{"id":"54955e58-6eb5-4a7c-90c3-8bd3ca9385d9","arxiv_id":"1908.01614","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A measurement-efficient method certifies a lower bound on the classical capacity of an unknown quantum channel using at most d^2 local measurement settings and classical optimization.","lead":"This paper introduces a measurement-efficient way to certify how many classical bits an unknown quantum channel can carry per use. It uses a few local measurements and a classical optimization to get a guaranteed lower bound on capacity, avoiding full process tomography.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-sample bias can push the estimated C_DET above the true capacity, so the practical 'accessible bound' needs error bars; the ideal-probability theorem in Eq. (5) is sound.","rationale":"The paper's central mathematical result is correct: I verified Eq. (3) by direct trace computation, and Eq. (5) follows because every I^(i) is the mutual information of an explicit prepare-measure scheme, hence bounded by the Holevo capacity of the same ensemble and therefore by C1(E). The Blahut-Arimoto recursion is the standard global maximizer for a DMC, so the numerical step is not a concern. The examples are internally consistent. The one weak point is the practical claim of experimental accessibility: because C_DET is a maximum and DMC capacity is convex in the transition matrix, finite-count fluctuations make the plug-in estimate positively biased; a direct simulation on a depolarizing channel would confirm this. This is exactly the reader's weakest assumption, so I agree with it. The bias is a standard experimental-statistics issue and does not undermine the ideal-probability theorem or the examples, so the reader's ACCEPT verdict should stand. A short error-analysis section or an explicit caveat would be sufficient to close the gap.","tokens_in":12159,"tokens_out":30136,"duration_ms":326857,"concrete_test":"Simulate the protocol for a completely depolarizing qubit channel (true C=C1=0). For a chosen set of Pauli bases and N = 10^4 copies per basis, draw multinomial counts from the ideal joint probabilities p(m,n)=1/d*p(m|n), estimate p-hat^(i)(m|n), run the Blahut-Arimoto maximization (Eq. (6)) for each i, and record C-hat_DET = max_i I^(i). Repeat 10^4 times. If the median C-hat_DET is positive, the uncorrected estimate is not a valid lower bound. Then construct a one-sided 95% bootstrap confidence bound from the same data and check that it covers the true value 0 in at least 95% of trials; this verifies the minimal error analysis needed to restore the practical claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Under exact probabilities the chain C(E) >= C1(E) >= max_i I^(i) is sound: Eq. (3) reconstructs p^(i)(m|n), and each I^(i) is bounded by C1 via the Holevo theorem. The load-bearing gap is experimental certification. C_DET is a maximum over i and over the input prior, and the capacity of a classical DMC is a convex function of its transition matrix. If p-hat^(i)(m|n) is an unbiased finite-sample estimate of p^(i)(m|n), Jensen's inequality gives E[max_i I^(i)(p-hat)] >= max_i I^(i)(p), so the naive plug-in estimate is biased upward. For a zero-capacity channel (e.g., completely depolarizing), any noise in the counts produces a positive C-hat_DET with high probability, so the uncorrected protocol is not a guaranteed lower bound. The paper does not provide a confidence-interval construction, one-sided tests, or sample-complexity bounds. This does not invalidate Eq. (5) as a mathematical statement, but it means the claim that the method 'detects' a certified lower bound in practice is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a method to obtain a lower bound on the classical capacity of an unknown quantum channel without full process tomography. The scheme prepares a maximally entangled state, applies the channel to the system half, measures local observables of the form X_i ⊗ X_i^T, and reconstructs conditional probabilities p^(i)(m|n) using the identity in Eq. (3). From these probabilities one computes I^(i), the capacity of the induced classical binary/dit channel, and defines C_DET as the maximum over i. The chain C ≥ C1 ≥ C_DET in Eq. (5) is the central result. The paper then derives explicit formulas for Pauli channels, generalized amplitude-damping channels, stretched-damping channels, extremal qubit channels, a