{"id":"4cee5cc6-375c-41fd-8a66-469c1c1b6bb0","arxiv_id":"1908.06669","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Every nontrivial tight correlation Bell inequality for bipartite binary-outcome systems is violated by some quantum strategy.","lead":"This paper proves that every tight Bell inequality for two-player correlation games is violated by quantum entanglement. It settles an open question and shows that classical and quantum boundaries can only coincide on lower-dimensional faces.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"The reader accepted the paper with high confidence, identifying strict positivity of the dual multipliers as the weakest assumption. I examined that step and found it sound: PSDness of the dual slack matrix forces t_i > 0 for exhaustive games, so the matrix F is well defined and the dimension bound follows. I also stress-tested the more terse part of the proof, the reduction from non-exhaustive to exhaustive games in Theorem 2, because a failure there would directly threaten the central claim. The stated codimension bounds are consistent with a monomial-counting analysis of the face: the available independent functions are the linear terms and the products u_i u_j, u_i s_l, r_k u_j, r_k s_l, which give exactly the claimed dimension upper bounds. The only case in which the correlation-polytope face can be a facet is m_A = m_B = 1, corresponding to the trivial inequalities |c_xy| <= 1. Thus the nontrivial theorem is well supported. The abstract and title overstate the scope by omitting 'nontrivial', since trivial facets have no quantum advantage, but this is a presentation flaw rather than a mathematical gap in the theorem as stated in the main text. No load-bearing objection to the central claim remains.","tokens_in":14915,"tokens_out":41003,"duration_ms":424061,"concrete_test":"Independently re-derive the complementary-slackness step: solve the dual SDP (14) explicitly for an exhaustive XOR game with known no quantum advantage (e.g. the n = 2 non-local computation game) and verify that every optimal dual solution has all t_i > 0 and that all optimal deterministic strategies satisfy beta = F alpha for the same matrix F. This directly tests the load-bearing positivity assumption and the linear-determinacy step on which the dimension bound relies.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The reader's flagged assumption, strict positivity of the dual multipliers t_i in Supplementary Material C after Eq. (14), is in fact forced: if t_i = 0, PSDness of Diag(t) - tilde{Phi} forces row i (or column i) of tilde{Phi} to vanish, contradicting exhaustiveness. Hence F = Lambda^{-1}Phi^T is well defined. The terse non-exhaustive extension in Theorem 2 also checks out: counting the monomials u_i u_j, u_i s_l, r_k u_j, and r_k s_l reproduces the stated codimension bounds, including the correlation-polytope bound Delta_0 >= M_A(m_B - m_A) + m_A(m_A+1)/2, whose only non-large case is m_A = m_B = 1, exactly the trivial |c_xy| <= 1 facets. There is a harmless factor-of-1/2 inconsistency in the displayed definition of F from the complementary-slackness equations, but only linear determinacy of Bob's strategy from Alice's is used. One presentation issue remains: the title and abstract omit the qualifier 'nontrivial', since the trivial facets have no quantum advantage; the theorem statement in the main text is correct as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies two-player XOR games with binary outcomes and the associated correlation Bell inequalities. It proves (Theorem 1) that if an exhaustive XOR game has no quantum advantage, the face of the local Bell polytope, and of the correlation polytope, cut out by the corresponding inequality has codimension strictly greater than one, and hence is not a facet. Theorem 2 extends the argument to non-exhaustive games and concludes that every nontrivial tight correlation Bell inequality has a quantum violation, answering Gill's question affirmatively for the correlation polytope. Theorem 3 transfers the geometric statement to the quantum correlation sets: nontrivial XOR games do not define facets of Qcom or Q⊗, and the set Q0 of quantum correlations has no nontrivial facets. The proofs use semidefinite programming duality, complementary slackness, and Tsirelson's characterization of quantum correlators, with full details in the Supplementary Material, including a separate treatment of non-local computation games.","tokens_in":15159,"tokens_out":15846,"duration_ms":160125,"significance":"If correct, this is a clean structural answer to a long-standing question: facet-defining correlation Bell inequalities are exactly the ones that are nontrivially violated by entanglement, apart from the trivial probability-positivity facets |c_xy| ≤ 1. The proof strategy is novel and elegant, deriving linear determinacy of optimal strategies from complementary slackness and then bounding the affine span of the face. The paper also improves on earlier non-local-computation results by giving explicit, asymptotically attainable dimension bounds, and it yields a striking by-product: the quantum correlation set has no nontrivial facets. The manuscript includes a rigorous SDP-duality proof with full supplementary details, and I verified the one step flagged by an initial reading as potentially missing: the strict positivity of the dual multipliers t_i follows from PSDness together with the assumption that the game matrix has no all-zero rows or columns. There are no fitted parameters or circular inputs; the earlier NLC result is cited but independently reproven.","major_comments":[],"minor_comments":[{"comment":"The title and the first