REVIEW 3 major objections 5 minor 30 references
Enhancing quantum noise characterization via extra energy levels
T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Qutrit-enhanced protocols shrink the gauge ambiguity in SPAM and gate Pauli noise characterization, with experiments on superconducting devices.
desk verdict A clean theory of qudit SPAM gauge ambiguities, with a plausible experimental demonstration; the main risk is the noiseless-control assumption, which the authors flag but do not quantify. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the subsystem depolarizing gauge: for any subset Ω of the n qudits, the map D^Ω_p(·)=(1-p)(·)+p Tr_Ω(·)⊗ I_Ω/$d^{{|Ω|}}$ commutes with every parallel single-qudit unitary, so transforming ρ0 with D^Ω_p and each measurement operator with its inverse leaves all outcome distributions unchanged. The paper linearizes these transformations to first order and shows the 2^n - 1 non-empty subsets give linearly independent gauge directions; it then constructs a protocol using generalized Pauli-X gates (permutation gates X_π) that determines every parameter orthogonal to all gauge directions, with no more than 2d^n circuits. The positivity constraints—every ε^S_j and ε^M_{l,k} is a probability—turn the gauge directions into a bounded interval, and it is the size of that interval that qutrit levels shrink.
What would settle it
Run the qutrit-enhanced SPAM protocol on a device while deliberately tuning the fidelity of the |1>-|2> gate; if the estimated gauge ambiguity interval A does not grow as that gate fidelity drops, or if the inferred ε^S_j and ε^M_{l,k} drift, the noiseless-control premise is falsified. A simpler check: compare qutrit-enhanced estimates of qubit-subspace SPAM with and without randomized compiling on the qutrit gates; they should agree if the gate noise is gate-independent.
Extended reading notes
Core claim
For an n-qudit system whose SPAM noise is incoherent and whose single-qudit unitary control is perfect, the paper proves (Proposition 1) that the noise model has exactly 2^n - 1 gauge degrees of freedom, generated by subsystem depolarizing maps D^Ω_p; no experiment within the gate set can distinguish the transformation ρ0 → D^Ω_p(ρ0), E_l → (D^Ω_p)^{-1}(E_l). Up to first order, each gauge parameter is confined by positivity: the initialization and measurement error probabilities must stay non-negative. For a single qudit the resulting ambiguity is A = min_{j≠0} ε^S_j + min_{k≠l} ε^M_{l,k}; since the population of the |2> level of a superconducting transmon is much smaller than that of |1>, adding qutrit information tightens A. The paper experimentally demonstrates this on one- and two-qubit/qutrit superconducting devices, showing that qutrit-enhanced SPAM characterization reduces the gauge ambiguity compared to qubit-only protocols both with and without heralding. It further feeds the tighter SPAM bounds into intercept cycle benchmarking of a two-qubit CZ gate, producing smaller intervals for non-identifiable Pauli fidelities such as λ_XI.
Load-bearing premise
Everything rests on the assumption that single-qudit control is noiseless (or that its noise is gate-independent and absorbable into the noise model), so if the extra-level gates themselves have appreciable noise, the claimed gauge reduction may not survive.
Editorial extensions
If this is right
- The gauge ambiguity for a single qudit shrinks from roughly ε^S_1 + min{ε^M_{1,0}, ε^M_{0,1}} to the qutrit-constrained minimum over j≠0 and k≠l, which is much smaller on thermal transmon devices.
- Tighter SPAM bounds directly tighten the estimate of Pauli fidelities λ_a for gates whose Clifford action changes the support of P_a, where λ_a is not individually identifiable.
- All learnable SPAM parameters of an n-qudit system can be learned with at most 2d^n circuits, within a factor of two of the information-theoretic lower bound of d^n + 1.
- Qutrit-enhanced characterization works alongside heralding; the two methods combine, since heralding reduces ε^S_1 and qutrit constraints further restrict the gauge.
