{"id":"1ee45ffb-6d77-4a19-a94c-79519a27bf8b","arxiv_id":"2506.12938","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For odd prime qudits, the sample complexity of direct fidelity estimation is controlled by the Wigner rank and mana of the target state or channel, with constant measurement settings and efficient protocols for stabilizer and low-magic targets.","lead":"This paper builds fidelity estimation protocols for quantum states and channels whose measurement cost is governed by the target's quantum 'magic' (nonstabilizerness), in odd prime dimensional qudit systems. It bridges resource theories of magic and the practical task of certifying noisy quantum devices.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Channel protocols' measurement primitive is underspecified; as printed the pseudocode is unphysical (applying a channel to an operator), and the claimed d^n bias depends on an unstated Choi-state reading.","rationale":"The reader's conditional verdict is appropriate: the state-side protocols (Protocols 1–3) are mathematically sound under the rescaled-outcome convention (O_j|k = λ/d^n with λ an eigenvalue of A_u), and the Wigner-rank/mana sample-complexity bounds follow from the given variance and Hoeffding steps. The main unresolved issue is in the channel protocols, where the reader identified a d^n normalization inconsistency. My analysis shows that the d^n bias is not intrinsic to the mathematics: under the natural Choi-state implementation, W_Λ(v|u) is directly the expectation of A_u^T⊗A_v on the normalized Choi state, so the estimator is unbiased. The real defect is that the paper does not state this implementation and instead gives a pseudocode that is operationally meaningless and even names the wrong channel (U instead of Λ). Thus the reader's specific 'off by d^n' diagnosis is only one possible reading of an underspecified procedure; the deeper problem is that the channel protocols are not executable as printed. This does not change the verdict: the paper should be conditionally accepted with the requirement that the authors specify a valid measurement primitive for channels (e.g., the Choi-state procedure), fix the U/Λ typo, and temper the 'requires' phrasing in the abstract and conclusions. A single analytical check—recomputing E[~Y] under the Choi-state reading—would settle whether any d^n factor remains. The 'requires exponential resources' claim is also unsupported as stated (only upper bounds are proven), but that is a presentation issue secondary to the channel measurement ambiguity.","tokens_in":29917,"tokens_out":20391,"duration_ms":236983,"concrete_test":"Derive E[~Y] for Protocol 4 under two readings and compare with the target F(U,Λ) = (1/d^{2n}) Tr[U†Λ]. Reading (a): literal 'apply Λ to A_u and measure A_v' gives E[~Y] = d^n F(U,Λ). Reading (b): Choi-state measurement of A_u^T⊗A_v on ρ_Λ = J_Λ/d^n gives E[~Y] = F(U,Λ) using Definition 4 and Eq. 18. If (b) is correct, amend Protocol 4 lines 12–14 to specify the Choi-state preparation and replace 'U' with 'Λ'; if (a) is intended, insert a 1/d^n factor in line 16 or redefine ∆. This settles whether the d^n bias is real or merely an implementation ambiguity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorems 18 and 19, the paper's central extension to quantum channels, rest on Protocol 4 (and Protocol 5), whose measurement step is not operationally defined. Protocol 4 line 13 says 'Prepare unitary channel U, apply it on the A_uk and measure the outcome on the A_vk basis.' Applying a channel to an operator A_u is not a physical operation, and the line refers to U instead of the unknown Λ. If the instruction is read literally as preparing some state, applying Λ, and measuring A_v, the measured expectation is Tr[A_v Λ(A_u)] = d^n W_Λ(v|u) by Definition 4, so the estimator in line 16 has expectation d^n W_Λ/W_U and the output estimates d^n F(U,Λ), not F(U,Λ); this is the reader's d^n bias. However, if the measurement is instead implemented by preparing the normalized Choi state ρ_Λ = J_Λ/d^n and measuring the joint observable A_u^T⊗A_v, the expectation is Tr[(A_u^T⊗A_v)ρ_Λ] = W_Λ(v|u) directly (using Definition 4 and Eq. 18), so the estimator is unbiased and ∆ = 1 is correct. The paper never specifies which operational reading is intended, so the channel protocols are not self-contained: a central measurement primitive is missing, and the pseudocode contains a typo. This is load-bearing because the channel extension is a major claimed contribution; without a specified implementable measurement, the claimed sample complexity does not yield a concrete procedure. The abstract's 'requires' phrasing in Section 8 is also an overstatement since only upper bounds are proven, but the channel measurement ambiguity is the more immediate correctness issue.