Unitary Channel Testing Under a Depolarizing Noise Assumption
Pith reviewed 2026-06-27 13:01 UTC · model grok-4.3
The pith
Under the depolarizing noise assumption, algorithms test if a black-box channel is a target unitary or ε-far in diamond distance with Θ(1/ε) queries.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Our optimal algorithms answer the following question: is the quantum channel implemented by a given black box identical to a target unitary or ε-far from it in the diamond distance, assuming that the deviation is a depolarizing channel with unknown parameter? Our algorithm has a query complexity of Θ(1/ε). The query complexity of the relaxed problem of testing whether the black-box channel is ε1-close to a target unitary or ε2-far in the diamond distance is Θ(ε2/(ε2 - ε1)^2). In both cases, we provide matching lower bounds that hold even for adaptive, ancilla-assisted protocols with multi-outcome incoherent measurements.
What carries the argument
Depolarizing channel model for deviations from target unitaries, which structures the diamond-distance testing problem to admit linear query complexity in 1/ε.
If this is right
- Optimal algorithms exist with Θ(1/ε) queries for exact vs ε-far under the depolarizing assumption.
- Relaxed testing has complexity Θ(ε2/(ε2-ε1)^2).
- Lower bounds match the upper bounds even for adaptive ancilla-assisted multi-outcome measurements.
- The results and optimality hold specifically when deviations are exactly depolarizing.
Where Pith is reading between the lines
- If real devices exhibit primarily depolarizing errors, these bounds could reduce the number of experiments needed to certify gate fidelity.
- The same modeling strategy might be applied to derive tight bounds for other structured noise families in channel testing.
- The gap between the exact and relaxed versions shows how relaxing the closeness threshold changes the scaling of required queries.
Load-bearing premise
Any deviation from the target unitary takes the exact form of a depolarizing channel whose parameter is unknown but fixed.
What would settle it
Implement a channel whose deviation is a non-depolarizing noise model and measure whether the number of queries needed to certify ε-far deviation matches the stated Θ(1/ε) bound or exceeds it.
read the original abstract
We present fast algorithms $\unicode{x2013}$ under the depolarizing noise assumption, often made in fault-tolerant quantum computations $\unicode{x2013}$ to test its strength. Our optimal algorithms answer the following question: is the quantum channel implemented by a given black box identical to a target unitary or $\varepsilon$-far from it in the diamond distance, assuming that the deviation is a depolarizing channel with unknown parameter? Our algorithm has a query complexity of $\Theta(1/\varepsilon).$ The query complexity of the relaxed problem of testing whether the black-box channel is $\varepsilon_1$-close to a target unitary or $\varepsilon_2$-far in the diamond distance is $\Theta\bigl(\varepsilon_2/(\varepsilon_2 - \varepsilon_1)^2\bigr).$ In both cases, we provide matching lower bounds that hold even for adaptive, ancilla-assisted protocols with multi-outcome incoherent measurements.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript develops query-efficient algorithms for distinguishing whether a black-box quantum channel equals a target unitary or is ε-far from it in diamond distance, under the modeling assumption that any deviation is exactly a depolarizing channel with unknown but fixed parameter. It claims an optimal query complexity of Θ(1/ε) for the exact testing task and Θ(ε₂/(ε₂ - ε₁)²) for the relaxed (ε₁, ε₂) version, together with matching lower bounds that apply even to adaptive, ancilla-assisted protocols using multi-outcome incoherent measurements.
Significance. If the claimed matching upper and lower bounds are correct, the work supplies tight characterizations of sample complexity for this restricted but practically motivated channel-testing problem. The depolarizing-noise restriction permits substantially better query bounds than the general case, and the lower bounds are strong because they survive powerful adaptive and ancilla-assisted protocols. The results are therefore potentially useful in fault-tolerant quantum computing settings where depolarizing noise is a standard modeling choice.
minor comments (2)
- The abstract states that the algorithms answer the testing question 'assuming that the deviation is a depolarizing channel with unknown parameter,' but does not indicate where in the manuscript the precise mathematical formulation of this assumption (e.g., the precise form of the channel family and the diamond-distance metric) is introduced; a forward reference would improve readability.
