{"id":"8ae93078-fb67-4722-bda1-a8e42f1919ca","arxiv_id":"2505.02297","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A Schmidt number criterion based on the trace norm of correlation matrices from informationally complete (N,M)-POVMs is derived and shown to encompass SIC and MUB criteria.","lead":"The authors derive a new algebraic test for detecting the Schmidt number of a bipartite quantum state using a flexible family of symmetric measurements called (N,M)-POVMs. The test generalizes two earlier criteria, and the paper provides examples where it detects entanglement that the earlier GSIC criterion misses.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed superiority over existing criteria is not established: all numerical examples fix the new criterion's free parameter t=0.01 while comparing against fixed GSIC parameters, and no optimization over either family is reported.","rationale":"The reader's weakest-assumption analysis targets exactly the parameter choices in the numerical comparisons, and I agree that this is the main vulnerability. The theorem's proof appears correct, and the special cases reduce properly, so I do not see an internal inconsistency that would invalidate the central mathematical construction. However, the paper's headline claim of superiority is not established by the examples as presented: the new criterion's free parameter is fixed to a single value, the competing criteria are evaluated at fixed literature parameters, and the measurement settings differ in size. These are all concrete, checkable issues rather than matters of taste. Because the same concern was already identified and the reader's verdict was CONDITIONAL, my stress test does not change the verdict. A full parameter optimization could either confirm or refute the claimed advantage, which is exactly the test to settle the concern.","tokens_in":11323,"tokens_out":14631,"duration_ms":172108,"concrete_test":"Recompute Examples 1, 2, and 4 by grid-searching the allowed ranges of t for the (N,M)-POVM criterion (as given by the positivity constraints in Section II) and, for the GSIC comparison, the allowed parameter a from Ref. [18]. For each example, record the widest interval of the mixing parameter where each criterion certifies Schmidt number at least 2. If the optimized (N,M)-POVM interval is not strictly wider than the optimized GSIC interval (and, where relevant, the realignment interval), the paper's superiority claim loses its numerical support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1 and Corollary 1 appear mathematically sound: Lemma 1 checks out for non-Hermitian operators, the pure-state reduction via convexity is valid, and the SIC/MUB limits reproduce the known criteria. The load-bearing weakness is the paper's central advertised claim, stated in the abstract and conclusions, that the symmetric-measurement criterion is 'more effective than and superior to' existing Schmidt-number criteria. This is supported only by Examples 1-4, and in every example the new criterion is evaluated at t=0.01 for the (N,M)-POVMs, while the competing GSIC criterion is evaluated at the fixed parameters a1=0.1277, a2=0.04984, a=0.04984 taken from Ref. [18]. These parameter values are not shown to be optimal or even representative. Since the (N,M)-POVM criterion has a free parameter t (equivalently x), and the GSIC criterion also has a free parameter a, the comparison conflates a particular choice of the new measurement with the best achievable performance of the new method. Moreover, in Example 1 the new measurements use more outcomes per local measurement than the GSIC comparison (6 and 20 effects versus 4 and 16), so some of the apparent advantage may simply reflect additional measurement resources rather than an intrinsic superiority of the criterion. A fair comparison requires optimizing over all allowed parameters for both methods, and the paper does not provide such an optimization or the code/data needed to verify it. Thus the mathematical result may stand, but the 'superiority' conclusion is conditional on an unperformed comparison.