{"id":"2f9ce482-feac-4e84-90e9-bbb28e63e505","arxiv_id":"2505.10462","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"For the simulated rank-1 qudit states, the Heisenberg-Weyl observable basis consistently needs fewer measurements than the generalized Gell-Mann basis as qudit dimension grows, while both give high-fidelity reconstruction.","lead":"This paper numerically compares two measurement bases for compressed sensing quantum state tomography of qudit systems. It finds that the Heisenberg-Weyl observable basis becomes more measurement-efficient than the generalized Gell-Mann basis as qudit dimension increases, while both bases remain viable.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Rank-1, N=2-only simulations do not support the unqualified claim that HWO becomes more efficient as k increases; mixed-rank or larger-N tests are needed.","rationale":"The paper's headline contribution is a quantitative scaling statement: HWO requires fewer measurements than GGM at large k, with a gap growing roughly as k^2. This is extracted from numerical CS-QST simulations performed only for rank-1 pure states of two qudits. The least secure link in the argument is the implicit extrapolation from these rank-1, N=2 results to the unqualified statement that \"the HWO basis becomes more efficient as the qudit dimension increases.\" Compressed sensing targets low-rank states generally, and the sample complexity depends on rank r; the coherence mechanism invoked to explain the trend is rank-sensitive, so the observed advantage might shrink, vanish, or reverse for modest ranks. The authors explicitly limit the simulations to r=1 and N=2, so the abstract and summary overreach. A concrete rerun with r=2 and r=4 would settle whether the central claim holds beyond pure states. The coherence-section normalization mismatch (Eq. 17 vs Eq. 31) is a genuine but secondary issue: it undermines the quantitative explanatory narrative, not the numerical comparison itself. Because the numerical comparison is the core result and the generalization is untested, a CONDITIONAL verdict remains appropriate; this stress test does not move the reader's verdict.","tokens_in":16384,"tokens_out":10827,"duration_ms":104966,"concrete_test":"Repeat the Sec. III B protocol unchanged except replacing the Ginibre rank parameter r=1 with r=2 and r=4, for N=2 at k=9, 12, and 15, using the same depolarizing noise, measurement noise, SVT settings, and m95% definition. Compute m95% for both bases from 50 repetitions and also from 200 repetitions to quantify sampling uncertainty. If the HWO-minus-GGM gap in m95% does not grow with k for r>1, or is within the repetition-induced error, the abstract claim must be restricted to rank-1 pure states.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central m95% comparison (Figs. 2-3, Eqs. 28-29) is computed exclusively for rank-1 pure states (Sec. III B: \"we consider only pure states with rank r=1\"), with a fixed 5% depolarizing channel and Gaussian measurement noise of standard deviation 0.1/d, for N=2 qudits. The abstract and summary state without qualification that \"the HWO basis becomes more efficient as the qudit dimension increases.\" Compressed sensing QST is motivated by low-rank states generally, not only pure states; the sample complexity bound is r d log d, and the recovery threshold can depend on rank through the coefficient distribution and coherence constants. No mixed-state or N>2 simulations are reported, and no error bars are given for the m95% values used in the fits, so the k^2 gap is an extrapolation from rank-1 behavior alone. If, for r=2 or r=4, the HWO advantage does not persist or the gap stops growing with k, the abstract's unqualified claim is false. The coherence argument in Sec. II C, which is offered as the mechanism, also has a normalization inconsistency: Eq. (17) defines nu_min via d max ||B_a||_2 (spectral norm), but Eq. (31) evaluates it as (max eps_n^2)^N, a squared-spectral-norm quantity off by a k- and N-dependent factor. This weakens the theoretical support for extrapolation but does not by itself invalidate the numerical comparison.