- The phrase 'to test its strength' in the first sentence of the abstract is slightly ambiguous; clarifying whether 'its' refers to the black-box channel or to the depolarizing-noise model would remove a minor source of confusion.
Simulated Author's Rebuttal
We thank the referee for their positive assessment of our work and for recommending minor revision. The referee's summary accurately captures the main contributions regarding optimal query complexities under the depolarizing noise assumption.
Circularity Check
No significant circularity
full rationale
The paper derives query complexity bounds (Θ(1/ε) and Θ(ε₂/(ε₂-ε₁)²)) as mathematical results under an explicit modeling assumption that deviations are exactly depolarizing channels with fixed unknown parameter. These are presented as derived from the problem setup and diamond-distance definitions, with matching lower bounds for adaptive protocols. No equations reduce a claimed prediction to a fitted input by construction, no load-bearing self-citations appear, and the central claims remain independent of the target result itself. The derivation is self-contained within the stated restricted channel class.
Axiom & Free-Parameter Ledger
axioms (2)
- standard math Diamond distance is the appropriate metric for distinguishing quantum channels.
- domain assumption The noise model is exactly depolarizing with a single unknown parameter.
Reference graph
Works this paper leans on
-
[1]
Learning general channels Θ(d4/ε2) [5]
-
[2]
Learning unitary channels Θ(d2/ε) [3]
-
[3]
Testing general channels Ω( √ d/ε) [18]
-
[4]
Testing unitary channels Θ(d/ε) [19]
-
[5]
Testing unitary channels under coherent noise assumption Θ( √ d/ε) [21]
-
[6]
individual
Testing unitary channels under depolarizing noise assumption Θ(1/ε) This Letter T ABLE II.Complexities of various learning and testing tasks, using an access model permitting ancilla, coher- ence, and adaptivity. All tasks quantifying proximity are in the diamond distance. Our algorithms are fast as their time complexities are independent of the number of...
-
[7]
1b): Each trial ap- plies the black box once to the input state, followed by a measurement
Parallel black-box access (Fig. 1b): Each trial ap- plies the black box once to the input state, followed by a measurement. This corresponds tom= 1, and henceN=L. We obtain standard-testing and tol- erant testing upper bounds using Algorithms 1 and 3 respectively
-
[8]
Sequential Black-Box access (Fig. 1c): Each trial preparesρand applies the black boxmtimes se- quentially, yielding the state N ◦m λ (ρ) =N λ ◦ Nλ ◦ · · · ◦ Nλ ◦ Nλ| {z } mtimes (ρ), and then measures using the binary POVM{ρ, I− ρ}. Repeating this procedure forLindependent trials givesN=mLtotal queries. We obtain a standard-testing upper bound in this acc...
-
[9]
Surawy-Stepney, J
T. Surawy-Stepney, J. Kahn, R. Kueng, and M. Guta, Quantum6, 844 (2022)
2022
-
[10]
Oufkir, in2023 IEEE International Symposium on Information Theory (ISIT)(2023) pp
A. Oufkir, in2023 IEEE International Symposium on Information Theory (ISIT)(2023) pp. 1919–1924
2023
-
[11]
J. Haah, R. Kothari, R. O’Donnell, and E. Tang, in2023 IEEE 64th Annual Symposium on Foundations of Com- puter Science (FOCS)(IEEE, 2023) pp. 363–390
2023
-
[12]
S. Chen, C. Oh, S. Zhou, H.-Y. Huang, and L. Jiang, Physical Review Letters132, 180805 (2024)
2024
-
[13]
A. A. Mele and L. Bittel, Optimal learning of quantum channels in diamond distance (2026), arXiv:2512.10214 [quant-ph]
arXiv 2026
-
[14]
Aharonov and M
D. Aharonov and M. Ben-Or, SIAM Journal on Comput- ing38, 1207 (2008)