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives a Schmidt-number criterion for bipartite quantum states based on the trace norm of the correlation matrix obtained from informationally complete (N,M)-POVMs. Theorem 1 gives a necessary condition for a state to have Schmidt number at most r, and Corollary 1 specializes to equal local dimensions. The authors also derive a symmetric-measurement lower bound on concurrence. They claim, on the basis of four examples, that their criterion is more effective than and superior to existing Schmidt-number criteria based on GSIC POVMs, MUBs, fidelity, and realignment.","tokens_in":11630,"tokens_out":4394,"duration_ms":57251,"significance":"The mathematical derivation of Theorem 1 and Corollary 1 appears coherent: Lemma 1 correctly generalizes the frame decomposition to non-Hermitian operators, the reduction to pure states via convexity is valid, and the limiting cases reproduce the GSIC and MUB criteria. If the superiority claim were properly supported, this would be a useful unifying framework for Schmidt-number detection. However, the paper's central advertised contribution—the claimed practical advantage over existing criteria—is not established by the examples as presented, because the comparisons fix the new criterion's free parameter at a single value and do not optimize over the competing criteria.","major_comments":[{"comment":"The central claim that the new criterion is 'more effective than and superior to' existing criteria is supported only by numerical examples in which the (N,M)-POVM parameter is fixed at t=0.01 while the competing GSIC criterion is evaluated at fixed parameters from Ref. [18]. Since both the (N,M)-POVM bound and the GSIC bound depend monotonically on their free parameters, a single choice is not representative of the achievable performance. The paper should either optimize over the allowed parameters for both methods, or prove analytically that t=0.01 (or the chosen value) is optimal or at least dominates the entire competing family. Without this, the examples do not substantiate the abstract's superiority claim.","section":"Section IV, Examples 1, 2, and 4"},{"comment":"The comparison in Example 1 does not control for measurement resources: the (3,2)-POVM on subsystem A has 6 effects and the (5,4)-POVM on subsystem B has 20 effects, whereas the GSIC comparison uses 4 and 16 effects, respectively. A correlation matrix from a larger informationally complete measurement can generically yield stronger bounds, so the observed advantage may be attributable to the larger number of outcomes rather than to an intrinsic superiority of the criterion. A fair comparison should either use the same number of outcomes or explicitly account for the additional resource cost.","section":"Section IV, Example 1"},{"comment":"The threshold values that support the superiority claim (e.g., 0.42115 in Example 1, 0.5219 in Example 2) are quoted from figures without numerical data, derivations, or code. Since the entire advertised advantage rests on these examples, the reader cannot verify the thresholds or the curves. The authors should provide the numerical data or a reproducibility script, or at least give the explicit algebraic expressions used to generate the figures.","section":"Section IV, Examples 1, 2, and 4"},{"comment":"The conclusion states that the criterion is 'more effective than and superior to' the fidelity criterion, but Example 3 only proves that for isotropic states the new criterion is not weaker than the fidelity criterion. The text explicitly says 'our criterion must not be weaker'—that is not a demonstration of strict superiority. The wording in the abstract and conclusions overstates what Example 3 shows.","section":"Section V and Example 3"}],"minor_comments":[{"comment":"The display defining H_{\\alpha,k} is garbled: the line '((\\sqrt{M}+1)G_{\\alpha,k}=M' is incomplete and the piecewise definition is not legible. This should be rewritten with the correct formula for both k=1,\\dots,M-1 and k=M.","section":"Section II, definition of H_{\\alpha,k}"},{"comment":"The appendix title refers to 'Example 3' but the (8,2)-POVM is used in Example 4; the numbering should be corrected.","section":"Section IV, Example 4 and Appendix B"},{"comment":"There are typographical errors such as 'green curve is is' and 'critera' instead of 'criteria' in the surrounding text.","section":"Section IV, Example 4"},{"comment":"The sentence 'in practical applications, Neither determining whether a given state is entangled nor describing the entanglement degree of a given entangled state is an easy task' contains a capitalization error and awkward phrasing; it should be revised.","section":"Section I, Introduction"},{"comment":"The remark that Fig. 3 'is also a supplement to Fig.4 in Ref.