{"id":"25984f4f-d0c1-4749-aae7-def2dc80ec59","arxiv_id":"2501.18546","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Semidefinite compatibility constraints tighten ground-state energy estimates from shot-noise-limited Pauli measurements on small XY spin chains.","lead":"The paper uses semidefinite programming to enforce consistency and physicality constraints on reduced density matrices reconstructed from noisy quantum measurements. In simulations of small spin chains, the constrained estimates give tighter energy brackets than standard tomography and stabilize an algorithmic cooling loop.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The SDP intervals are not certified to contain the true state, so the reported tightness may be an artifact of excluding the truth; a calibration test is needed.","rationale":"The reader's weakest assumption and my concern coincide: the reported energy bounds are not certified to contain the true state because alpha is tuned by feasibility, not by a coverage guarantee. This is the single most load-bearing issue because the abstract and the numerical section explicitly claim tighter bounds and a 10^1--10^2 sample reduction; if the SDP intervals are not valid confidence intervals, those comparisons are apples-to-oranges. The paper is transparent about the low-shot failure mode, which is a credit, but transparency does not supply the missing guarantee. The proposed calibration test would settle whether the advertised advantage is genuine or an artifact of excluding the true state. Other issues, such as the lack of shipped code and the absence of N-representability baselines, are secondary; the coverage problem is what makes the central numerical claim potentially unsound. Since the reader's verdict was already CONDITIONAL, and my concern is exactly the one they flagged, I do not move the verdict; it stays CONDITIONAL pending the calibration check.","tokens_in":17211,"tokens_out":3842,"duration_ms":52156,"concrete_test":"Run a calibration experiment for the n=6 XY chain ground state. For each shot count n_s in {10^1, 10^2, 10^3, 10^4}, draw many independent simulated measurement datasets (e.g., 1000). For each dataset, run Algorithm 1 for both minimization and maximization to obtain the SDP interval, and also form the standard 99% QST interval. Compute the empirical coverage of the exact ground-state energy by each interval. If the SDP interval has materially lower coverage than 99% (or lower than the QST interval's coverage) while being narrower on average, the 'tighter bounds' claim is not a fair comparison. If coverage is comparable, the concern is resolved. Repeat on random Haar states, not just ground states, to rule out special structure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that SDP-assisted tomography yields tighter bounds at the same shot count than standard tomography. This comparison is only meaningful if the SDP min/max solutions are valid bounds on the true energy, which requires the true state's local RDMs to lie inside the SDP feasible set. Algorithm 1 chooses alpha by bisection to the smallest value for which the SDP is feasible, not by a coverage guarantee. Feasibility only certifies that some global-consistent set of local RDMs fits the noisy intervals; it does not certify that the actual state does. The paper itself concedes the failure mode: Section III C 2 states that in low-shot regimes the lower-bound curves can 'exceed the exact ground-state energy.' When the feasible set excludes the true state, the SDP minimum is not a valid lower bound, and the apparent 10^1--10^2 sample reduction relative to standard 99% QST intervals may result from comparing a non-calibrated interval with a calibrated one. The absence of a benchmark against the closest prior N-representability-under-shot-noise methods (Refs. [49,50]) compounds this: there is no evidence that the gain survives when both methods are required to provide equal coverage. This is the load-bearing assumption: the bounds are treated as confidence intervals without a coverage guarantee.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a semidefinite-programming (SDP) post-processing step for local overlapping quantum tomography. Starting from noisy estimates of local reduced density matrices (RDMs) obtained via random Pauli measurements, the method enforces positivity, overlapping-compatibility (OC), and enhanced-compatibility (EC) constraints, then minimizes and maximizes the energy expectation value over the resulting feasible set. The authors claim that, for a fixed number of measurements, this yields tighter energy bounds than standard tomography without compatibility constraints, with reported sample reductions of 10^1 to 10^2 in lower-bounding the ground-state energy of the 1D XY model. They also embed the SDP reconstruction in an algorithmic cooling (AC) procedure and present numerical simulations for n=3,...,8 qubits and various sample counts.","tokens_in":17438,"tokens_out":3975,"duration_ms":46566,"significance":"If the reported tightening is statistically valid, the method would be a useful classical post-processing tool for near-term quantum experiments, allowing better energy estimates from the same measurement budget. The SDP