qutrit V-shaped decay channel, dephasing on an unknown basis, and Pauli channels followed by an unknown phase rotation, using both analytic and numerical (Blahut–Arimoto) optimization.","tokens_in":12379,"tokens_out":17882,"duration_ms":172891,"significance":"The mathematical core is a clean and correct inequality: for exact conditional probabilities, the mutual information achievable with any fixed orthonormal input ensemble and output measurement is a lower bound on the Holevo capacity, and hence on the classical capacity. The identity in Eq. (3) provides a practical way to extract these conditional probabilities with local measurements. If the paper's protocol is implemented with rigorous statistical treatment, it would offer a valuable experimental tool for certifying a guaranteed lower bound on classical capacity for completely unknown channels, complementing the authors' earlier QDET bound for quantum capacity. The examples are well chosen and demonstrate that the bound is often tight, sometimes exactly equal to the true capacity.","major_comments":[{"comment":"The derivation of Eq. (5) is sound for exact conditional probabilities p^(i)(m|n), but the paper's claim that the method provides an 'experimentally accessible bound' is not justified under finite statistics. In practice the probabilities are estimated from a finite number of trials, and the estimated C_DET is a maximum over i and over the input prior. Because the capacity of a classical discrete memoryless channel is a convex function of its transition matrix, Jensen's inequality gives E[C(p-hat)] ≥ C(p), so the plug-in estimate is biased upward. For a zero-capacity channel, any statistical noise produces a positive estimated bound with high probability. The paper does not provide a confidence-interval construction, a one-sided test, or a sample-complexity bound. Without such a correction, the protocol does not certify a lower bound in a real experiment. I recommend adding a paragraph (or a short section) that either gives a finite-size confidence-interval procedure or explicitly states that the bound applies to ideal probabilities and that a rigorous statistical extension is needed.","section":"Detection strategy, Eq. (5)"}],"minor_comments":[{"comment":"The definition of z in Eq. (A.3) is ambiguous in the typeset text: it should read z = 2^{(H[ϵ0]−H[ϵ1])/(1−ϵ0−ϵ1)}, with the parentheses in the exponent made explicit.","section":"Supplemental Material, Eq. (A.3)"},{"comment":"The statement that C = C1 = CDET for Pauli channels rests on additivity of the Holevo capacity for unital qubit channels, but the cited reference [46] is not the standard theorem for qubit additivity; the authors should cite King's theorem on additivity for all qubit channels (or for unital qubit channels specifically) to make the justification precise.","section":"Eq. (12) and surrounding text"},{"comment":"The claim 'We numerically checked that the maximum is always achieved by the first term in Eq. (17)' is a strong statement over the full parameter range; a brief analytic argument or a plot of the difference would make the example more convincing.","section":"Generalized amplitude-damping example, Eq. (17)"},{"comment":"Similarly, the numerical check that the first term in Eq. (A.8) is always the maximizer is reported without supporting data; a short proof or figure would strengthen the example.","section":"Extremal qubit channels, Eq. (A.8)"},{"comment":"The phrase 'number of local measurements that scales at most as d^2' should be clarified: each measurement setting X_i ⊗ X_i^T yields d^2 joint probabilities, so d^2 settings produce d^4 numbers, which is already enough for full process tomography. The advantage of the method is that a smaller number of settings still yields a nontrivial bound; this distinction should be stated explicitly to avoid a misleading scalability claim.","section":"General remarks"}],"recommendation":"major_revision","confidential_remarks":"The paper is theoretically sound and the main inequality is correct; the examples are convincing. The single load-bearing issue is the absence of a finite-size statistical analysis, which is central to the claim of an 'accessible' experimental bound. The fix is local and does not affect the mathematical results, so I see this as a major-revision rather than a rejection. The authors' citation of their own prior work is appropriate, though the dependence on [24] and [26] is heavy; the novelty here is the classical-capacity bound, which is distinct from the quantum-capacity bound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is a neat, honest extension of the authors' QDET detection method to classical capacity. The core chain C(E) ≥ C1(E) ≥ C_DET = max_i I^(i) is correct: reconstructing the conditional probabilities via the maximally entangled state identity (3) and then computing mutual information for each basis gives a genuine lower bound, since each I^(i) corresponds to a specific encoding-decoding scheme. The math is standard and the examples check out. For unital qubit channels, C_DET equals the Holevo capacity because they are pseudoclassical; for nonunital channels the paper gives analytic formulas and honestly notes when the bound is loose. The qutrit V-shaped channel example is a good demonstration that the method can produce nontrivial bounds for channels where the capacity isn't well known.\n\nWhat's genuinely new is the adaptation to classical capacity with d^2 measurement scaling, plus the explicit reconstruction recipe and the new analytic expressions. The citation pattern is appropriate, including the self-citations, since this builds directly on their QDET work. No circularity.\n\nThe soft spot is finite-sample statistics. C_DET is a maximum over bases and priors, and mutual information is convex in the transition matrix. If you plug in an unbiased estimate of p^(i)(m|n) from finite counts, Jensen's inequality gives an upward bias in the estimated C_DET; for a completely depolarizing channel, noise alone will produce a positive number with high probability. The paper never mentions this. The ideal-probability theorem is fine, but the claim that the method 'detects' a certified lower bound in practice is not established without confidence intervals or sample-complexity bounds. This doesn't sink the paper, but it should be stated as a limitation or addressed in a follow-up. I'd also note that the bound tests only orthogonal ensembles, so there are channels where it's far below C1; the paper acknowledges this.\n\nOverall, this is a solid Letter for experimentalists benchmarking unknown channels and for theorists interested in detection methods. It deserves a serious referee; I'd ask for a finite-size error analysis or an explicit caveat in the main text. Worth a quick read at the next group meeting.","headline":"Sound and useful: a d^2-measurement lower bound for classical capacity, but the experimental 'accessible' claim needs a finite-sample error analysis.","tokens_in":12890,"tokens_out":2960,"would_cite":true,"duration_ms":30997,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Hk"],"model":"deepseek-v4-flash","headline":"A few local measurements certify a lower bound on any quantum channel's classical capacity","keywords":["classical capacity","quantum channel","lower bound","local measurements","Holevo capacity","process tomography","Blahut-Arimoto algorithm","qubit channels"],"falsifier":"Run the protocol on a channel whose Holevo capacity is known analytically but whose optimal ensemble is nonorthogonal (e.g., a qubit channel violating pseudoclassicality), using finite measurement statistics, and check whether the estimated $C_{\\rm DET}$ ever exceeds the known $C_1$; exceeding it would invalidate the bound as an experimental guarantee. Alternatively, add a small depolarizing noise to the reference system and observe whether the reconstructed probabilities break the assumptions of the trace identity.","tokens_in":11968,"feed_emoji":"📡","tokens_out":6926,"duration_ms":63511,"temperature":0.7,"pith_summary":"This paper presents a method to certify a lower bound on the classical capacity of a noisy quantum communication channel without performing full process tomography. The method prepares a system qudit and a noiseless reference qudit in a maximally entangled state, measures a small set of local observables of the form $X_i \\otimes X_i^T$, and reconstructs the conditional probabilities of the channel acting on the eigenstates of each $X_i$. A classical optimization over input priors, using the Blahut–Arimoto recursion, yields a mutual information $I^{(i)}$, and the detected bound is $C_{\\rm DET} = \\max_i I^{(i)}$, which provably satisfies $C(\\mathcal{E}) \\ge C_1(\\mathcal{E}) \\ge C_{\\rm DET}$. The number of measurement settings scales at most as $d^2$ for a $d$-dimensional