[35]' is unexplained and should be either elaborated or removed.","section":"Section IV, Example 4"}],"recommendation":"major_revision","confidential_remarks":"The mathematical criterion in Theorem 1 appears sound and is a legitimate generalization of existing results. The main obstacle to publication is the unsupported superiority claim: all numerical evidence fixes the new parameter at a single value and does not optimize the competing criteria, and at least one comparison uses more measurement outcomes for the new method. I would require a fair parameter-resource comparison with reproducible data before accepting the paper's advertised conclusion."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read on Wang-Fei (arXiv:2505.02297).\n\nThe main inequality, Theorem 1, is a real generalization of the Tavakoli-Morelli correlation trace-norm criterion to informationally complete (N,M)-POVMs. The proof of Lemma 1 is a straightforward frame decomposition, and I checked the special cases: for a single GSIC POVM it reduces to the criterion of Wang et al. [18], for MUBs it reduces to the MUB criterion of [17], and for sigma=I the normalization works out. The proof of Corollary 1 via conical 2-designs is a nice observation, and the concurrence lower bound (Eq. 3) is a legitimate add-on. So the mathematical core is solid and, as far as I can see, new.\n\nThe soft spot is the advertised \"more effective and superior\" claim. It rests on Examples 1-4, and in every example the new criterion is evaluated at t=0.01 while the GSIC competitor is evaluated at the fixed parameters a1=0.1277, a2=0.04984 from Ref. [18]. There is no optimization over the free parameter in either family, and the paper gives no code or data to reproduce the figures. On top of that, in Example 1 the new measurements use more outcomes (6 and 20) than the GSIC ones (4 and 16), so some of the apparent advantage may just be more measurement resources. That doesn't kill the theorem, but it means the abstract's claim is too broad as written. A fair comparison would either optimize both families or at least state that the advantage holds for the chosen parameters rather than claim universal superiority.\n\nThere are also minor presentational issues: a few typos (\"critera\", \"is is\"), and the figures lack explicit numeric curves. I'd ask the authors to provide the data/code and to either optimize the parameters or substantially soften the superiority claim.\n\nVerdict: this deserves a serious referee. The inequality is correct and useful for the Schmidt number detection toolbox, and the generalization to symmetric measurements is a legitimate step forward. But the paper needs revision before acceptance, mainly to make the comparison honest and reproducible.\n\nRecommendation: send to peer review, and in the report ask for code/data and a fair parameter comparison.","headline":"The main inequality is correct and the generalization is real, but the superiority claim over existing criteria is not established by the examples as presented.","tokens_in":12177,"tokens_out":6435,"would_cite":true,"duration_ms":72252,"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":"Any bipartite state with Schmidt number at most r must have a correlation-matrix trace norm below a threshold set by the symmetric measurements; exceeding it certifies Schmidt number at least r+1.","keywords":["Schmidt number","symmetric measurements","quantum entanglement","correlation matrix","(N,M)-POVM","Schmidt number criterion","concurrence","entanglement detection"],"falsifier":"For the family $\\rho(0.9,q)$ in Example 1, compute the entanglement-detection threshold of the GSIC POVM criterion with its free parameter $a$ optimized rather than fixed at $0.1277$ and $0.04984$; if the optimized GSIC criterion detects entanglement for $q$-values in $[0.42115,1]$ or beyond, the claimed advantage of the $(3,2)/(5,4)$-POVM choice with $t=0.01$ fails. More broadly, optimize $t$ in Theorem 1 on each example family and compare with optimized older criteria to see whether the reported detection intervals are actually the best this method can achieve.","tokens_in":11124,"feed_emoji":"⚛️","tokens_out":11411,"duration_ms":114468,"temperature":0.7,"pith_summary":"This paper tries to establish a practical necessary condition for the Schmidt number of a bipartite quantum state: if $\\rho_{AB}$ has Schmidt number at most $r$, then the trace norm of its correlation matrix, built from outcomes of a pair of informationally complete symmetric measurements, cannot exceed $L/K + (r-1)R/K$. Exceeding that bound certifies that the state has Schmidt number at least $r+1$, which matters because higher Schmidt numbers give a quantifiable advantage in tasks like channel discrimination. The result is framed in terms of $(N,M)$-POVMs, a family that contains the previously used SIC POVMs and mutually unbiased bases as special cases, so the criterion