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes direct fidelity estimation (DFE) protocols for pure states and unitary channels in odd-prime-dimensional qudit systems, using the discrete Wigner function and two magic measures: the newly studied Wigner rank and the mana. The state protocols (Protocols 1–3) express the fidelity as an expectation value over randomly sampled phase-space point operators, with sampling probabilities set by the target's Wigner function; the channel protocols (Protocols 4–6) do the same for the entanglement fidelity of a unitary channel versus an unknown channel. The main formal claims are Theorem 15 (state fidelity via Wigner rank, sample complexity O(1/(ε²δ) + 2^{χ_log(ψ)}/ε² ln(1/δ))), Theorem 16 (analogous mana bound), Theorem 18 (channel fidelity via Wigner rank, O(1/(ε²δ) + 2^{χ_log(U)}/ε² ln(1/δ))), Theorem 19 (channel mana bound), and Propositions 17 and 20 for stabilizer states and Clifford channels. The paper also establishes properties of Wigner rank: faithfulness, additivity, subadditivity, and an upper bound on mana.","tokens_in":30138,"tokens_out":11425,"duration_ms":143383,"significance":"If the channel measurement primitive is made precise, the paper would provide a clean operational interpretation of two magic measures—Wigner rank and mana—as sample-complexity quantifiers for fidelity estimation, extending the qubit-based stabilizer-Rényi-entropy results of Leone, Oliviero, and Hamma to odd-prime-dimensional qudits and to unitary channels. The state-side derivations are careful and largely check out: the sampling identity of Eq. (41), the variance bound via Lemma 2, and the Chebyshev/Hoeffding two-stage analysis in Section 4.2 are internally consistent, and the sample-count calculation leading to Eq. (71) is explicit. The paper also credits prior work, especially [Bu24] for the state Wigner rank and [WWS19] for channel mana. The main deficiency is that the channel protocols in Sections 6–7 are not self-contained, and the abstract and concluding section state a hardness claim that goes beyond the proven upper bounds.","major_comments":[{"comment":"The measurement primitive for the channel protocols is not operationally defined. The instruction “Prepare unitary channel U, apply it on the A_{u_k} and measure the outcome on the A_{v_k} basis” cannot be executed as written: a quantum channel acts on states, not on operators, and the unknown channel is Λ, not U. Since Definition 4 defines W_Λ(v|u) = Tr[A_v Λ(A_u)]/d^n, a measurement whose outcomes have mean Tr[A_v Λ(A_u)] would give the Protocol 4 estimator ~X an expectation of d^n W_Λ(v|u)/W_U(v|u), so ~Y would estimate d^n F(U,Λ) rather than F(U,Λ). If the intended implementation is to prepare the normalized Choi state J_Λ/d^n and measure the joint observable A_u^T⊗A_v, then the estimator is unbiased and ∆=1 is correct, but this is never stated. Theorems 18 and 19 rest entirely on this primitive, so the channel protocols require a precise specification of the Choi-state measurement, its normalization, and a correction of the U/Λ typo.","section":"Section 6.1, Protocol 4 line 13; Section 7.1, Protocol 5 line 12"},{"comment":"The claim that fidelity estimation “requires resources that scale exponentially with nonstabilizerness” is stronger than what is proven. Theorems 15, 16, 18, and 19 are upper bounds on the sample complexity of the authors’ specific protocols; no matching lower bound or converse is established for arbitrary protocols. The first paragraph of Section 8 repeats this overstatement (“requires exponential resources”). The text should be rephrased to say that the proposed protocols have sample complexity scaling with the target’s Wigner rank or mana, unless a genuine lower bound is added.","section":"Abstract and Section 8"},{"comment":"The channel Wigner-rank sample-complexity bound inherits a subtlety that should be stated explicitly: χ(U) counts the number of nonzero entries of the channel Wigner matrix, which can be as large as d^{4n}, so 2^{χ_log(U)} is not a practically small quantity for generic unitary channels. The text correctly notes the worst case in the Protocol 4 summary, but the abstract’s framing of a “fundamental trade-off” conflates