estimator generalizes from the training ensemble to real-device non-Clifford gates, the method would supply a practical, low-shot alternative to full tomography or Clifford-restricted benchmarking for involutory non-Clifford operations. The explicit, tunable character of the training ensemble is a methodological strength that allows incorporation of device-specific noise priors.

major comments (2)

[Abstract] Abstract: validation is performed exclusively on two Clifford circuits that remain Clifford after transpilation; the central claim that the EFS estimator applies to the non-Clifford CCZ therefore rests on the untested assumption that the chosen ensemble is representative of the device noise for the target gate. No independent ground-truth comparison (IRB or otherwise) is reported for the CCZ itself.
[Abstract] Abstract: the reported estimation precision of approximately 0.01 is stated without accompanying details on ensemble construction, ridge-regularization parameter selection, checks for overfitting, or error propagation; these omissions are load-bearing for evaluating whether the quoted precision is robust or an artifact of the training procedure.

minor comments (1)

[Abstract] The abstract refers to “involutory multi-qubit gates” without specifying the precise algebraic condition (e.g., U² = I) used to define the class; a short clarifying sentence would improve scope precision.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful review and constructive feedback. We address the two major comments point by point below. We agree that the abstract requires additional methodological details and will revise it. Regarding validation of the non-Clifford target, we clarify the rationale for the chosen benchmarks while acknowledging the indirect nature of the evidence.

read point-by-point responses

Referee: [Abstract] Abstract: validation is performed exclusively on two Clifford circuits that remain Clifford after transpilation; the central claim that the EFS estimator applies to the non-Clifford CCZ therefore rests on the untested assumption that the chosen ensemble is representative of the device noise for the target gate. No independent ground-truth comparison (IRB or otherwise) is reported for the CCZ itself.

Authors: We agree that no direct IRB comparison is possible for the CCZ, as IRB requires Clifford gates by construction; this is precisely the motivation for developing EFS. The two validation benchmarks were chosen because they are structurally related to the transpiled CCZ (same qubit layout and similar gate decomposition) and remain Clifford after transpilation, allowing independent IRB. The training ensemble is explicitly constructed from an ensemble of noisy channels that incorporates measured device noise priors (e.g., dominant depolarizing and coherent error channels on ibm_kingston). We will add a new paragraph in the revised manuscript (Section 5) explicitly stating the assumption of ensemble representativeness, its justification from the device-specific training, and the limitation that direct ground-truth validation for non-Clifford gates is unavailable with current methods. revision: partial
Referee: [Abstract] Abstract: the reported estimation precision of approximately 0.01 is stated without accompanying details on ensemble construction, ridge-regularization parameter selection, checks for overfitting, or error propagation; these omissions are load-bearing for evaluating whether the quoted precision is robust or an artifact of the training procedure.

Authors: We accept this criticism. The abstract currently omits these details. In the full manuscript, ensemble construction is described in Section 3 (candidate pool of 200 circuits, reduced to 12 via feature selection), the ridge parameter λ is chosen by 5-fold cross-validation on the training ensemble (Section 4.2), overfitting is checked via hold-out validation on a separate noise realization set with reported R^{2} > 0.95, and precision is obtained via bootstrap resampling of the linear estimator (1000 resamples) yielding a standard error of ~0.01. We will revise the abstract to include a concise sentence summarizing these elements and will ensure the methods section makes the full procedure reproducible. revision: yes

Circularity Check

0 steps flagged

No circularity: estimator trained offline on independent simulated ensemble

full rationale

The EFS protocol trains the linear estimator via ridge regression on an explicit simulated ensemble of noisy channels chosen to reflect hardware noise, then applies the fixed weights to real-device measurement outcomes on the target gate. Validation against IRB occurs on separate Clifford circuits using independent hardware data. No equation or step reduces the reported infidelity for CCZ to a quantity fitted from the CCZ measurements themselves; the training data and target data are disjoint by construction. This matches the default expectation of a self-contained protocol with no load-bearing self-definition or fitted-input-as-prediction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that a linear estimator trained via ridge regression on a physically motivated ensemble of noisy channels can accurately predict process infidelity from a small set of real-device measurements; the ensemble itself is treated as an input that encodes prior noise knowledge.

axioms (2)

standard math Ridge regression produces stable weights for a linear combination of measurement outcomes that estimate process infidelity
The method explicitly uses ridge regression to learn the estimator weights from the training ensemble.
domain assumption An ensemble of simulated noisy channels can be constructed to capture the dominant error mechanisms present on the target hardware
The abstract states that the training ensemble incorporates prior knowledge about dominant hardware noise mechanisms and is a tunable component of the protocol.