2008
-
[15]
Malhotra, A
P. Malhotra, A. Kumar, and S. Garhwal, International Journal of Theoretical Physics63, 278 (2024)
2024
-
[16]
Proctor, K
T. Proctor, K. Young, A. D. Baczewski, and R. Blume- Kohout, Nature Reviews Physics7, 105 (2025)
2025
-
[17]
T. Rohe, F. H. Ruiloba, S. Egger, S. von Beck, J. Stein, and C. Linnhoff-Popien, Quantum computer benchmark- ing: An explorative systematic literature review (2025), arXiv:2509.03078 [quant-ph]
arXiv 2025
-
[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, Nature Communications10, 5347 (2019)
2019
-
[19]
Harper, S
R. Harper, S. T. Flammia, and J. J. Wallman, Nature Physics16, 1184 (2020). 6
2020
-
[20]
S. Chen, Y. Liu, M. Otten, A. Seif, B. Fefferman, and L. Jiang, Nature Communications14, 52 (2023)
2023
-
[21]
S. T. Flammia and J. J. Wallman, ACM Transactions on Quantum Computing1, 3:1 (2020)
2020
-
[22]
Fawzi, A
O. Fawzi, A. Oufkir, and D. Stilck Fran¸ ca, IEEE Trans- actions on Information Theory71, 2642 (2025)
2025
-
[23]
Y. R. Sanders, J. J. Wallman, and B. C. Sanders, New Journal of Physics18, 012002 (2015)
2015
-
[24]
Montanaro and R
A. Montanaro and R. d. Wolf,A Survey of Quantum Property Testing, Graduate Surveys No. 7 (Theory of Computing Library, 2016) pp. 1–81
2016
-
[25]
Fawzi, N
O. Fawzi, N. Flammarion, A. Garivier, and A. Oufkir, inProceedings of Thirty Sixth Conference on Learning Theory, Proceedings of Machine Learning Research, Vol. 195, edited by G. Neu and L. Rosasco (PMLR, 2023) pp. 1822–1884
2023
-
[26]
G. Rosenthal, H. Aaronson, S. Subramanian, A. Datta, and T. Gur, Quantum channel testing in average-case distance (2024), arXiv:2409.12566 [quant-ph]
arXiv 2024
-
[27]
K. Chen, Q. Wang, and Z. Zhang, Strict hierarchy for quantum channel certification to unitary (2026), arXiv:2604.26900 [quant-ph]
Pith/arXiv arXiv 2026
-
[28]
Chen and W
S. Chen and W. Gong, PRX Quantum6, 020323 (2025)
2025
-
[29]
Jeon and C
S. Jeon and C. Oh, npj Quantum Information12, 2 (2025)
2025
-
[30]
Raussendorf, J
R. Raussendorf, J. Harrington, and K. Goyal, New Jour- nal of Physics9, 199 (2007)
2007
-
[31]
Takada and K
Y. Takada and K. Fujii, PRX Quantum5, 030352 (2024)
2024
-
[32]
J. Old, S. Tasler, M. J. Hartmann, and M. M¨ uller, Phys. Rev. Lett.135, 240601 (2025)
2025
-
[33]
Sasaki, M
M. Sasaki, M. Ban, and S. M. Barnett, Physical Review A66, 022308 (2002)
2002
-
[34]
Z. Ji, G. Wang, R. Duan, Y. Feng, and M. Ying, IEEE Transactions on Information Theory54, 5172 (2008)
2008
-
[35]
S. Khatri and M. M. Wilde, Principles of quantum communication theory: A modern approach (2024), arXiv:2011.04672 [quant-ph]
arXiv 2024
-
[36]
F. Scholz, Confidence bounds and intervals for param- eters relating to the binomial, negative binomial, pois- son and hypergeometric distributions with applications to rare events (2008)
2008
-
[37]
A. V. Gerbessiotis, A survey of chernoff and hoeffding bounds (2025), arXiv:2506.15612 [cs.DM]
arXiv 2025
-
[38]
Wu, Lecture notes on information-theoretic methods for high-dimensional statistics (2020)
Y. Wu, Lecture notes on information-theoretic methods for high-dimensional statistics (2020). 7 Appendix A: Lemmas
2020
-
[39]
ThenN λ(U ρU†) =UN λ(ρ)U †,and therefore NU ◦ Nλ =N λ ◦ NU
Depolarizing Channel Lemma 1.1(Unitary Covariance of the Depolarizing Channel).LetN λ(ρ) = (1−λ)ρ+λI/dandN U(ρ) = U ρU†. ThenN λ(U ρU†) =UN λ(ρ)U †,and therefore NU ◦ Nλ =N λ ◦ NU. Proof.UsingUIU † =I, Nλ(U ρU†) = (1−λ)U ρU † +λ I d (A1) =U (1−λ)ρ+λ I d U † (A2) =UN λ(ρ)U † (A3) Lemma 1.2(Query Complexity of Depolarizing Param- eter Estimation).LetH ∼= Cd...
