{"id":"15e04353-a54b-4aa8-be67-8b982c183e19","arxiv_id":"2601.02893","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Some symmetric Bell inequalities can only be maximally violated by asymmetric minimal-dimension quantum strategies, while the symmetric CGLMP family admits symmetric maximizers up to dimension 19.","lead":"This paper asks whether enforcing party-exchange symmetry costs extra Hilbert-space dimension when maximizing Bell-inequality violations. It finds symmetric inequalities where the optimal minimal-dimension quantum strategy must be asymmetric, and a family (CGLMP) where symmetry costs nothing.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Load-bearing concern: the SQS upper bounds (Fig. 2, Table I) come from a modified QDimSum hierarchy whose constraints are learned from random samples rather than proven; a spurious sampled constraint could artificially lower the bound and create the claimed trade-off.","rationale":"The reader's weakest assumption correctly pinpoints the validity of the modified QDimSum hierarchy in Appendix A1b. This is the single most load-bearing premise for the central claim: the entire trade-off for IS(2) and the six dm=2 inequalities in Table I rests on the SQS upper bounds being true upper bounds. If those bounds are invalid (too low), the reported gap between symmetric-qubit and asymmetric-qubit maxima could be an artifact. The concern is not about disagreement with consensus or about the pure structural results (Propositions 1, 5–7, 9), which appear sound; it is about the numerical method used to prove suboptimality. The paper provides no code, no explicit QS for the dm=2 rows, and no dual certificates, and it openly reports LB<UB for three rows, so the numerical bounds are not independently checkable. A rigorous, certified SDP relaxation for the two-qubit SQS problem would settle whether the gap is real. Since the reader already flagged this and issued a CONDITIONAL verdict, my analysis does not move the verdict; it reinforces the need for that condition.","tokens_in":31207,"tokens_out":17433,"duration_ms":166312,"concrete_test":"For IS(2) at α=2, carry out a rigorous, certified SDP relaxation of the two-qubit SQS problem (e.g., using the Navascués–Vértesi hierarchy for dimension 2 with explicit Cayley–Hamilton constraints and the constraint that both parties use the same POVM), and compare its upper bound with 9; if the certified upper bound exceeds 9, the modified QDimSum bound is too low and the central claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Appendix A1b computes upper bounds for symmetric quantum strategies (Eq. A1) by modifying QDimSum: it samples pure states from the symmetric/antisymmetric subspace and sets Bob's POVM equal to Alice's. QDimSum 'learns' algebraic constraints by random sampling; these constraints are not proven. If a constraint that holds for the finite random sample but not for all SQS enters the SDP, the feasible set is too small and the reported upper bound can lie below the true SQS maximum. This is exactly the regime of the central claim: for IS(2) (α=2), the SQS bound equals the local bound 9, while the asymmetric two-qubit value is 9.1407; a false lower bound would erase the gap. The same issue underlies the six dm=2 rows of Table I; explicit QSs matching the bounds are not published, and for the dm=3 rows the caption admits LB<UB, so the numerical evidence is not independently checkable. The paper also gives no dual SDP certificates (e.g., rational feasible solutions) that would certify the bounds. Thus the existence of a symmetry-dimension trade-off for these inequalities is not yet rigorously established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates, for symmetric (party-permutation-invariant) Bell inequalities, whether the maximal quantum violation can always be attained by a symmetric quantum strategy (SQS) acting on the minimal local Hilbert-space dimension. It proves structural results (mirror-symmetric strategies give symmetric correlations; asymmetric maximizers force a one-dimensional flat boundary of the quantum set and preclude self-testing) and presents a mix of analytic and numerical evidence. For the CGLMP family the authors find SQSs matching NPA upper bounds for dimensions 2–19, indicating no trade-off. For the IS(α) family, nine (2,4,2) facet inequalities, and a nine-setting inequality I9, they report a gap between the maximal quantum value and the value attainable by SQSs in the minimal dimension, concluding that some symmetric Bell inequalities are maximally violated only by asymmetric minimal-dimension strategies. The I9 result is supported by an analytic proof (Prop. 9); the other trade-off claims rely on numerical SDP bounds from a modified QDimSum hierarchy described in Appendix A1b.","tokens_in":31522,"tokens_out":3267,"duration_ms":34779,"significance":"If fully established, the paper's central claim would be a clean example of asymmetry as a resource in Bell scenarios: some symmetric Bell inequalities force asymmetric minimal-dimension strategies, which in turn implies flat regions of the quantum correlation set and limits