the protocol-specific upper bound with a statement about the intrinsic hardness of all fidelity-estimation procedures. Please add an explicit sentence distinguishing protocol-dependent sample complexity from fundamental lower bounds.","section":"Section 6.1 and Appendix E, Eq. (172)"}],"minor_comments":[{"comment":"The pseudocode says “Prepare unitary channel U” but the unknown channel is Λ; in all three channel protocols this line should refer to the unknown channel or, in the Choi-state formulation, to applying Λ to half of a maximally entangled state.","section":"Protocol 4, line 13; Protocol 5, line 12; Protocol 6, line 12"},{"comment":"There are unmatched closing parentheses in the definitions of K: for example, K = ⌈8∆ψ/ε²δ)⌉ and K = ⌈8/ε²δ)⌉ should be K = ⌈8∆ψ/ε²δ⌉ and K = ⌈8/ε²δ⌉.","section":"Protocol 2, line 7; Protocol 4, line 7; Protocol 5, line 7"},{"comment":"The proof header refers to “Theorem 20” but the statement being proved is Proposition 20; please correct the cross-reference.","section":"Appendix G"},{"comment":"The text says that O_{j|k} is rescaled by d^n, but the pseudocode does not show the rescaling step explicitly; clarify whether the recorded eigenvalue is that of A_u or of d^n A_u so that the stated range [−1,1] is unambiguous.","section":"Protocol 3, lines 13–15"},{"comment":"There are several typographical errors, including “emperical” in Protocol 3, “speep up” in Section 7.2, and “⌈8∆ψ/ε²δ)⌉” style parentheses in theorem statements; these should be fixed in a revision.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The state-side contribution is largely sound and publishable after revision. The channel extension is the main advertised novelty, and it is currently missing an operational measurement specification; this is fixable but requires more than copy-editing. I would also ask the authors to soften the abstract’s ‘requires’ language unless a lower bound is added. No concerns about citation ethics or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good paper to know about, but not ready as printed. The state-side fidelity estimation bounds are real—I re-derived the Chebyshev/Hoeffding split and it works. The sampling identity and the variance bound are clean, and the Wigner-rank and mana versions give genuine operational meaning to these magic measures in a benchmarking task. The channel-level extension is the new part: the channel Wigner rank with faithfulness, additivity, subadditivity, and the associated protocols are not in the cited literature. The authors also cite prior qubit work honestly and don't oversell the novelty there.\n\nThe soft spot is the channel protocol. Protocol 4 line 13 says 'Prepare unitary channel U, apply it on the A_uk and measure the outcome on the A_vk basis.' That is not an implementable instruction: you cannot apply a channel to an operator and then measure in an operator basis. As written, the estimator's expectation depends on which of two readings you adopt. Literal reading yields a d^n bias and also applies U instead of the unknown Λ; a Choi-state joint measurement yields the correct expectation with Δ = 1. The paper never specifies which. This is a genuine gap in the main new contribution, not a nitpick. It looks fixable without changing the exponential-in-magic structure, but the authors need to give an explicit measurement circuit and settle the normalization.\n\nTwo smaller things: the abstract and conclusions say DFE 'requires' exponential resources in nonstabilizerness, but only upper bounds are proven; and the protocols assume phase-space point measurements that are not yet standard, which the authors acknowledge. Neither is fatal, and the lack of code or experiment is fine for a theory paper.\n\nThis deserves a real referee. The state-side results are solid, the channel extension is worth having, and the gap is localized and repairable. Send it to review with the expectation of a reasonable revision round on the channel sections.","headline":"State-side fidelity bounds are solid; the new channel protocols are promising but need the measurement primitive spelled out before the d^n scaling ambiguity can be resolved.","tokens_in":30862,"tokens_out":7654,"would_cite":true,"duration_ms":82771,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The sample complexity of direct fidelity estimation is governed by the target's nonstabilizerness, measured by Wigner rank or mana, with stabilizer targets costing only $O(\\varepsilon^{-2}\\ln(1/\\delta))$ copies.","keywords":["nonstabilizerness","direct fidelity estimation","Wigner rank","mana","qudit","discrete Wigner function","sample complexity","magic states"],"falsifier":"Numerically simulate the state protocol on a qutrit magic state with known logarithmic Wigner rank, sampling $A_u$ according to $\\Pr(u)=d^n W_\\psi^2(u)$ and using the prescribed $N_k$ measurement rounds; if the empirical number of copies needed to reach accuracy $\\varepsilon$ and confidence $1-\\delta$ systematically exceeds $O(1/(\\varepsilon^2\\delta) + 2^{\\chi_{\\log}}/\\varepsilon^2\\ln(1/\\delta))$, the variance bound would be contradicted. Equivalently, on a real device, measure the qutrit state $|T\\rangle\\langle T|$ and compare the achieved fidelity error at the predicted copy count.","tokens_in":29508,"feed_emoji":"🎯","tokens_out":7135,"duration_ms":75825,"temperature":0.7,"pith_summary":"The paper establishes that the sample complexity of directly estimating the fidelity of a quantum state or unitary channel is governed by the target's nonstabilizerness—its 'magic' as measured by Wigner rank or mana. For odd-prime-dimensional qudit systems, the authors design protocols that sample phase-space point operators according to an importance distribution derived from the target's Wigner function, and they prove that the number of copies needed grows like $O(2^{\\chi_{\\log}(\\psi)}/\\varepsilon^2 \\ln(1/\\delta))$ for a state of logarithmic Wigner rank $\\chi_{\\log}(\\psi)$. Stabilizer states and Clifford channels, which have zero magic, reduce to the copy-optimal $O(1/\\varepsilon^2 \\ln(1/\\delta))$ regime. This gives the abstract measures of nonstabilizerness an operational meaning as the cost of benchmarking, and it shows that states which are hard to simulate classically are also proportionally harder to certify.","feed_headline":"Fidelity estimation cost scales with a state's magic","feed_subtitle":"New protocols show that verifying a quantum state or gate costs copies that grow exponentially with its nonstabilizerness.","key_machinery":"The central object is the discrete Wigner function and its phase-space point operators $\\{A_u\\}$, which form an orthogonal operator basis. The fidelity $\\mathrm{Tr}[\\psi\\rho]$ is rewritten as $\\sum_u \\Pr(u) X_u$ with $\\Pr(u) = d^n W_\\psi^2(u)$, turning estimation into importance sampling; the variance of $X_u$ is bounded by $\\mathrm{Tr}[\\rho^2]\\le 1$, and the number of nonzero Wigner coefficients of the target fixes how many copies of the state are needed. For channels the same identity runs through the channel Wigner function $W_U(v|u)$ and the distribution $\\Pr(v,u) = W_U^2(v|u)/d^{2n}$. The logarithmic Wigner rank $\\chi_{\\log}(\\psi) := \\log\\chi(\\psi) - \\log d(\\psi)$ is the quantity that appears in the exponential sample-complexity term.","core_discovery":"The central claim is a resource-theoretic price tag for direct fidelity estimation (DFE). For a pure target state $\\psi$ in $H_d^{\\otimes n}$, the proposed protocol samples $K = \\lceil 8/\\varepsilon^2\\delta\\rceil$ phase-space point operators $A_u$ with probability proportional to $W_\\psi^2(u)$, and the expected number of copies of the unknown state $\\rho$ is $O(1/(\\varepsilon^2\\delta) + 2^{\\chi_{\\log}(\\psi)}/\\varepsilon^2 \\ln(1/\\delta))$, where $\\chi_{\\log}(\\psi)$ is the logarithmic Wigner rank, the log of the number of nonzero Wigner coefficients minus the log dimension. The same structure holds for unitary channels with logarithmic Wigner rank $\\chi_{\\log}(U)$ and for mana-based versions with $2^{M(\\psi)}$ or $2^{M(U)}$ in place of the rank. For stabilizer states and Clifford channels the magic terms vanish, leaving a constant-copy protocol independent of system size. Along the way the authors prove that Wigner rank is faithful, additive, and an upper bound on mana, and they extend it to channels with analogous structural properties.","pith_inferences":["A direct corollary the paper leaves implicit is that the protocol is efficient for any target whose Wigner rank is small, not only for stabilizer states; this suggests a hierarchy of 'low-rank' states that are cheap to certify despite being non-stabilizer.","The practical bottleneck is the measurement of $A_u$; if phase-space point operator measurements remain noisy or expensive, the constant-copy advantage will not transfer to hardware, and a testable prediction is that the achievable accuracy will be limited by the measurement fidelity rather than the sampling bound.","The channel