formulation is clearly presented, the constraints are polynomial-size, and the numerical experiments span several system sizes and shot counts, with an explicit disclosure that AC is a heuristic and that no single-run advantage is guaranteed. However, the central comparison to standard tomography currently lacks a calibration or coverage guarantee, so the claimed improvement may be an artifact of comparing an uncalibrated feasible set with a calibrated confidence interval. A coverage analysis would be needed to establish the significance of the numerical claims.","major_comments":[{"comment":"The bracketing inequality in Eq. (22) is stated conditionally on the search region being large enough to contain the true RDMs, but Algorithm 1's bisection only certifies non-emptiness of the feasible set, not containment of the true state. The text in Section III.C.2 explicitly concedes that in low-shot regimes the SDP lower bound can exceed the exact ground-state energy. Consequently, the SDP interval is not a calibrated confidence interval, and the claimed 10^1 to 10^2 sample reduction relative to the 99% standard-QST interval compares an uncalibrated region with a calibrated one. Please add an empirical coverage analysis (e.g., the fraction of trials in which the true ground-state energy lies between the SDP bounds for each shot count) and, if coverage is below the nominal level, either adjust the alpha values to restore coverage or rephrase the claims as heuristic tightening rather than certified bounds.","section":"Section III.C.2, Eq. (22), Algorithm 1"},{"comment":"The free parameter alpha, which sets the width of the per-coefficient intervals epsilon_i^j, is tuned by feasibility bisection rather than by any coverage guarantee. The Chernoff bound motivation applies to each individual Pauli coefficient estimate, but containment of the true state's RDMs in the global feasible set is a joint, correlated condition; per-coefficient intervals with any finite alpha do not automatically imply joint containment. The paper should state explicitly what statistical guarantee, if any, the SDP feasible set carries, or provide numerical or analytical evidence of coverage.","section":"Section III.B, Eq. (13), Algorithm 1"},{"comment":"The closest prior approaches that enforce N-representability conditions on tomographic estimates under shot noise (Refs. [49,50]) are mentioned only as future work in the Conclusion, and no numerical comparison is provided. Since the paper's headline claim is tighter bounds at equal shot count, a comparison against these methods under matched coverage would be needed to establish that the proposed method improves on the state of the art rather than only on independent unconstrained tomography.","section":"Conclusion, Refs. [49,50]"}],"minor_comments":[{"comment":"The 1D nearest-neighbor XY model is described as a 'frustrated Hamiltonian' in the caption of Fig. 5, but this model is not frustrated in the standard sense (no conflicting interaction constraints); please reword or justify the terminology.","section":"Section IV.C, Fig. 5 caption"},{"comment":"The caption states that the green curves use tolerance parameters Δ0 = 0.1 and Δ1 = 0.001, but both are labeled Δ0 in the text; please clarify which tolerance applies to minimization and which to maximization.","section":"Fig. 3 caption"},{"comment":"In Algorithm 1, lines 17 and 18 update the solution and energy estimate inside the while loop even when the SDP is infeasible (in which case S may be undefined from a previous iteration); the pseudocode should guard these assignments so they occur only after a feasible solution is found.","section":"Algorithm 1"},{"comment":"The code availability statement says the code is available 'upon reasonable request'; for a numerical methods paper, depositing the code in a public repository would improve reproducibility and allow readers to verify the coverage concerns raised above.","section":"Section VII"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is clearly written and the SDP formulation is sensible, but the central comparison to standard tomography is not yet calibrated. I would encourage the editor to require a coverage analysis before publication. The lack of comparison to Refs. [49,50] is also worth addressing, as those works appear closely related. The self-citation pattern is moderate and not problematic."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. The method is a reasonable heuristic for post-processing noisy local tomography data, and the numerical study is honest. But the headline claim, tighter bounds for the same measurements, rests on intervals that are not guaranteed to contain the true energy. The apparent 10^1-10^2 sample reduction is likely an artifact of comparing an uncalibrated SDP interval with a calibrated 99% QST interval.\n\nWhat is actually new is the specific combination: random-Pauli overlapping tomography, SDP feasibility with alpha-scaled error bars, overlapping and enhanced compatibility constraints, and the use of the resulting intervals inside an algorithmic cooling loop. That combination is not in the cited literature. The paper is also transparent about the key failure mode: Section III C 2 concedes that in low-shot regimes the SDP lower bound can exceed the exact ground-state energy. That disclosure is to their credit.