-
[40]
Prepare the stateρand apply the channelN λ
-
[41]
Perform the binary POVM measurement{ρ,I−ρ} to obtain an outcomeX i ∈ {0,1}. Letˆp= 1 N PN i=1 Xi be the empirical mean and define the estimator: ˆλ:= d(1−ˆp) d−1 .(A5) Then, for any precisionε∈(0,1)and failure probability δ∈(0,1/2), we have: Pr ∥Nˆλ(ρ)− N λ(ρ)∥1 ≥ε (A6) = Pr |ˆλ−λ| ≥ ε 2(1−1/d) ≤δ.(A7) whenever the number of samples satisfies: N≥ 2 ε2 ln ...
-
[42]
(A12) and requiring the probability of error be at mostδ: 2 exp −2N ε 2 2 ≤δ⇐ ⇒N≥ 2 ε2 ln 2 δ .(A13)
Substitutingt= ε 2 into Eqn. (A12) and requiring the probability of error be at mostδ: 2 exp −2N ε 2 2 ≤δ⇐ ⇒N≥ 2 ε2 ln 2 δ .(A13)
-
[43]
Distance Lemmas The following distance lemmas will be used in the up- per and lower bound calculations in the following sec- tions. Lemma 2.1(Trace distance between depolarizing chan- nel and the identity channel).The distance betweenN λ and the identity channel,id, in trace distance, is: dTr(Nλ,id) := max ρ∈Cd×d ∥(Nλ −id)(ρ)∥ 1 = 2λ(1− 1 d). (A14) Furthe...
-
[44]
Letn∈Nandk∈ {0,
Other Important Lemmas These lemmas will be used in the tolerant tester upper bound. Letn∈Nandk∈ {0, . . . , n}. Forp∈[0,1], let S(p) ∼Binomial(n, p). Lemma 3.1(Monotonicity of Binomial Lower Tail [28]). The cumulative distribution functionPr(S (p) ≤k)is nonincreasing inp. Specifically, for any0≤p 1 ≤p 2 ≤1, Pr(S(p2) ≤k)≤Pr(S (p1) ≤k).(A54) Lemma 3.2(Mono...
-
[45]
Proof.From Algorithm 1, the probability of error under H0 is 0, since,∀k∈[L], Pr(X k = 1|H0) = 0
Standard T ester : When sequential black box applications are not allowed Theorem B.1.For the problem inP2.1, there exists an ancilla-free, non-adaptive algorithm using individual two-outcome measurements that achieves an upper bound ofO(1/ε)queries. Proof.From Algorithm 1, the probability of error under H0 is 0, since,∀k∈[L], Pr(X k = 1|H0) = 0. UnderH 1...
-
[46]
1c) satis- fying: L= Ω (ln(1/δ))andmL > 2 ε ln 1 δ
Standard T ester: When sequential black box applications are allowed Theorem B.2.There exists an ancilla-free, non- adaptive tester solving ProblemP2.1withmsequential black-box applications andLinput states (Fig. 1c) satis- fying: L= Ω (ln(1/δ))andmL > 2 ε ln 1 δ . Proof.From Algorithm 2, the probability of error under H0 is 0, since: Pr(Xk = 1|H 0) = 0∀k...
-
[47]
T olerant T ester Theorem B.3.For the tolerant testing problem inP2.2, we give an ancilla-free, non-adaptive algorithm using in- dividual measurements that achieves an upper bound of O ε2/(ε2 −ε 1)2 queries. Proof.Letpbe the probability of obtaining an outcome 1, at each step, when the depolarizing parameter isλ, i.e., p= 1− ⟨0|N λ(|0⟩⟨0|)|0⟩=λ(1−1/d).(B2...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.