- The same framework applies to correlated multi-qubit SPAM noise: for two qutrits the three gauge DOFs are the two single-qudit depolarizing gauges plus the joint {1,2} depolarizing gauge.
Reading between the lines
- If single-qutrit gate noise is gate-independent, it can likely be absorbed into the gauge framework just as single-qubit gate noise is absorbed in Pauli noise learning; testing this would require interleaving qutrit gate calibrations into the protocol.
- The positivity-constraint mechanism might apply to leakage detection: population in |2> is itself a leakage signature, so the same extra-level data could jointly bound leakage and SPAM gauge in transmons.
- A natural stress test is to run the protocol on a device with controllable |1>-|2> gate error; the predicted gauge interval should widen monotonically as that gate's error increases.
- The gauge DOF count being independent of d suggests that using qutrits rather than qubits is always at least as informative even if higher levels are slightly noisier, because the added levels give more positivity constraints without adding new gauge directions.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes using higher energy levels (qudits) to reduce gauge ambiguities in qubit-subspace SPAM and Pauli-gate noise characterization. The authors develop a first-order gauge theory for n-qudit SPAM noise under the assumption of noiseless single-qudit control, proving that the gauge freedoms are exactly the 2^n−1 subsystem depolarizing transformations and giving an explicit protocol with at most 2d^n circuits that determines all learnable parameters. They then use positivity constraints to bound the residual gauge ambiguity, arguing that qutrit levels with smaller populations tighten the bound. Experimental results on one- and two-superconducting-qutrit devices compare qubit-only and qutrit-enhanced protocols, showing smaller ambiguity intervals for the latter, and a gate-noise application bounds non-identifiable Pauli fidelities for CZ more tightly using qutrit-enhanced SPAM data.
Significance. If the central claim holds, this is a practical and elegant contribution: it extends the self-consistent Pauli-noise-learning framework to qudit systems, gives a proof of the gauge-DOF count, and demonstrates a concrete use of higher levels already present in superconducting devices. The positivity-based ambiguity bound (Eq. (9)) and the explicit circuit constructions in the SM are useful. The paper also provides an experimental demonstration, which strengthens the case. The main caveat is that the entire gauge analysis assumes noiseless single-qudit gates; the authors acknowledge this but do not quantify how the benefit degrades with qutrit gate error. The proof of Proposition 1 is explicit and the protocol is concrete, which is a strength, as is the inclusion of experimental data rather than only numerics.
major comments (3)
- [SM Eq. (S6)–(S7); 'Single-qudit SPAM characterization'] The noiseless single-qudit control assumption is load-bearing for the central claim. The gauge transformation in Eq. (S6) preserves the gate set only because Λ_p commutes with every single-qudit unitary (S7). When the generalized X gates used in the protocol are noisy, conjugation by the gate no longer maps the noise model to itself, and the first-order equations (S15) acquire additional contributions from qutrit gate errors. The Discussion explicitly states that single-qudit gates 'might be more noisy due to their complexity' and lists leakage as a possible source of the experimental discrepancy. However, the paper does not report qutrit gate fidelities for the device used in Figs. 1–3 and does not provide a sensitivity analysis showing how the reported reduction in gauge ambiguity depends on single-qutrit gate error. Since the claimed enhancement is the difference between qubit-only and qutrit-enhanced intervals, a per-gate qutrit error comparable to that difference could erase the benefit. This needs to be quantified, e.g., by reporting gate fidelities and by numerical simulation of the protocol with finite single-qutrit gate error.
- [Figs. 1–3 and Eq. (9)] The experimental demonstration does not provide statistical uncertainty on the ambiguity reduction itself. The light bars are described as one standard error for deciding the region, but the ambiguity width is a nonlinear function of the estimated error probabilities, involving minima (Eq. (9)); the plug-in estimate of a minimum is biased, and the reported reduction could be within noise. The authors should provide confidence intervals or bootstrap/tolerance intervals for the ambiguity intervals, or a statistical test comparing the qubit-only and qutrit-enhanced widths.