self-testing from maximal violation. The paper contains several valuable analytic contributions: Prop. 5 and Prop. 6 characterize mirror-symmetric qubit strategies; Prop. 7 and Cor. 8 are simple but important implications for geometry and self-testing; Prop. 9 gives a rigorous proof of a symmetry–dimension trade-off for I9. The explicit quantum strategies for IS(2) and J42, and the matching of CGLMP SQS values with NPA bounds, are also useful data points. However, the numerical upper bounds for SQSs in Fig. 2 and Table I are load-bearing and currently lack rigorous certification.","major_comments":[{"comment":"The SQS upper bounds in Fig. 2 and Table I are obtained by a modified QDimSum hierarchy in which the algebraic constraints are learned from random sampling of symmetric/antisymmetric states and of identical POVMs for both parties. These constraints are not proven to be valid for all SQSs, and no dual feasible solutions are provided. For IS(2), the SQS bound is reported to equal the local bound 9, while the asymmetric two-qubit value is 9.1407; a single spurious sampled constraint would erase the claimed gap. The same issue affects all six dm=2 rows of Table I. The authors should either (i) prove the sampled constraints hold for every SQS of the relevant dimension, (ii) provide rigorous dual certificates for the reported SDP bounds, or (iii) explicitly relabel these as heuristic upper bounds and consequently limit the strength of the trade-off claim.","section":"Appendix A1b / Fig. 2 / Table I"},{"comment":"The central trade-off claim for IS(α) — in particular the statement that symmetric qubit strategies cannot even violate the inequality for α ∈ (1.975, 3] — rests entirely on the unproven numerical SQS bound from Appendix A1b. The asymmetric lower bound 9.1407 for IS(2) and its matching with the NPA hierarchy are convincing for the quantum maximum, but the SQS side is not. Without a certified or analytic upper bound on SQSs, the existence of a symmetry–dimension trade-off for this family is not rigorously established. Given that the abstract states 'we show that symmetric quantum strategies ... can only lead to a suboptimal Bell violation', this gap is load-bearing and should be addressed, either by added proof/certification or by softening the claim to 'provide numerical evidence'.","section":"Section V.A / Eq. (26)"},{"comment":"For the three dm=3 inequalities (I8, I19, I13), the caption itself admits that the best lower bound falls short of the SDP upper bound. These rows therefore do not demonstrate a trade-off, only an upper bound on SQS values. For the six dm=2 rows, the lower bounds are said to match the upper bounds, but the explicit QS parameters are not published, making the matching non-reproducible. The paper would be strengthened by providing the QS parameters (e.g., in a supplementary file) or a reproducible script, especially because the entire trade-off claim for these inequalities rests on the numerical agreement.","section":"Table I / caption"}],"minor_comments":[{"comment":"The title contains a typo: 'Bell ineq uality' should be 'Bell inequality'.","section":"Title/Abstract"},{"comment":"The symbol '⟳' is used without definition; the text says it refers to 'six remaining terms' but the reader must guess the precise symmetry convention. Please define it explicitly.","section":"Eq. (B9)"},{"comment":"The entry 'I2233 [54]: Eq. (22)' appears inconsistent with the notation I22dd used elsewhere; please check the label.","section":"Table II"},{"comment":"The proof of Prop. 7 assumes the symmetric group action commutes with the linear functional in the sense used in Eq. (54). This is fine, but it may be worth explicitly noting that Vσ maps Q to itself, which is needed for the conclusion that the convex combination remains in Q. This is implicit but should be stated.","section":"Section VII.A"},{"comment":"The definition of the J42 inequality in Eq. (B9) uses a placeholder '⟳' in the sum, which obscures the actual expression. A fully explicit form is essential for reproducing the numerical values in Table I.","section":"Appendix B.4"}],"recommendation":"major_revision","confidential_remarks":"The central structural results (Props 5–9) are clean and likely correct. The main risk is the numerical SQS upper bound method in Appendix A1b: it is a heuristic sampling-based relaxation with no proved validity or dual certificates. Since the paper's headline trade-off claims for IS(2) and the (2,4,2) inequalities depend on these bounds, the manuscript would need either a rigorous certification of those bounds or a substantial reframing as numerical evidence, not a proof. The analytic I9 example is a strong positive result and could anchor a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper is worth reading and refereeing, but the headline claim is not yet proven. The cleanest example (IS(2)) rests on a numerical SQS upper bound from a modified QDimSum hierarchy whose sampling-derived constraints are not certified, so the gap could be an artifact. Also, Table I calls three rows trade-offs where the lower bound falls short of the upper bound.