pith-pipeline@v0.9.1-grok · 5700 in / 1638 out tokens · 34421 ms · 2026-07-02T11:34:40.587222+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages · 1 internal anchor

[1]

Magesan, J

E. Magesan, J. M. Gambetta, and J. Emerson, Phys. Rev. Lett.106, 180504 (2011)

work page 2011
[2]

Magesan, J

E. Magesan, J. M. Gambetta, B. R. Johnson, C. A. Ryan, J. M. Chow, S. T. Merkel, M. P. Da Silva, G. A. Keefe, 9 M. B. Rothwell, T. A. Ohki, M. B. Ketchen, and M. Stef- fen, Phys. Rev. Lett.109, 080505 (2012)

work page 2012
[3]

A. W. Cross, E. Magesan, L. S. Bishop, J. A. Smolin, and J. M. Gambetta, npj Quantum Inf2, 16012 (2016)

work page 2016
[4]

Gu, W.-F

Y. Gu, W.-F. Zhuang, X. Chai, and D.-L. Deng, Nat. Commun.14, 5880 (2023)

work page 2023
[5]

H. Ye, G. Liu, and X. Ma, arXiv:2510.15554 (2025)

work page arXiv 2025
[6]

Genois, J

´E. Genois, J. A. Gross, A. Di Paolo, N. J. Stevenson, G. Koolstra, A. Hashim, I. Siddiqi, and A. Blais, PRX Quantum2, 040355 (2021)

work page 2021
[7]

Cimini, M

V. Cimini, M. Valeri, E. Polino, S. Piacentini, F. Cecca- relli, G. Corrielli, N. Spagnolo, R. Osellame, and F. Scia- rrino, Advanced Quantum Technologies7, 2300474 (2024)

work page 2024
[8]

Papiˇ c, A

M. Papi´ c, A. Auer, and I. de Vega, arXiv preprint arXiv:2305.08916 10.48550/arXiv.2305.08916 (2023)

work page doi:10.48550/arxiv.2305.08916 2023
[9]

Proctor, A

T. Proctor, A. Tran, X. Liu, A. Dhumuntarao, S. Seritan, A. Green, and N. M. Linke, arXiv:2504.12575 (2025), arXiv:2504.12575

work page arXiv 2025
[10]

Guyon and A

I. Guyon and A. Elisseeff, J. Mach. Learn. Res.3, 1157 (2003)

work page 2003
[11]

Bol’on-Canedo and A

V. Bol’on-Canedo and A. Alonso-Betanzos, Inf. Fusion 52, 1 (2019)

work page 2019
[12]

A. D. Hill, M. J. Hodson, N. Didier, and M. J. Reagor, arXiv:2108.01652 10.48550/arXiv.2108.01652 (2021), arXiv:2108.01652 [quant-ph]

work page doi:10.48550/arxiv.2108.01652 2021
[13]

Phys.18, 783 (2022)

Y.Kim, A.Morvan, L.B.Nguyen, R.K.Naik, C.J¨ unger, L.Chen, J.M.Kreikebaum, D.I.Santiago,andI.Siddiqi, Nat. Phys.18, 783 (2022)

work page 2022
[14]

Hashim, L

A. Hashim, L. B. Nguyen, N. Goss, B. Marinelli, R. K. Naik, T. Chistolini, J. Hines, J. P. Marceaux, Y. Kim, P. Gokhale, T. Tomesh, S. Chen, L. Jiang, S. Ferracin, K. Rudinger, T. Proctor, K. C. Young, R. Blume-Kohout, and I. Siddiqi, arXiv:2408.12064 10.48550/arXiv.2408.12064 (2024), arXiv:2408.12064 [quant-ph]

work page doi:10.48550/arxiv.2408.12064 2024
[15]

Erhard, J

A. Erhard, J. J. Wallman, L. Postler, M. Meth, R. Stricker, E. A. Martinez, P. Schindler, T. Monz, J. Emerson, and R. Blatt, Nat Commun10, 5347 (2019)

work page 2019
[16]

J. J. Wallman and J. Emerson, Phys. Rev. A94, 052325 (2016)

work page 2016
[17]