system, requires no a priori knowledge of the channel, and the protocol returns an explicit encoding that attains the detected bound. For all unital qubit channels and for nonunital qubit channels satisfying the pseudoclassicality condition, this detected bound is exactly the Holevo capacity.","feed_headline":"Few measurements bound the classical capacity of unknown channels","feed_subtitle":"The bound uses at most d^2 measurement settings, with no prior noise knowledge.","key_machinery":"The load-bearing identity is $\\mathrm{Tr}[(A \\otimes B^T)(\\mathcal{E} \\otimes I_R)|\\varphi^+\\rangle\\langle\\varphi^+|] = \\frac{1}{d}\\mathrm{Tr}[A\\,\\mathcal{E}(B)]$, which converts local measurements of $X_i \\otimes X_i^T$ on the entangled state into the channel's transition probabilities. The classical optimization of the mutual information $I^{(i)}$ over the input prior is performed with the Blahut–Arimoto recursion, guaranteeing convergence to the optimal prior for each measured basis. The detection bound is then $C_{\\rm DET} = \\max_i I^{(i)}$, and the paper shows this quantity satisfies $C \\ge C_1 \\ge C_{\\rm DET}$, where $C_1$ is the Holevo capacity.","core_discovery":"For any completely positive trace-preserving quantum channel $\\mathcal{E}$ on a $d$-dimensional system, the authors construct a quantity $C_{\\rm DET} = \\max_i I^{(i)}$ from conditional probabilities $p^{(i)}(m|n) = \\langle \\varphi^{(i)}_m | \\mathcal{E}(|\\varphi^{(i)}_n \\rangle \\langle \\varphi^{(i)}_n|) | \\varphi^{(i)}_m \\rangle$, which are obtained by measuring local observables $X_i \\otimes X_i^T$ on a maximally entangled state and using the identity (3). They prove the chain $C(\\mathcal{E}) \\ge C_1(\\mathcal{E}) \\ge C_{\\rm DET}$, establishing $C_{\\rm DET}$ as an experimentally accessible lower bound to the classical capacity. The bound improves monotonically as more observables are measured, and for important qubit-channel classes — all unital qubit channels and pseudoclassical nonunital channels — it saturates to the Holevo capacity, so the certification is tight in those cases.","pith_inferences":["Because $C_{\\rm DET}$ is a maximum over measured bases and optimized priors, finite measurement statistics can in principle push the estimated value above the true capacity; a rigorous experimental protocol would need to report confidence intervals or use a penalized estimator.","The same local correlation measurements could be reused to certify other capacity-like quantities, such as a lower bound on the private classical capacity via the known inequality chain, though this may be considerably looser.","An adaptive strategy that chooses the next observable $X_i$ based on previously estimated transition probabilities could approach the Holevo capacity with fewer measurement settings than a fixed set of bases.","Applying the method to random measurement bases for high-dimensional channels could give a statistical estimate of how many bases are needed to get close to the Holevo capacity, since the bound is monotonically nondecreasing with added bases."],"forward_implications":["For a completely unknown qubit or qutrit channel, the protocol certifies a lower bound on classical capacity using only 1 to $d^2-1$ local measurement settings, whereas full process tomography requires $d^4$ parameters.","For all unital qubit channels the detected bound is tight: $C_{\\rm DET} = C_1 = C$.","For nonunital qubit channels satisfying the pseudoclassicality condition — the shift vector parallel to a principal axis and a sufficiently thin ellipsoid — the detected bound again equals the Holevo capacity.","When the measurement bases are badly matched to the unknown channel, the bound can be loose; for a dephasing channel on an unknown basis the worst-case detected capacity is $1-H(2p/3)$, illustrating the trade-off between basis choice and tightness.","The protocol outputs an explicit encoding (input states and prior probabilities) that achieves the detected lower bound, so it can be directly used to implement a communication scheme."],"supporting_citations":[{"why":"Supplies the trace identity that converts local $X_i \\otimes X_i^T$ measurements on the entangled state into the channel's conditional probabilities; without it the detection protocol does not go through.","marker":"[32]"},{"why":"The prior quantum-capacity detection method that this work extends; provides the entangled-input and