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper numerically compares compressed sensing quantum state tomography (CS-QST) for qudit systems using two operator bases: the generalized Gell-Mann (GGM) basis and the Heisenberg-Weyl observable (HWO) basis. The authors generate rank-1 Ginibre states, apply 5% depolarizing noise and Gaussian measurement noise, reconstruct via singular value thresholding, and evaluate fidelity and trace distance. For two-qudit systems (N=2) with qudit dimension k=3,6,9,12,15, they report that the HWO basis requires fewer randomly selected measurements to achieve a mean fidelity minus one standard deviation of at least 0.95, and they fit the measured m_95% to functions of k (Eqs. 28-29), claiming the GGM-HWO gap grows approximately as k^2. They also compare fixed-dimension d=16 systems and attribute the observed differences to the coherence properties of the two bases, defined via a 'minimum coherence' ν_min (Eq. 17). The abstract states without qualification that 'the HWO basis becomes more efficient as the qudit dimension increases.'","tokens_in":16656,"tokens_out":15324,"duration_ms":134509,"significance":"If the reported trend is robust, the paper offers practically useful guidance for choosing measurement bases in qudit CS-QST, an area relevant to high-dimensional quantum information processing. The numerical protocol is clearly described (Ginibre states, SVT, AIC model selection), and the paper provides explicit fitting formulas that can be tested against other simulations or experiments. The main theoretical support, however, rests on a coherence calculation that, as written, is inconsistent with the paper's own definition of ν_min, and the empirical claims are based on a restricted set of simulations (rank-1, N=2) without error bars on the m_95% values. These issues do not, by themselves, invalidate the numerical comparison, but they make the central claim stronger than the evidence presented.","major_comments":[{"comment":"The coherence measure ν_min defined in Eq. (17) as d max_a ||B_a||_2 is inconsistent with the values reported in Eq. (31) and Table I for the HWO basis. With B_a = (ω_{i1}⊗...⊗ω_{iN})/√(k^N) as in Eq. (2), which the text calls an orthonormal basis, the spectral norm of B_a is (max_i ||ω_i||_2)^N / k^{N/2}, so ν_min = k^{N/2} (max_i ||ω_i||_2)^N, not (max_i ||ω_i||_2)^N as stated in Sec. II C. For the HWO basis, using the eigenvalue expression in Eq. (30), this gives ν_min = k^{N/2} (max_n ε_n^2)^{N/2}, not (max_n ε_n^2)^N as in Eq. (31). For example, for k=3, max_n ε_n^2 = 1.866 but ||W(0,1)||_2 = 1.366, so the value 1.866 reported in Table I is a squared eigenvalue, not a spectral-norm coherence. Under the corrected definition, HWO ν_min grows as k^{N/2}(√2)^N for k a multiple of 8, rather than remaining constant. This invalidates the statement in Sec. II C that ν_min for the HWO basis is 'nearly constant regardless of k' and undercuts the coherence-based explanation in Sec. III C. The authors should correct the normalization in Eq. (31) and revise Fig. 1, Table I, and the related discussion accordingly.","section":"Eq. (17), Eq. (31), Table I, Fig. 1"},{"comment":"The abstract and Sec. IV state without qualification that 'the HWO basis becomes more efficient as the qudit dimension increases,' but the numerical evidence is obtained only for N=2 two-qudit systems with rank-1 pure states (Sec. III B: 'we consider only pure states with rank r=1'), a fixed 5% depolarizing channel, and Gaussian measurement noise with standard deviation 0.1/d. No simulations for mixed states (r>1) or for N>2 are reported, so the unqualified claim is an extrapolation from a narrow regime. The abstract and summary should either explicitly restrict the claim to the simulated conditions (e.g., 'for two-qudit pure states under the noise model considered') or the authors should add simulations for other ranks and numbers of qudits to support the general statement.","section":"Abstract and Secs. III B, III C 1"},{"comment":"The quantitative claim that the GGM-HWO gap in m_95% grows roughly as k^2 is based on least-squares fits to m_95% values, but the m_95% data points are reported without error bars. Each m_95% is determined from 50 repetitions (Sec. III B), so the sample mean and standard deviation used in the condition 'mean fidelity minus one standard deviation ≥ 0.95' carry statistical uncertainty that propagates to m_95%. Since no uncertainties are reported for the fitted parameters in Eqs. (28)-(29), the fitted k^2 scaling is not statistically grounded. The authors should provide error estimates for m_95% (e.g., via bootstrapping) and report confidence intervals for the fit parameters, or otherwise demonstrate that the observed trend is robust to the sampling uncertainty.","section":"Sec. III C 1, Eqs. (28)-(29), Fig. 3"}],"minor_comments":[{"comment":"The text describes the target state as a rank-1 pure state but then applies depolarizing noise to obtain \\tilde{ρ}, which has full