\n\nWhere it is soft: Eq. (22) is presented as a bracketing inequality, but the feasible set is defined by alpha chosen via bisection on feasibility, not by a coverage guarantee. Feasibility only tells you that some global-consistent set of local RDMs fits the noisy intervals; it does not tell you the true state is inside. When the true state falls outside, the SDP minimum is not a lower bound. The paper openly states this later, but the abstract and the sample-reduction sentence overstate what is shown. The lack of a benchmark against Refs. [49,50], which already apply N-representability constraints to shot-noise reduction, is a real gap; without that comparison the claimed advantage is not established. No code or data are shipped, and \"available on request\" is thin. These issues are addressable, not fatal.\n\nWho gets value: researchers working on practical RDM estimation for variational algorithms, and anyone interested in SDP relaxations of the quantum marginal problem as a post-processing tool. This is a contribution to heuristics, not a certified method.\n\nI would accept it for peer review, but would ask for a calibration test or an explicit re-framing as a heuristic with uncalibrated intervals, a comparison with the prior N-representability-under-shot-noise methods, and code release. With those changes, the paper could be a useful practical contribution.","headline":"A useful heuristic for post-processing noisy local tomography data, but the advertised 10^1-10^2 sample reduction compares uncalibrated SDP intervals with calibrated QST intervals and needs a coverage test before the bounds claim is taken at face value.","tokens_in":17970,"tokens_out":3018,"would_cite":false,"duration_ms":31518,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Classical SDP post-processing of local quantum measurement data tightens ground-state energy bounds and can cut required samples by 10 to 100 times.","keywords":["reduced density matrices","semidefinite programming","shot noise","quantum state tomography","overlapping tomography","quantum marginal problem","algorithmic cooling","variational quantum algorithms"],"falsifier":"Repeatedly run Algorithm 1 on synthetic data from a known 4-qubit XY ground state with a small fixed shot budget, and record the fraction of runs where the reported SDP interval excludes the true energy; if that fraction exceeds the nominal confidence level (e.g., more than 1% of runs at the 99% level), the bracketing claim fails.","tokens_in":16973,"feed_emoji":"⚛️","tokens_out":6100,"duration_ms":61151,"temperature":0.7,"pith_summary":"The paper claims that shot noise in estimating local reduced density matrices (RDMs) can be partially undone in classical post-processing. By solving a polynomial-size semidefinite program that re-imposes positivity, unit trace, and approximate compatibility constraints on overlapping RDMs, the method produces tighter two-sided bounds on ground-state energy than standard tomography alone. Simulations on the 1D XY model show that the SDP approach matches standard tomography's precision with 10 to 100 times fewer samples, and it makes an algorithmic-cooling loop more monotone at low shot counts.","feed_headline":"SDP post-processing cuts quantum tomography samples 10 to 100 times","feed_subtitle":"Re-imposing compatibility on reduced density matrices tightens energy estimates for the same shot budget.","key_machinery":"The central object is a hierarchy of semidefinite relaxations of the quantum marginal problem applied to estimated local RDMs, with the Hamiltonian represented as a hypergraph whose hyperedges are the local interaction terms. Overlapping-compatibility (OC) constraints force two RDM estimates to agree on their shared qubits; enhanced-compatibility (EC) constraints require that each pair of neighboring RDMs be marginals of some larger positive semidefinite matrix. A bisection search (Algorithm 1) tunes a per-coefficient tolerance, proportional to the estimator variance, to the smallest feasible search region, which yields the energy interval compatible with the data.","core_discovery":"The central claim is that compatibility constraints inherited from the quantum marginal problem can be re-imposed on noisy tomographic estimates of overlapping RDMs at polynomial cost, and that doing so yields tighter bounds on the energy of a local Hamiltonian than estimating each RDM independently. The paper constructs a semidefinite program whose decision variables are the local RDMs, each constrained to be positive semidefinite, unit trace, pairwise consistent on shared qubits (overlapping-compatibility), and pairwise extendable to a larger positive matrix (enhanced-compatibility), with every Pauli coefficient allowed to vary inside a confidence interval set by the shot-noise variance. Minimizing and maximizing the energy over this feasible set brackets the exact ground-state energy. Numerically, for the 1D XY model with 3 to 8 qubits, the SDP bracket is narrower than the 99% confidence interval from standard tomography at the same shot budget, and the paper reports a $10^{1}$ to $10^{2}$ reduction in required samples for