generalizes both earlier Schmidt-number criteria. The authors also show, in examples, that the new condition can detect entanglement and Schmidt number above 2 in parameter regions where the GSIC, fidelity, and realignment criteria fail, and they extract a family of concurrence lower bounds from the same proof.","feed_headline":"One correlation-matrix bound reveals Schmidt number","feed_subtitle":"Symmetric measurements unify older entanglement tests and catch states those tests miss.","key_machinery":"The load-bearing object is the correlation matrix $P(\\rho_{AB})$ assembled from an informationally complete $(N,M)$-POVM: a set of $N$ $d$-dimensional POVMs with $M$ outcomes each that satisfy fixed trace, overlap, and cross-overlap conditions, with $N(M-1)=d^2-1$ for informational completeness. The key identity is Lemma 1, a frame-style resolution formula: for every linear operator $\\sigma$, $$\\sum_{\\$\\alpha$,k}|\\mathrm{tr}(E_{\\$\\alpha$,k}\\$\\sigma$)|^2 = \\frac{d($M^{2}$ x-d)\\mathrm{tr}(\\$\\sigma$\\$\\sigma$^\\dagger) + ($d^{3}$ - $M^{2}$ x)|\\mathrm{tr}(\\$\\sigma$)|^2}{dM(M-1)},$$ which converts measurement probabilities into Hilbert-Schmidt data. The proof of the theorem splits a pure Schmidt-rank-$r$ state into diagonal parts $\\langle ss|E^A_{\\alpha,k}\\otimes E^B_{\\beta,l}|ss\\rangle$ and off-diagonal parts $\\langle ss|E^A_{\\alpha,k}\\otimes E^B_{\\beta,l}|tt\\rangle$, bounds the trace norm of each piece with Lemma 1, and uses convexity plus $(\\sum_s \\lambda_s)^2 \\leq r$ to reach the bound. The measurement family is what does the generalizing work: GSIC POVMs and mutually unbiased bases are the two limiting cases of the same construction.","core_discovery":"The central claim is Theorem 1: for any bipartite state $\\rho_{AB}$ on $H_A \\otimes H_B$ with Schmidt number at most $r$, the trace norm of the correlation matrix $P(\\rho_{AB})$ satisfies $$\\|P(\\rho_{AB})\\|_{\\mathrm{tr}} \\leq \\frac{L}{K} + \\frac{(r-1)R}{K},$$ where $K,L,R$ depend only on the local dimensions and on the parameters $x_A,x_B$ of the chosen informationally complete $(N,M)$-POVMs. Here $P(\\rho_{AB})$ is the matrix of joint measurement probabilities $p_{\\alpha,k;\\beta,l} = \\mathrm{tr}(\\rho_{AB}(E^A_{\\alpha,k}\\otimes E^B_{\\beta,l}))$. Since the condition is necessary, any state whose measured trace norm exceeds the $r$-threshold provably has Schmidt number at least $r+1$. Setting $M_A=d_A^2$, $M_B=d_B^2$, $N_A=N_B=1$ recovers the GSIC POVM criterion, and setting $M_A=d_A$, $N_A=d_A+1$ (and similarly for $B$) with $x=1$ recovers the MUB criterion, so the theorem is a common generalization of both. The proof also yields a class of lower bounds on concurrence for heterogeneous-dimensional bipartite states from the same correlation matrix, generalizing the known equal-dimensional bound.","pith_inferences":["Because the proof only invokes the algebraic identity that makes the measurements a conical 2-design, the same trace-norm bound should be derivable for any measurement family that forms a conical 2-design, not just the symmetric POVMs treated here.","The examples fix the free parameter at $t=0.01$; optimizing $t$ for each state family could lower the detection thresholds further than the reported intervals, a direct testable extension of the paper's numerics.","The correlation matrix is built from joint probabilities of local measurements, so in principle it can be estimated from experimental coincidence counts; the criterion could be turned into a data-driven Schmidt-number witness, though the paper does not perform such an estimation."],"forward_implications":["If the measured trace norm exceeds the bound for a given $r$, the state's Schmidt number is at least $r+1$, so the criterion gives a directly computable lower bound on entanglement dimensionality.","The criterion specializes to the GSIC POVM criterion when each side uses one $d^2$-outcome measurement and to the MUB criterion when using $d+1$ $d$-outcome bases, so improvements in detection carry over to both settings.","For $r=1$ the inequality reduces to the known symmetric-measurement separability criterion, making the result an entanglement witness as well as a Schmidt-number test.","The same proof supplies lower bounds on concurrence for bipartite systems of unequal dimensions, extending the equal-dimensional bound previously obtained.","In the worked examples, the new condition certifies entanglement or Schmidt number greater than 2 on wider parameter intervals than the GSIC, fidelity, and realignment criteria, indicating stronger detection on those