- [Eqs. (5), (8), (9); Figs. 3–4] The quantitative claims are first-order in ε, but the manuscript does not bound the size of the neglected O(ε^2) terms for the experimentally measured error rates (up to about 3.5%). Figure 4's blue shaded region shows a visible spread attributed to the first-order approximation, indicating that these corrections are not always negligible relative to the reported ambiguity differences. The authors should provide an estimate or upper bound on second-order contributions for the parameters in Figs. 1–3, or validate the first-order bounds against a higher-order or numerical calculation for those parameters.
minor comments (5)
- [SM text before Eq. (S8)] The sentence 'We can all the independent noise parameters in a vector v' contains a typo and should read 'We can call' or 'We can collect'.
- [SM paragraph after Eq. (S20)] The phrase 'the experiment of X_{π*(k)(0)}' appears to be a typo; the probability expression in Eq. (S20) uses the gate X_{π*(k)}, not X_{π*(k)(0)}. Please clarify.
- [SM Eq. (S19)] The notation '(0,h_i)' in the definition of π*(k) is not defined in that context; please specify the transposition or cycle notation used for the permutation.
- [Fig. 4 caption] The caption says the blue shaded region is 'due to the first-order approximation'; since the max/min curves also differ because of the remaining gauge choices, the caption should distinguish the two sources of spread.
- [Experimental sections] The manuscript does not report standard experimental details such as the number of shots, single-qutrit gate decompositions, device calibration parameters, or the exact procedure used to assign the light error bars in Figs. 1–3; providing these details in the Supplemental Materials would improve reproducibility.
Circularity Check
No significant circularity: the qutrit gauge theory and ambiguity reduction are derived in-paper from measured level populations and positivity bounds; self-citations are background support, not load-bearing.
full rationale
The derivation chain is self-contained. The gauge DOF count is proven in the main text and SM: Eqs. (S6)-(S12) exhibit 2^n-1 independent subspace depolarizing gauges, and the constructive protocol of Eqs. (S13)-(S20) plus the completeness argument fixes all remaining parameters, so no more gauges exist. The qutrit ambiguity reduction is not a fitted prediction: Eq. (9) defines the ambiguity A as the sum of the smallest off-diagonal state-preparation element and the smallest off-diagonal measurement element; the experimental demonstration evaluates this bound using independently estimated populations and confusion rates, and the smaller qutrit interval follows from the physical fact (thermal populations, readout classifier thresholds) that |2> rates are below |1> rates. The gate-noise application re-derives the exponential decay model in Eq. (13) and uses Ref. [3]'s identifiability criterion only to classify which Pauli eigenvalues are gauge-dependent; that cited theorem is a parameter-free prior result whose assumptions do not include qutrits, so the self-citation is ordinary background support rather than a smuggled conclusion. The Discussion's caveat about noisy single-qutrit control and leakage is an acknowledged robustness limitation, not a circular step.
Assumptions & free parameters
assumptions (5)
- domain assumption Noiseless single-qudit control
- domain assumption Incoherent (diagonal) SPAM noise, enforceable by random phase twirling
- domain assumption SPAM noise is small (first-order approximation)
- standard math Positivity of probabilities
- standard math Depolarizing map properties
Cite this review
Pith. "Pith review of Enhancing quantum noise characterization via extra energy levels." pith.science (2026). https://pith.science/paper/TGS5CN6Z
@misc{pith2026250609131,
author = {Pith},
title = {Pith review of: Enhancing quantum noise characterization via extra energy levels},
year = {2026},
howpublished = {\url{https://pith.science/paper/TGS5CN6Z}},
note = {Machine review of arXiv:2506.09131}
}
read the original abstract
Noise is a major challenge for building practical quantum computing systems. Precise characterization of quantum noise is crucial for developing effective error mitigation and correction schemes. However, state preparation and measurement (SPAM) errors on many current platforms can introduce large ambiguity into conventional noise characterization methods. In this work, we propose a scheme for enhancing quantum noise characterization using additional energy levels. We first develop a comprehensive theory on the identifiability of n-qudit SPAM noise given high-quality single-qudit control, showing the existence of gauge freedoms which can be completely described using subsystem depolarizing maps. We then show how to use these extra energy levels to reduce the gauge ambiguity in characterizing both SPAM and gate noise in the qubit subspace. We experimentally implement these ideas on a superconducting quantum computing device and demonstrate a qutrit-enabled enhancement in noise characterization precision.