rank. Since the fidelity in Eq. (25) is computed against the original ρ, the effective reconstruction target is a pure state corrupted by noise, not the depolarized state itself; this is a reasonable protocol, but it would be clearer to state explicitly that ρ, not \\tilde{ρ}, is the target for the fidelity and trace distance.","section":"Sec. III B"},{"comment":"The sentence 'Section II C presents our numerical simulations' is a typo; it should refer to Section III.","section":"Sec. II C"},{"comment":"The sentence 'The work of was supported' is incomplete; the author names are missing.","section":"Acknowledgments"},{"comment":"The connection between the fitted m_95% curves and the theoretical scaling m ∼ O(rd log^2 d) is only qualitative; the fitted forms have p_1 >> p_0, so the leading behavior is m_95% ∼ p_1 k^2, and the paper would benefit from a brief explanation of why the logarithmic factor is subleading in the simulated range.","section":"Sec. III C 1"},{"comment":"Equation (31) uses the notation (max_n ε_n^2)^N without explicitly stating that ε_n are the eigenvalues of the single-qudit HWO operator; relating this to the corrected definition of ν_min discussed above would improve clarity.","section":"Appendix B"}],"recommendation":"major_revision","confidential_remarks":"This is a useful numerical study with a clear protocol, but the coherence analysis contains a concrete normalization error (Eq. (17) vs. Eq. (31)) that affects the theoretical interpretation of the central result. The error is fixable by recomputing ν_min with the proper B_a normalization and revising the coherence plots and discussion. Additionally, the abstract's unqualified claim goes beyond the simulated rank-1, N=2 setting, and the m_95% fits lack error bars. I do not see grounds for rejection; the underlying numerical comparison may well be correct, but the paper needs a major revision to align its claims with its evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper gives a clean numerical answer to a narrow question: for CS-QST of two-qudit systems on rank-1 pure states, the HWO basis needs fewer measurements than GGM as k grows, and the gap in m95% scales roughly as k^2. The qualitative result was already expected from coherence theory; the quantitative fits are new and could be a useful rule of thumb for experimentalists.\n\nWhat the paper does well: the simulation protocol is standard and well specified — Ginibre states, 5% depolarizing, Gaussian noise 0.1/d, SVT with all parameters stated, and AIC used for model selection. The appendices with explicit GGM/HWO matrices for k=4 and the coherence values in Table I are helpful. The fixed-dimension comparison (d=16) across qubit, k=4 qudit, and k=16 qudit is a nice addition.\n\nThe soft spots are real but not fatal. The abstract and introduction claim \"the HWO basis becomes more efficient as the qudit dimension increases\" without mentioning that all scaling data come from N=2 and rank r=1. Section III B is explicit: \"we consider only pure states with rank r=1.\" Since CS-QST is motivated by low-rank but generally mixed states, the unqualified claim goes beyond the evidence. I would expect the qualitative trend to survive for mixed states because the coherence advantage is state-independent, but the m95% thresholds could shift, and the k^2 gap is fitted from pure-state data alone. A few r=2 or r=4 runs would fix this.\n\nThere is also a genuine notation/calculation mismatch in the coherence section. Eq. (17) defines ν_min as d times the spectral norm of B_a, but the values in Table I and Eq. (31) are computed as max eigenvalue squared — i.e., squared spectral norm, without the factor of d. For GGM, this changes the scaling from O(√k) to O(k). The qualitative conclusion survives either way, but the paper should not cite the definition and then evaluate a different quantity without comment.\n\nMinor: no code, data, or m95% tables are provided, so exact reproduction is inconvenient. The AIC fits rely on 14 points with no error bars on the m95% values, so the k^2 gap inference is softer than the fits suggest.\n\nWho this is for: people planning qudit CS-QST experiments who want a practical comparison of basis choices, and theorists working on coherence-based bounds in tomography. It deserves a serious referee. The question is legitimate, the numerics are honest, and the overgeneralization is fixable. With the rank/N restrictions stated in the abstract and the coherence formula corrected, this becomes a useful reference for the subfield. I would send it to peer review.","headline":"Solid, clearly described numerical comparison of GGM vs HWO bases for two-qudit CS-QST on rank-1 states, but the abstract overgeneralizes beyond what was simulated.","tokens_in":17237,"tokens_out":6510,"would_cite":false,"duration_ms":58809,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.