states."],"supporting_citations":[{"why":"Defines the (N,M)-POVM class and gives the Hermitian version of Lemma 1 that this paper generalizes to non-Hermitian operators; the measurement family is the basis of the criterion.","marker":"[22]"},{"why":"Supplies the correlation trace-norm methodology and the SIC/MUB Schmidt-number criteria that Theorem 1 extends and recovers.","marker":"[17]"},{"why":"Gives the GSIC POVM Schmidt-number criterion, the immediate predecessor that Theorem 1 generalizes and the main comparison in the examples.","marker":"[18]"},{"why":"Introduces the Schmidt number and the fidelity criterion, which Example 3 shows the new criterion is never weaker than, and sets the definition the paper tests.","marker":"[1]"},{"why":"Provides the symmetric-measurement lower bound on concurrence that Theorem 1's proof generalizes to heterogeneous systems.","marker":"[25]"},{"why":"Supplies the Horodecki bound entangled states used to construct the state families in Examples 1 and 4 where the new criterion outperforms earlier ones.","marker":"[32]"},{"why":"Supplies the mixed two-ququart state family used in Example 2 to compare against the realignment criterion.","marker":"[33]"},{"why":"Gives the realignment/CCNR Schmidt-number criterion used as a comparison baseline in Examples 2 and 4.","marker":"[5, 6]"}],"fun_headline_variants":["Symmetric-measurement trace norm sets Schmidt number bound","One trace norm suffices: Schmidt number from symmetric POVMs","Schmidt number criterion unifies GSIC and MUB tests","Tighter Schmidt number test from symmetric measurements"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The superiority over previous criteria rests on the specific parameter choices in the examples — $t=0.01$ for the symmetric measurements and $a=0.1277$, $a=0.04984$ for the GSIC POVMs — and could shrink or vanish if those comparison criteria were optimized over their free parameters.","fun_headline_variants_meta":{"raw":{"variants":["Symmetric-measurement trace norm sets Schmidt number bound","One trace norm suffices: Schmidt number from symmetric POVMs","Schmidt number criterion unifies GSIC and MUB tests","Tighter Schmidt number test from symmetric measurements"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000145,"raw_usage":{"total_tokens":1151,"prompt_tokens":888,"completion_tokens":263,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":196}},"tokens_in":504,"tokens_out":263,"duration_ms":3933,"temperature":1.0,"reasoning_tokens":196,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:56:59.937632+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the family $\\rho(0.9,q)$ in Example 1, compute the entanglement-detection threshold of the GSIC POVM criterion with its free parameter $a$ optimized rather than fixed at $0.1277$ and $0.04984$; if the optimized GSIC criterion detects entanglement for $q$-values in $[0.42115,1]$ or beyond, the claimed advantage of the $(3,2)/(5,4)$-POVM choice with $t=0.01$ fails. More broadly, optimize $t$ in Theorem 1 on each example family and compare with optimized older criteria to see whether the reported detection intervals are actually the best this method can achieve.","supporting_citations":[{"cited_title":"Kalev and G","cited_arxiv_id":null,"evidence_quote":"Defines the (N,M)-POVM class and gives the Hermitian version of Lemma 1 that this paper generalizes to non-Hermitian operators; the measurement family is the basis of the criterion."},{"cited_title":"Tavakoli and S","cited_arxiv_id":null,"evidence_quote":"Supplies the correlation trace-norm methodology and the SIC/MUB Schmidt-number criteria that Theorem 1 extends and recovers."},{"cited_title":"rP s=1 λ2 s (L−R) + rP s=1 λs 2 R # = L K + R K","cited_arxiv_id":null,"evidence_quote":"Gives the GSIC POVM Schmidt-number criterion, the immediate predecessor that Theorem 1 generalizes and the main comparison in the examples."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the Schmidt number and the fidelity criterion, which Example 3 shows the new criterion is never weaker than, and sets the definition the paper tests."},{"cited_title":"Siudzinska, Informationally overcomplete measurements from generalized equiangular tight frames, J","cited_arxiv_id":null,"evidence_quote":"Provides the symmetric-measurement lower bound on concurrence that Theorem 1's proof generalizes to heterogeneous systems."},{"cited_title":"Huang, F","cited_arxiv_id":null,"evidence_quote":"Supplies the Horodecki bound entangled states used to construct the state families in Examples 1 and 4 where the new criterion outperforms earlier ones."}],"review_version":1}