J. J. Wallman and J. Emerson, System and methods for local randomized benchmarking, US Patent Application 2019/0026211 A1 (2019), filed Jul. 23, 2018; published Jan. 24, 2019

work page 2019
[18]

S. K. Grigorova, N. V. Vitanov, and B. T. Torosov, arXiv 10.48550/arXiv.2606.30327 (2026), 2606.30327 [quant- ph]

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2606.30327 2026
[19]

M. A. Nielsen, Physics Letters A303, 249 (2002)

work page 2002
[20]

D. C. McKay, C. J. Wood, S. Sheldon, J. M. Chow, and J. M. Gambetta, Phys. Rev. A96, 022330 (2017)

work page 2017
[21]

IBM Quantum, Ibm quantum platform (2026), accessed: 2026-06-06

work page 2026
[22]

Krantz, M

P. Krantz, M. Kjaergaard, F. Yan, T. P. Orlando, S. Gus- tavsson, and W. D. Oliver, Appl. Phys. Rev.6, 021318 (2019)

work page 2019
[23]

Krinner, S

S. Krinner, S. Lazar, A. Remm, C. K. Andersen, N. Lacroix, G. J. Norris, C. Hellings, M. Gabureac, C. Eichler, and A. Wallraff, Phys. Rev. Applied14, 024042 (2020)

work page 2020
[24]

A. N. Tikhonov and V. Y. Arsenin,Solutions of Ill-Posed Problems(Winston, 1977)

work page 1977
[25]

A. E. Hoerl and R. W. Kennard, Technometrics12, 55 (1970)

work page 1970
[26]

Kohavi and G

R. Kohavi and G. H. John, Artificial Intelligence97, 273 (1997)

work page 1997
[27]

T. Q. A. D. Team, Qiskit aer,https://github.com/ Qiskit/qiskit-aer(2025)

work page 2025
[28]

Kohavi, inProceedings of the Fourteenth International Joint Conference on Artificial Intelligence (IJCAI)(Mor- gan Kaufmann, 1995) pp

R. Kohavi, inProceedings of the Fourteenth International Joint Conference on Artificial Intelligence (IJCAI)(Mor- gan Kaufmann, 1995) pp. 1137–1143. 10 Appendix A: Gate Circuits This appendix shows the particular transpiled bench- mark circuits used in the experimental validation of the EFS estimator. These circuits are not unique; many al- ternative real...

work page 1995

[1] [1]

Magesan, J

E. Magesan, J. M. Gambetta, and J. Emerson, Phys. Rev. Lett.106, 180504 (2011)

work page 2011

[2] [2]

Magesan, J

E. Magesan, J. M. Gambetta, B. R. Johnson, C. A. Ryan, J. M. Chow, S. T. Merkel, M. P. Da Silva, G. A. Keefe, 9 M. B. Rothwell, T. A. Ohki, M. B. Ketchen, and M. Stef- fen, Phys. Rev. Lett.109, 080505 (2012)

work page 2012

[3] [3]

A. W. Cross, E. Magesan, L. S. Bishop, J. A. Smolin, and J. M. Gambetta, npj Quantum Inf2, 16012 (2016)

work page 2016

[4] [4]

Gu, W.-F

Y. Gu, W.-F. Zhuang, X. Chai, and D.-L. Deng, Nat. Commun.14, 5880 (2023)

work page 2023

[5] [5]

H. Ye, G. Liu, and X. Ma, arXiv:2510.15554 (2025)

work page arXiv 2025

[6] [6]

Genois, J

´E. Genois, J. A. Gross, A. Di Paolo, N. J. Stevenson, G. Koolstra, A. Hashim, I. Siddiqi, and A. Blais, PRX Quantum2, 040355 (2021)

work page 2021

[7] [7]

Cimini, M

V. Cimini, M. Valeri, E. Polino, S. Piacentini, F. Cecca- relli, G. Corrielli, N. Spagnolo, R. Osellame, and F. Scia- rrino, Advanced Quantum Technologies7, 2300474 (2024)

work page 2024

[8] [8]

Papiˇ c, A

M. Papi´ c, A. Auer, and I. de Vega, arXiv preprint arXiv:2305.08916 10.48550/arXiv.2305.08916 (2023)

work page doi:10.48550/arxiv.2305.08916 2023

[9] [9]

Proctor, A

T. Proctor, A. Tran, X. Liu, A. Dhumuntarao, S. Seritan, A. Green, and N. M. Linke, arXiv:2504.12575 (2025), arXiv:2504.12575

work page arXiv 2025

[10] [10]