protocols as printed rely on a normalization in Definition 4; a reader implementing them should verify the $d^n$ factor, since an off-by-$d^n$ shift would change the estimator's expectation value.","A natural extension is to continuous-variable systems, where the same importance-sampling identity might connect quadrature measurement costs to Wigner-function negativity."],"forward_implications":["Stabilizer states and Clifford operations can be fidelity-estimated with a number of copies that does not grow with the number of qudits, $O(1/\\varepsilon^2 \\ln(1/\\delta))$.","For a target with logarithmic Wigner rank $\\chi_{\\log}$, the copy count scales as $2^{\\chi_{\\log}}$, so each additional unit of magic doubles the expected sample cost.","The protocols use only single-qudit phase-space point measurements, and the measurement type is uniform across the procedure.","The mana-based protocols give a second, $\\ell^1$-norm route to the same benchmarking task, with sample complexity $O(2^{M(\\psi)}/\\varepsilon^2\\delta + 2^{2M(\\psi)}/\\varepsilon^2 \\ln(1/\\delta))$.","Because positive-Wigner states and CPWP channels are classically simulable, the results imply that DFE remains tractable exactly for the targets that admit efficient classical simulation."],"supporting_citations":[{"why":"Supplies the direct fidelity estimation framework whose sample complexity this work tightens using nonstabilizerness.","marker":"[FL11]"},{"why":"Establishes the qubit analogue of the result that nonstabilizerness determines DFE hardness, which this paper extends to odd-prime qudits.","marker":"[LOH23]"},{"why":"Provides the discrete Wigner function formalism and the discrete Hudson's theorem that the protocols rely on.","marker":"[Gro06]"},{"why":"Defines mana as a magic measure for states; used as the $\\ell^1$-based cost in Protocols 2 and 5.","marker":"[VMGE14]"},{"why":"Defines the discrete Wigner function and mana of quantum channels, which the channel protocols are built on.","marker":"[WWS19]"},{"why":"First defined the Wigner rank of quantum states, which this paper extends to channels and uses to bound sample complexity.","marker":"[Bu24]"},{"why":"Introduced the Pauli rank for qubits that inspired the Wigner rank definition.","marker":"[BK19]"},{"why":"Shows positive-Wigner operations are classically simulable, motivating the claim that DFE is feasible for classically simulable targets.","marker":"[ME12]"}],"fun_headline_variants":["Verifying quantum states costs more when they're magical","Magic in quantum states drives up verification sample costs","Nonstabilizerness sets the price of fidelity estimation","Sample complexity of verification grows with quantum magic","Magic quantifies the cost of verifying quantum devices"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The protocols presuppose that phase-space point operator expectation values are measurable with the required accuracy and that the target's Wigner function is known classically; if either fails, the sample-complexity savings do not carry over to practice.","fun_headline_variants_meta":{"raw":{"variants":["Verifying quantum states costs more when they're magical","Magic in quantum states drives up verification sample costs","Nonstabilizerness sets the price of fidelity estimation","Sample complexity of verification grows with quantum magic","Magic quantifies the cost of verifying quantum devices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000605,"raw_usage":{"total_tokens":2834,"prompt_tokens":970,"completion_tokens":1864,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":586,"completion_tokens_details":{"reasoning_tokens":1791}},"tokens_in":586,"tokens_out":1864,"duration_ms":15227,"temperature":1.0,"reasoning_tokens":1791,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:45:40.157586+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically simulate the state protocol on a qutrit magic state with known logarithmic Wigner rank, sampling $A_u$ according to $\\Pr(u)=d^n W_\\psi^2(u)$ and using the prescribed $N_k$ measurement rounds; if the empirical number of copies needed to reach accuracy $\\varepsilon$ and confidence $1-\\delta$ systematically exceeds $O(1/(\\varepsilon^2\\delta) + 2^{\\chi_{\\log}}/\\varepsilon^2\\ln(1/\\delta))$, the variance bound would be contradicted. Equivalently, on a real device, measure the qutrit state $|T\\rangle\\langle T|$ and compare the achieved fidelity error at the predicted copy count.","supporting_citations":[],"review_version":1}