\n\nWhat's genuinely new: the symmetry-vs-dimension framing, the explicit symmetric CGLMP strategies for d=2..19, the mirror-symmetric strategies that produce symmetric correlations, and the clean geometric consequence (Prop 7/Cor 8) that an asymmetric maximizer implies a flat boundary and rules out self-testing from the maximal violation alone. Props 5 and 6 are solid. The CGLMP evidence is decent: the SQS value matches the NPA upper bound, so there is no trade-off there up to numerical precision.\n\nMain soft spot: the SQS upper bounds in Fig 2 and Table I come from a modified QDimSum that 'learns' constraints from a finite sample of symmetric/antisymmetric states and identical POVMs. Those constraints are not proven to hold for all SQSs. If a sampled constraint is spurious, the SDP can underestimate the true SQS maximum. For IS(2) the SQS bound equals the local bound, so the entire trade-off for that inequality rests on this unverified bound. The paper does not supply dual certificates or rational feasible solutions. That is a load-bearing gap, not a technicality.\n\nSecond, the dm=3 rows in Table I are explicitly incomplete (LB<UB in the caption), yet the text groups them with the matched rows. That is an overclaim. Also, the explicit QSs for the six dm=2 rows are not published, so the lower bounds are not independently checkable. No code or data is shipped.\n\nSome 'observations' are numerical rather than proven (the continuous IS(α) curve, CGLMP d=6..19). That is acceptable if labelled as evidence, but the paper sometimes uses 'show' where 'provide evidence for' is more accurate.\n\nWho it's for: quantum information researchers working on Bell inequalities, dimension bounds, and self-testing. The conceptual parts are useful. But the main existence result should be treated as conditional until the SQS upper-bound method is either proven or replaced by a certified bound.\n\nRecommendation: send to peer review, but with major revision requested: prove or certify the SQS upper bound for the key examples, publish the QS parameters, and mark the unmatched rows as partial. If the authors can do that, this becomes a solid subfield contribution. If not, it remains a set of interesting numerical observations.","headline":"A credible and interesting phenomenon, but the central claim leans on unproven numerical upper bounds and a couple of overclaims; worth refereeing, but needs a rigorous revision before I'd treat the main theorem as established.","tokens_in":32078,"tokens_out":3972,"would_cite":false,"duration_ms":38433,"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":"Some symmetric Bell inequalities can only reach their maximum quantum violation through asymmetric strategies operating in the minimal Hilbert-space dimension.","keywords":["Bell inequality","quantum nonlocality","party-permutation symmetry","Hilbert-space dimension","asymmetric quantum strategy","self-testing","quantum correlations","semidefinite programming"],"falsifier":"Find an explicit symmetric two-qubit strategy (a party-symmetric state with identical measurements on both sides) that attains the value 1/3(13+4√13) ≈ 9.1407 for IS(2), or that exceeds the local bound 2α+5 for some α in (1.975, 3]. Equivalently, run a higher-level SDP hierarchy with the symmetry constraints and show that the upper bound drops below the asymmetric two-qubit value for IS(2).","tokens_in":1564,"feed_emoji":"⚛️","tokens_out":2823,"duration_ms":62593,"temperature":0.7,"pith_summary":"This paper asks whether a symmetry of a Bell test—invariance under exchanging the two parties—can be kept when one insists on using the smallest possible quantum system to get the maximal violation. The authors find that for several symmetric Bell inequalities, the answer is no: any strategy that respects the symmetry and uses qubits is strictly worse than an asymmetric two-qubit strategy. They show this for a family of three-setting inequalities and for nine four-setting facet-defining inequalities, and they prove a general consequence: if an asymmetric correlation maximizes a symmetric inequality, the set of quantum maximizers contains a flat one-parameter region, which makes self-testing from the maximal violation alone impossible. For the CGLMP family, by contrast, no such trade-off appears: symmetric strategies of minimal dimension already achieve the maximum.","feed_headline":"Some symmetric Bell inequalities demand asymmetric quantum strategies","feed_subtitle":"Identical qubit measurements fall short; asymmetric strategies win, blocking self-testing from maximal violation alone.","key_machinery":"The central object is the family IS(α) of symmetric Bell inequalities in correlator form, together with nine symmetric facet-defining inequalities in the (2,4,2) scenario. The key mechanism is Proposition 7: because the Bell functional is linear and the quantum set is convex, an asymmetric maximizer of a symmetric inequality forces a one-parameter flat boundary in the quantum set. The numerical bound on symmetric qubit strategies comes from a dimension-bounded SDP hierarchy in which states are sampled from the symmetric/antisymmetric