-a","03.65.Wj"],"model":"deepseek-v4-flash","headline":"Compressed sensing quantum state tomography for qudits is more measurement-efficient with the Heisenberg-Weyl observable basis than with the generalized Gell-Mann basis, and the advantage grows with the qudit dimension.","keywords":["compressed sensing quantum state tomography","qudit tomography","Heisenberg-Weyl observable basis","generalized Gell-Mann basis","coherence","singular value thresholding","fidelity","measurement basis selection"],"falsifier":"Repeat the same protocol for two-qudit states of ranks $r=2$ and $r=3$ at $k=9,12,15$ and for $N=3$ or $4$ qudits, keeping the same noise model, and compute $m_{95\\%}^{\\mathrm{GGM}}-m_{95\\%}^{\\mathrm{HWO}}$; if the gap stops growing as $k^2$, or reverses, the central claim holds only within the simulated regime.","tokens_in":16135,"feed_emoji":"⚛️","tokens_out":9898,"duration_ms":77763,"temperature":0.7,"pith_summary":"This paper asks whether the choice of measurement-operator basis changes how many measurements compressed sensing quantum state tomography (CS-QST) needs to reconstruct an unknown qudit state. By numerically reconstructing random pure states of two qudits under 5% depolarizing and Gaussian measurement noise, it establishes that the Heisenberg-Weyl observable (HWO) basis reaches a mean fidelity minus one standard deviation of 0.95 with fewer randomly selected measurements than the generalized Gell-Mann (GGM) basis, and that the gap grows roughly as $k^2$ with qudit dimension $k$. The paper attributes this to coherence: the GGM basis has minimum coherence $\\nu_{\\min}$ scaling as $(k-1)^N$, while the HWO basis stays near $2^N$. If correct, the result gives a concrete rule for choosing bases in high-dimensional quantum tomography, while leaving GGM as a workable alternative at small $k$.","feed_headline":"Heisenberg-Weyl basis cuts measurement count for high-dim qudits","feed_subtitle":"The gap versus the Gell-Mann basis grows roughly as the square of the qudit dimension, simulations show.","key_machinery":"The load-bearing object is the minimum coherence $\\nu_{\\min}\\equiv d\\max_a\\|B_a\\|_2$ defined in Eq. (17) from compressed sensing theory, which measures how concentrated tomographic weight is across the basis elements. For the GGM basis, the diagonal operators have spectral norms that grow with $k$, giving $\\nu_{\\min}=(k-1)^N$, while for the HWO basis all non-identity elements share the same spectral norm, giving $\\nu_{\\min}\\approx 2^N$ (exactly $2^N$ when $k$ is a multiple of 8, and exactly $1$ for $k=4$). The paper uses this coherence difference to predict and explain why HWO requires fewer random measurements, alongside the singular value thresholding (SVT) algorithm used for reconstruction.","core_discovery":"The central claim is that in CS-QST of two-qudit systems, the HWO basis becomes more efficient than the GGM basis as the qudit dimension $k$ increases: the minimum number $m_{95\\%}$ of randomly selected measurement operators needed to keep mean fidelity minus one standard deviation at or above 95% is smaller for HWO, and the gap between the two bases grows approximately as $k^2$ (Eqs. 28-29, Fig. 3). The paper explains this through the minimum coherence $\\nu_{\\min}=d\\max_a\\|B_a\\|_2$: for GGM it grows as $(k-1)^N$, whereas for HWO it is nearly constant in $k$, taking values close to $2^N$ and equalling $2^N$ exactly when $k$ is a multiple of 8. The same coherence argument accounts for the paper's fixed-dimension comparison at $d=16$, where HWO-based qudit tomography matches Pauli-based qubit tomography while GGM degrades with increasing $k$.","pith_inferences":["A direct testable extension is to run the same comparison for mixed low-rank states ($r=2,3$); if the coherence argument is the true driver, the HWO advantage should persist because coherence is a property of the basis rather than of the state rank.","The coherence analysis suggests a mitigation for GGM: sampling measurement operators with probabilities weighted by their spectral norms, or always including the diagonal operators, could reduce the gap; this is worth testing.","The same basis-efficiency logic likely applies to other compressed-sensing recovery tasks in quantum information, such