equal precision in lower-bounding the energy.","pith_inferences":["If a calibration or coverage certificate could be added to the feasibility test, the reported 10- to 100-fold sample reduction might transfer from the XY model to other local Hamiltonians, but this is untested in the paper.","The same feasible-set construction could certify expectation values of non-linear local quantities, such as entanglement witnesses, with the same caveat that feasibility does not by itself guarantee coverage of the true state.","A testable extension would be to use the SDP-reconstructed RDMs to pre-process classical shadow data, potentially reducing the number of shadow samples needed for local observables.","Systematic measurement errors are outside the paper's scope; combining the SDP with error-calibration strategies could determine whether the bounds stay valid on real hardware."],"forward_implications":["For variational algorithms that rely on local RDMs, the same measurement budget can yield tighter energy estimates, reducing the number of shots or iterations needed.","The SDP feasible set bounds not only energy but any linear function of the local RDMs, so correlation functions and other local observables inherit the same tightening.","Because the SDP is polynomial in system size for fixed interaction locality, the post-processing remains efficient as the number of qubits grows, unlike full quantum state tomography.","Embedding the SDP in algorithmic cooling makes the energy descent more monotone in the low-shot regime, making the heuristic more robust to shot noise."],"supporting_citations":[{"why":"Supplies the overlapping tomography measurement framework whose data are post-processed.","marker":"[12]"},{"why":"Supplies the semidefinite relaxation approach for ground-state energies that the hierarchy extends.","marker":"[37]"},{"why":"Provides certification of ground-state properties via SDP, the baseline relaxation method compared in the numerics.","marker":"[34]"},{"why":"Classical shadow tomography, the main alternative measurement scheme that the method is compared against in spirit.","marker":"[6]"},{"why":"Quantum digital cooling, the algorithmic cooling routine into which the SDP reconstruction is embedded.","marker":"[22]"},{"why":"Establishes QMA-completeness of local density matrix consistency, justifying the relaxation to OC and EC constraints.","marker":"[15]"}],"fun_headline_variants":["SDP constraints slash quantum tomography sample needs 10-100x","Quantum tomography sample counts drop 10-100x via SDP","SDP reimposes compatibility, cutting tomography samples 10-100x","SDP constraints on overlapping RDMs save 10-100x samples"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The bounds are trustworthy only if the true state's reduced density matrices actually lie inside the SDP feasible set whenever the bisection finds the set non-empty; the tolerance is tuned by feasibility, not by a coverage guarantee, and the paper concedes that in low-shot regimes the lower bound can dip below the exact ground-state energy.","fun_headline_variants_meta":{"raw":{"variants":["SDP constraints slash quantum tomography sample needs 10-100x","Quantum tomography sample counts drop 10-100x via SDP","SDP reimposes compatibility, cutting tomography samples 10-100x","SDP constraints on overlapping RDMs save 10-100x samples"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000977,"raw_usage":{"total_tokens":4165,"prompt_tokens":975,"completion_tokens":3190,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":591,"completion_tokens_details":{"reasoning_tokens":3112}},"tokens_in":591,"tokens_out":3190,"duration_ms":24744,"temperature":1.0,"reasoning_tokens":3112,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T23:02:55.018905+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeatedly run Algorithm 1 on synthetic data from a known 4-qubit XY ground state with a small fixed shot budget, and record the fraction of runs where the reported SDP interval excludes the true energy; if that fraction exceeds the nominal confidence level (e.g., more than 1% of runs at the 99% level), the bracketing claim fails.","supporting_citations":[{"cited_title":"Cramer, M","cited_arxiv_id":null,"evidence_quote":"Supplies the overlapping tomography measurement framework whose data are post-processed."},{"cited_title":"Jiang, T","cited_arxiv_id":null,"evidence_quote":"Supplies the semidefinite relaxation approach for ground-state energies that the hierarchy extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides certification of ground-state properties via SDP, the baseline relaxation method compared in the numerics."},{"cited_title":"We begin by intro- ducing the cooling principle","cited_arxiv_id":null,"evidence_quote":"Classical shadow tomography, the main alternative measurement scheme that the method is compared against in spirit."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Quantum digital cooling, the algorithmic cooling routine into which the SDP reconstruction is embedded."},{"cited_title":"Huang, R","cited_arxiv_id":null,"evidence_quote":"Establishes QMA-completeness of local density matrix consistency, justifying the relaxation to OC and EC constraints."}],"review_version":1}