Figures
Figures from the paper (1 more)
Reference graph
Works this paper leans on
-
[1]
M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information: 10th Anniversary Edition, 10th ed. (Cambridge University Press, USA, 2011)
2011
-
[2]
Nielsen, J
E. Nielsen, J. K. Gamble, K. Rudinger, T. Scholten, K. Young, and R. Blume-Kohout, Gate Set Tomography, Quantum5, 557 (2021)
2021
-
[3]
S. Chen, Y. Liu, M. Otten, A. Seif, B. Fefferman, and L. Jiang, The learnability of pauli noise, Nature Commu- nications14, 52 (2023)
work page 2023
-
[4]
E. H. Chen, S. Chen, L. E. Fischer, A. Eddins, L. C. Govia, B. Mitchell, A. He, Y. Kim, L. Jiang, and A. Seif, Disambiguating pauli noise in quantum computers, arXiv preprint arXiv:2505.22629 (2025)
arXiv 2025
-
[5]
R. Laflamme, J. Lin, and T. Mor, Algorithmic cooling for resolving state preparation and measurement errors in quantum computing, Phys. Rev. A106, 012439 (2022)
work page 2022
-
[6]
J. Lin, J. J. Wallman, I. Hincks, and R. Laflamme, In- dependent state and measurement characterization for quantum computers, Phys. Rev. Res.3, 033285 (2021)
work page 2021
-
[7]
A.Hashim, R.K.Naik, A.Morvan, J.-L.Ville, B.Mitchell, J. M. Kreikebaum, M. Davis, E. Smith, C. Iancu, K. P. O’Brien, I. Hincks, J. J. Wallman, J. Emerson, and I. Sid- diqi, Randomized compiling for scalable quantum com- puting on a noisy superconducting quantum processor, Phys. Rev. X11, 041039 (2021)
work page 2021
-
[8]
E. Van Den Berg, Z. K. Minev, A. Kandala, and K. Temme, Probabilistic error cancellation with sparse pauli–lindblad models on noisy quantum processors, Na- ture Physics (2023)
work page 2023
Show all 30 references
-
[9]
Ferracin, A
S. Ferracin, A. Hashim, J.-L. Ville, R. Naik, A. Carignan- Dugas, H. Qassim, A. Morvan, D. I. Santiago, I. Siddiqi, and J. J. Wallman, Efficiently improving the performance of noisy quantum computers, Quantum8, 1410 (2024)
2024
-
[10]
S. Chen, Z. Zhang, L. Jiang, and S. T. Flammia, Effi- cient self-consistent learning of gate set pauli noise, arXiv preprint arXiv:2410.03906 (2024)
2024
-
[11]
Morvan, V
A. Morvan, V. V. Ramasesh, M. S. Blok, J. M. Kreike- baum, K. O’Brien, L. Chen, B. K. Mitchell, R. K. Naik, D. I. Santiago, and I. Siddiqi, Qutrit randomized bench- marking, Physical review letters126, 210504 (2021)
2021
-
[12]
M. A. Yurtalan, J. Shi, M. Kononenko, A. Lupascu, and S. Ashhab, Implementation of a walsh-hadamard gate in a superconducting qutrit, Physical review letters125, 180504 (2020)
2020
-
[13]
See Supplemental Materials
-
[14]
Krantz, M
P. Krantz, M. Kjaergaard, F. Yan, T. P. Orlando, S. Gus- tavsson, and W. D. Oliver, A quantum engineer’s guide to superconducting qubits, Applied physics reviews6(2019)
2019
-
[15]
Bravyi, S
S. Bravyi, S. Sheldon, A. Kandala, D. C. Mckay, and J. M. Gambetta, Mitigating measurement errors in multiqubit experiments, Phys. Rev. A103, 042605 (2021)
2021
-
[16]
S. T. Flammia and J. J. Wallman, Efficient estimation of Pauli channels, ACM Transactions on Quantum Comput- ing1, 10.1145/3408039 (2020)