subspace and Bob's POVMs are forced equal to Alice's; lower bounds come from explicit two-qubit strategies with degenerate measurements.","core_discovery":"For several symmetric (party-permutation-invariant) Bell inequalities, including the family IS(α) in the (2,3,2) scenario and nine symmetric facet-defining inequalities in the (2,4,2) scenario, the maximal quantum violation can only be attained by an asymmetric quantum strategy of minimal dimension. Numerical evidence shows that symmetric qubit strategies strictly underperform the global maximum, while explicit asymmetric two-qubit strategies match the upper bound from the Navascués-Pironio-Acín hierarchy. Consequently, the set of quantum maximizers contains a flat one-parameter region (Prop. 7, Cor. 8), making self-testing from the maximal violation alone impossible. In contrast, the CGLMP-","pith_inferences":["The flat-region phenomenon identified here suggests that device-independent protocols relying on the maximal violation of a symmetric Bell inequality may need to add extra constraints (e.g., on the number of outcomes or on other moments) to pin down the underlying state and measurements; otherwise the one-parameter family of maximizing correlations leaves genuine freedom.","The mirror-symmetric construction (Eqs. 29–30) provides a template for building further examples of asymmetric strategies that produce symmetric correlations; a systematic search over such strategies could reveal more inequalities with a symmetry–dimension trade-off, possibly in simpler Bell scenarios.","The numerical gap reported for the three dm=3 inequalities in Table I (where lower and upper bounds differ) suggests that a higher-level SDP or an analytic sum-of-squares certificate could either close the gap or reveal that those examples require an even larger symmetric dimension than currently found.","If the trade-off holds generally, then in experiments aiming at maximal Bell violation with minimal-dimensional systems, one should expect the optimal implementations to be lopsided; this has practical consequences for how measurement settings are chosen in device-independent experiments."],"forward_implications":["For the symmetric inequalities IS(α) (with α in the relevant range) and the nine listed four-setting inequalities, any quantum strategy reaching the maximum must break the party-exchange symmetry; respecting the symmetry forces a higher-dimensional system or a suboptimal violation.","The set of quantum correlations that maximally violate these inequalities contains a one-parameter flat boundary, so the maximal violation alone cannot be used to self-test a specific reference strategy (Corollary 8).","For J^42_{4422}, the hierarchy of values 0.5682 (qubit SQS) < 0.6012 (qubit symmetric correlations) < 0.6722 (general qubit maximum) implies that a symmetric correlation can still certify that the underlying qubit strategy is asymmetric, giving a semi-device-independent witness of asymmetry.","The CGLMP-type symmetric inequalities I22_dd do not exhibit a trade-off for d = 2,...,19: a symmetric strategy of dimension d achieves the maximal violation, so symmetry and minimal dimensionality can coexist there.","There exist asymmetric Bell inequalities (the two-parameter family Ir0,r1) whose maximal violation is attained by a symmetric correlation, showing that the direction of the trade-off is not universal."],"fun_headline_variants":["Symmetry costs Bell-inequality violation","Asymmetric qubits win maximal Bell violation","Symmetric Bell inequalities can't self-test","Trade symmetry for dimension in Bell tests"],"cache_read_input_tokens":33280,"weakest_assumption_plain":"The load-bearing premise is that the modified QDimSum hierarchy (Appendix A1b) correctly upper-bounds the maximum of a symmetric Bell inequality over symmetric quantum strategies of bounded local dimension; if that bound is not tight or is flawed, the observed gaps between symmetric and asymmetric qubit values could be numerical artifacts.","fun_headline_variants_meta":{"raw":{"variants":["Symmetry costs Bell-inequality violation","Asymmetric qubits win maximal Bell violation","Symmetric Bell inequalities can't self-test","Trade symmetry for dimension in Bell tests"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00016,"raw_usage":{"total_tokens":1068,"prompt_tokens":743,"completion_tokens":325,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":487,"completion_tokens_details":{"reasoning_tokens":280}},"tokens_in":487,"tokens_out":325,"duration_ms":4250,"temperature":1.0,"reasoning_tokens":280,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T12:26:55.886665+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find an explicit symmetric two-qubit strategy (a party-symmetric state with identical measurements on both sides) that attains the value 1/3(13+4√13) ≈ 9.1407 for IS(2), or that exceeds the local bound 2α+5 for some α in (1.975, 3]. Equivalently, run a higher-level SDP hierarchy with the symmetry constraints and show that the upper bound drops below the asymmetric two-qubit value for IS(2).","supporting_citations":[],"review_version":1}