Guyon and A

I. Guyon and A. Elisseeff, J. Mach. Learn. Res.3, 1157 (2003)

work page 2003

[11] [11]

Bol’on-Canedo and A

V. Bol’on-Canedo and A. Alonso-Betanzos, Inf. Fusion 52, 1 (2019)

work page 2019

[12] [12]

A. D. Hill, M. J. Hodson, N. Didier, and M. J. Reagor, arXiv:2108.01652 10.48550/arXiv.2108.01652 (2021), arXiv:2108.01652 [quant-ph]

work page doi:10.48550/arxiv.2108.01652 2021

[13] [13]

Phys.18, 783 (2022)

Y.Kim, A.Morvan, L.B.Nguyen, R.K.Naik, C.J¨ unger, L.Chen, J.M.Kreikebaum, D.I.Santiago,andI.Siddiqi, Nat. Phys.18, 783 (2022)

work page 2022

[14] [14]

Hashim, L

A. Hashim, L. B. Nguyen, N. Goss, B. Marinelli, R. K. Naik, T. Chistolini, J. Hines, J. P. Marceaux, Y. Kim, P. Gokhale, T. Tomesh, S. Chen, L. Jiang, S. Ferracin, K. Rudinger, T. Proctor, K. C. Young, R. Blume-Kohout, and I. Siddiqi, arXiv:2408.12064 10.48550/arXiv.2408.12064 (2024), arXiv:2408.12064 [quant-ph]

work page doi:10.48550/arxiv.2408.12064 2024

[15] [15]

Erhard, J

A. Erhard, J. J. Wallman, L. Postler, M. Meth, R. Stricker, E. A. Martinez, P. Schindler, T. Monz, J. Emerson, and R. Blatt, Nat Commun10, 5347 (2019)

work page 2019

[16] [16]

J. J. Wallman and J. Emerson, Phys. Rev. A94, 052325 (2016)

work page 2016

[17] [17]

J. J. Wallman and J. Emerson, System and methods for local randomized benchmarking, US Patent Application 2019/0026211 A1 (2019), filed Jul. 23, 2018; published Jan. 24, 2019

work page 2019

[18] [18]

S. K. Grigorova, N. V. Vitanov, and B. T. Torosov, arXiv 10.48550/arXiv.2606.30327 (2026), 2606.30327 [quant- ph]

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2606.30327 2026

[19] [19]

M. A. Nielsen, Physics Letters A303, 249 (2002)

work page 2002

[20] [20]

D. C. McKay, C. J. Wood, S. Sheldon, J. M. Chow, and J. M. Gambetta, Phys. Rev. A96, 022330 (2017)

work page 2017

[21] [21]

IBM Quantum, Ibm quantum platform (2026), accessed: 2026-06-06

work page 2026

[22] [22]

Krantz, M

P. Krantz, M. Kjaergaard, F. Yan, T. P. Orlando, S. Gus- tavsson, and W. D. Oliver, Appl. Phys. Rev.6, 021318 (2019)

work page 2019

[23] [23]

Krinner, S

S. Krinner, S. Lazar, A. Remm, C. K. Andersen, N. Lacroix, G. J. Norris, C. Hellings, M. Gabureac, C. Eichler, and A. Wallraff, Phys. Rev. Applied14, 024042 (2020)

work page 2020

[24] [24]

A. N. Tikhonov and V. Y. Arsenin,Solutions of Ill-Posed Problems(Winston, 1977)

work page 1977

[25] [25]

A. E. Hoerl and R. W. Kennard, Technometrics12, 55 (1970)

work page 1970

[26] [26]

Kohavi and G

R. Kohavi and G. H. John, Artificial Intelligence97, 273 (1997)

work page 1997

[27] [27]

T. Q. A. D. Team, Qiskit aer,https://github.com/ Qiskit/qiskit-aer(2025)

work page 2025

[28] [28]

Kohavi, inProceedings of the Fourteenth International Joint Conference on Artificial Intelligence (IJCAI)(Mor- gan Kaufmann, 1995) pp

R. Kohavi, inProceedings of the Fourteenth International Joint Conference on Artificial Intelligence (IJCAI)(Mor- gan Kaufmann, 1995) pp. 1137–1143. 10 Appendix A: Gate Circuits This appendix shows the particular transpiled bench- mark circuits used in the experimental validation of the EFS estimator. These circuits are not unique; many al- ternative real...

work page 1995