as channel or process tomography, where the measurement basis is also chosen rather than fixed.","Because the paper fixes the SVT algorithm and its hyperparameters, the size of the gap may shift with a different reconstruction algorithm; repeating the comparison with another matrix-completion method would separate basis effects from algorithm effects."],"forward_implications":["For two-qudit systems, the number of measurements needed to reach 95% mean fidelity minus one standard deviation grows more slowly with $k$ for the HWO basis than for the GGM basis, and the gap grows roughly as $k^2$.","At a fixed Hilbert space dimension $d=16$, qudit tomography using the HWO basis matches the efficiency of four-qubit tomography using the Pauli basis, while GGM efficiency degrades as $k$ increases.","The coherence analysis predicts that GGM becomes less reliable at small $m/d^2$ because its diagonal operators have larger spectral norms, and the simulations confirm larger error bars for GGM in that regime.","Both bases eventually achieve high fidelity, so an experiment that can more easily implement GGM measurements can still use GGM at low or moderate $k$; high-dimensional implementations should prefer HWO."],"supporting_citations":[{"why":"Supplies the compressed sensing quantum state tomography framework and the $O(rd\\log^2 d)$ scaling law used as the reference for the paper's fitted measurement counts.","marker":"[4]"},{"why":"Defines the coherence conditions that the paper uses to compare the GGM and HWO bases and to predict recoverability.","marker":"[27]"},{"why":"Constructs the Heisenberg-Weyl observable basis and notes that all HW observables have maximal rank, motivating their universality.","marker":"[24]"},{"why":"Defines the generalized Gell-Mann basis as the SU(k) generators used as the alternative measurement basis.","marker":"[23]"},{"why":"Provides the singular value thresholding algorithm used in all the paper's reconstructions.","marker":"[5]"},{"why":"Supplies the random-state generation procedure used to create the target states for the simulations.","marker":"[32]"}],"fun_headline_variants":["HWO basis wins for high-dim qudit state tomography","Gell-Mann vs Heisenberg-Weyl: HWO scales better for qudits","Qudit CS-QST: Heisenberg-Weyl basis requires fewer measurements","High-dim qudits: Heisenberg-Weyl basis beats Gell-Mann efficiency","Which basis cuts measurement count in qudit tomography? HWO"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central claim rests on simulations limited to rank-1 pure states of two qudits with a fixed 5% depolarizing channel and Gaussian measurement noise of standard deviation $0.1/d$; the general conclusion that HWO becomes more efficient as qudit dimension increases assumes this ordering survives for mixed low-rank states, more qudits, and other noise levels.","fun_headline_variants_meta":{"raw":{"variants":["HWO basis wins for high-dim qudit state tomography","Gell-Mann vs Heisenberg-Weyl: HWO scales better for qudits","Qudit CS-QST: Heisenberg-Weyl basis requires fewer measurements","High-dim qudits: Heisenberg-Weyl basis beats Gell-Mann efficiency","Which basis cuts measurement count in qudit tomography? HWO"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000223,"raw_usage":{"total_tokens":1481,"prompt_tokens":995,"completion_tokens":486,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":611,"completion_tokens_details":{"reasoning_tokens":384}},"tokens_in":611,"tokens_out":486,"duration_ms":4870,"temperature":1.0,"reasoning_tokens":384,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:09:28.566401+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the same protocol for two-qudit states of ranks $r=2$ and $r=3$ at $k=9,12,15$ and for $N=3$ or $4$ qudits, keeping the same noise model, and compute $m_{95\\%}^{\\mathrm{GGM}}-m_{95\\%}^{\\mathrm{HWO}}$; if the gap stops growing as $k^2$, or reverses, the central claim holds only within the simulated regime.","supporting_citations":[{"cited_title":"By varyingNandkwhile keepingd=k N constant, we com- pare the relative advantages of different representations","cited_arxiv_id":null,"evidence_quote":"Supplies the compressed sensing quantum state tomography framework and the $O(rd\\log^2 d)$ scaling law used as the reference for the paper's fitted measurement counts."},{"cited_title":"H¨ affner et al., Nature (london), Nature438, 643 (2005)","cited_arxiv_id":null,"evidence_quote":"Provides the singular value thresholding algorithm used in all the paper's reconstructions."}],"review_version":1}