2020 doi
-
[17]
J. J. Wallman and J. Emerson, Noise tailoring for scalable quantum computation via randomized compiling, Physical Review A94, 052325 (2016)
2016
-
[18]
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, Characterizing large-scale quantum computers via cycle benchmarking, Nature Communications10, 5347 (2019)
2019
-
[19]
S. Chen, S. Zhou, A. Seif, and L. Jiang, Quantum ad- vantages for pauli channel estimation, Physical Review A 105, 032435 (2022)
2022
-
[20]
Rosenblatt, H
Y.Kim, A.Eddins, S.Anand, K.X.Wei, E.VanDenBerg, S. Rosenblatt, H. Nayfeh, Y. Wu, M. Zaletel, K. Temme, et al., Evidence for the utility of quantum computing before fault tolerance, Nature618, 500 (2023)
2023
-
[21]
P. Iyer, A. Jain, S. D. Bartlett, and J. Emerson, Efficient diagnostics for quantum error correction, Phys. Rev. Res. 4, 043218 (2022)
2022
-
[22]
E. H. Chen, T. J. Yoder, Y. Kim, N. Sundaresan, S. Srini- vasan, M. Li, A. D. Córcoles, A. W. Cross, and M. Takita, Calibrated decoders for experimental quantum error cor- rection, Phys. Rev. Lett.128, 110504 (2022)
2022
-
[23]
Hashim, S
A. Hashim, S. Seritan, T. Proctor, K. Rudinger, N. Goss, R. K. Naik, J. M. Kreikebaum, D. I. Santiago, and I. Sid- diqi, Benchmarking quantum logic operations relative to thresholds for fault tolerance, npj Quantum Information 9, 109 (2023)
2023
-
[24]
Huang, S
H.-Y. Huang, S. T. Flammia, and J. Preskill, Founda- tions for learning from noisy quantum experiments (2022), arXiv:2204.13691 [quant-ph]
2022 arXiv
-
[25]
van den Berg and P
E. van den Berg and P. Wocjan, Techniques for learning sparse pauli-lindblad noise models, Quantum8, 1556 (2024)
2024
-
[26]
Carignan-Dugas, D
A. Carignan-Dugas, D. Dahlen, I. Hincks, E. Ospadov, S. J. Beale, S. Ferracin, J. Skanes-Norman, J. Emerson, and J. J. Wallman, The error reconstruction and compiled calibration of quantum computing cycles, arXiv preprint arXiv:2303.17714 (2023)
2023 arXiv
-
[27]
Wagner, H
T. Wagner, H. Kampermann, D. Bruß, and M. Kliesch, Learning logical pauli noise in quantum error correction, Physical review letters130, 200601 (2023)
2023
-
[28]
statement
C. J. Haupt, A. C. Vazquez, L. E. Fischer, S. Woerner, and D. J. Egger, State-preparation and measurement error mitigation with non-computational states (2025), in preparation. 7 Supplemental Materials: Enhancing quantum noise characterization via extra energy levels THEORY OF...
2025
-
[29]
learnable
What and how many degrees of freedom are identifiable, (or, “learnable”)?
-
[30]
Furthermore, there exists a protocol using no more than 2dn circuits to estimate all the learnable degrees of freedom of the noise
How many experiments are needed to extract all learnable information? Our answers are summarized in the following, Proposition 1.For an n-qudit system, given perfect single-qudit control, there are2n− 1gauge degrees of freedom for incoherent SPAM noise, up to a first-order app...
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.