local-measurement approach and the benchmark for comparing bound looseness.","marker":"[24]"},{"why":"Blahut–Arimoto recursion used to maximize the mutual information $I^{(i)}$ over input priors for each measured basis.","marker":"[34-36]"},{"why":"Experimental demonstration of the analogous detection scheme, referenced to argue the protocol is implementable with present-day technology.","marker":"[26]"},{"why":"Defines pseudoclassical channels and gives the conditions under which the Holevo capacity is saturated by orthogonal preparations; used to show when $C_{\\rm DET}=C_1$.","marker":"[47]"},{"why":"Establish the regularized definition of classical capacity and the single-letter Holevo bound that underlies the inequality chain $C\\ge C_1\\ge C_{\\rm DET}$.","marker":"[2-4]"},{"why":"Additivity of unital qubit channels, used to identify $C=C_1=C_{\\rm DET}$ for Pauli channels.","marker":"[46]"},{"why":"Shows projective measurements on commuting output states attain the Holevo information, used to argue the $z$-basis measurement saturates $C_1$ for pseudoclassical channels.","marker":"[52]"}],"fun_headline_variants":["Sparse measurements bound classical capacity of unknown channels","Efficient classical capacity bounds from few measurements","Few measurements certify classical capacity of quantum channels","No tomography: capacity bounds from a few measurements"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The detection scheme presupposes a noiseless reference system and ideal projective measurements; under real finite statistics the reconstructed conditional probabilities are noisy, and the paper does not analyze how statistical overestimates could make the detected bound exceed the true capacity.","fun_headline_variants_meta":{"raw":{"variants":["Sparse measurements bound classical capacity of unknown channels","Efficient classical capacity bounds from few measurements","Few measurements certify classical capacity of quantum channels","No tomography: capacity bounds from a few measurements"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000835,"raw_usage":{"total_tokens":3570,"prompt_tokens":797,"completion_tokens":2773,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":413,"completion_tokens_details":{"reasoning_tokens":2716}},"tokens_in":413,"tokens_out":2773,"duration_ms":22062,"temperature":1.0,"reasoning_tokens":2716,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:08:10.417744+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the protocol on a channel whose Holevo capacity is known analytically but whose optimal ensemble is nonorthogonal (e.g., a qubit channel violating pseudoclassicality), using finite measurement statistics, and check whether the estimated $C_{\\rm DET}$ ever exceeds the known $C_1$; exceeding it would invalidate the bound as an experimental guarantee. Alternatively, add a small depolarizing noise to the reference system and observe whether the reconstructed probabilities break the assumptions of the trace identity.","supporting_citations":[{"cited_title":"King and M","cited_arxiv_id":null,"evidence_quote":"Supplies the trace identity that converts local $X_i \\otimes X_i^T$ measurements on the entangled state into the channel's conditional probabilities; without it the detection protocol does not go through."},{"cited_title":"Chruscinski, C","cited_arxiv_id":null,"evidence_quote":"The prior quantum-capacity detection method that this work extends; provides the entangled-input and local-measurement approach and the benchmark for comparing bound looseness."},{"cited_title":"Macchiavello and M","cited_arxiv_id":null,"evidence_quote":"Experimental demonstration of the analogous detection scheme, referenced to argue the protocol is implementable with present-day technology."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines pseudoclassical channels and gives the conditions under which the Holevo capacity is saturated by orthogonal preparations; used to show when $C_{\\rm DET}=C_1$."},{"cited_title":"Braun, O","cited_arxiv_id":null,"evidence_quote":"Additivity of unital qubit channels, used to identify $C=C_1=C_{\\rm DET}$ for Pauli channels."},{"cited_title":"Daems, Phys","cited_arxiv_id":null,"evidence_quote":"Shows projective measurements on commuting output states attain the Holevo information, used to argue the $z$-basis measurement saturates $C_1$ for pseudoclassical channels."}],"review_version":1}