sentence of the abstract claim that every tight correlation Bell inequality has a quantum advantage, omitting the qualifier 'nontrivial' that appears in Theorem 2; as written they are contradicted by the trivial facets |c_xy| ≤ 1, which are tight in the correlation polytope but have no quantum advantage. Please add the qualifier in both places.","section":"Title and abstract"},{"comment":"The definition of Γ as (1/2)Σ ⊕ Λ and the subsequent equations Σ|α_c⟩ = Φ|β_c⟩ and Λ|β_c⟩ = Φ^T|α_c⟩ contain an inconsistent factor of 1/2, so the displayed F = Λ^{-1}Φ^T is not the literal consequence of the preceding line; since only linear determinacy is used this is harmless, but the scaling should be fixed for internal consistency.","section":"Supplementary Material C, around Eq. (15)"},{"comment":"The sentence 'by our assumption that Φ has no all-zero rows or columns, we even can conclude that all t_i > 0' is asserted without proof; it follows from the 2x2 principal minor condition t_i t_j ≥ (Φ_ij/2)^2 together with exhaustiveness, so adding a one-line justification would remove a potential concern for the reader.","section":"Supplementary Material C, after Eq. (14)"},{"comment":"The dimension count for the non-exhaustive case is very compressed: the codimension bound is stated as 'we thus arrive at' without exhibiting the affine-span calculation for the free marginals and the cross terms. Expanding this step, or explicitly referencing a supplementary section that contains it, would make the proof of the titular theorem easier to verify.","section":"Theorem 2 proof, main text"},{"comment":"The term 'nontrivial' is used in Theorem 2 but is only characterized in Supplementary Material B; a parenthetical definition in the main text (inequalities other than ±c_xy ≤ 1) would improve readability.","section":"Main text, after Theorem 2"}],"recommendation":"minor_revision","confidential_remarks":"The paper is technically sound; the mathematical arguments check out, including the strict positivity step that initially looked under-supported. The remaining issues are presentational, but the title/abstract overclaim is visible enough that it should be fixed before publication. I see no basis for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing to know: this paper settles, for the correlation polytope, the old question of whether every tight Bell inequality has a quantum violation. It shows that any two-player XOR game (equivalently, any correlation Bell inequality) with no quantum advantage defines a face of high codimension, not a facet. The contrapositive gives the title result: all nontrivial tight correlation Bell inequalities are violated by quantum strategies.\n\nThe novelty is in the proof. They write the quantum optimum as an SDP, use strong duality, and from complementary slackness derive that when the classical and quantum maxima coincide, Bob's optimal strategy is linearly determined by Alice's (β_c = F α_c). The dimension of the face is then bounded by the number of free parameters in Alice's strategy plus the off-diagonal of α α^T, which is far below the facet dimension. This is a fresh approach and also yields a nice byproduct: the quantum correlation set Q_0 has no nontrivial facets. The supplementary material is solid; the NLC special case is reproven with sharper dimension bounds.\n\nSoft spots are minor. The title and abstract say 'all tight correlation Bell inequalities' without the qualifier 'nontrivial,' which will confuse readers since the trivial |c_xy|≤1 inequalities are facets without quantum advantage. The proof of Theorem 2 in the main text is compressed to the point of being hard to follow—the codimension formula is garbled around 'of the face of' and there's a notation slip (M_A vs m_A). You'll want to do the dimension count yourself, but it works out. The strict positivity of the dual multipliers t_i is asserted rather than proved; it follows from exhaustiveness, so it's not a gap. In the supplement there's a harmless factor-of-1/2 inconsistency in the complementary-slackness equations defining F, but only linear determinacy is used.\n\nWho should read this? Anyone working on Bell inequalities, nonlocal games, or the geometry of quantum correlations. It resolves a question that has been around since 2005 and gives a clean geometric explanation for why no-quantum-advantage XOR games are rare in the tightness sense. I'd assign it in a reading group, and I'd cite the main theorem in my own work.\n\nMy recommendation: this deserves a serious referee. It's correct, the result is significant, and the writing issues are fixable with a careful revision. Send it to review.","headline":"A correct and significant answer to Gill's question for XOR games: no-quantum-advantage correlation Bell inequalities are never facets, so all nontrivial tight ones are quantum violated.","tokens_in":15650,"tokens_out":30751,"would_cite":true,"duration_ms":274153,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P40","81P45","52B11","90C22"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that every nontrivial tight correlation Bell inequality is violated by quantum entanglement.","keywords":["Bell inequalities","XOR games","quantum violation","facet-defining","local polytope","semidefinite programming duality","quantum correlations","non-local computation"],"falsifier":"Search for an exhaustive XOR game with $\\xi_Q=\\xi_c$ whose optimal classical face has dimension $D-1$ in the full Bell polytope, or dimension $m_A m_B$ in the correlation polytope; the theorem says neither can happen. A direct numerical check on small cases, say $m_A=2$, $m_B=3$, would enumerate all optimal sign vectors and compute the affine dimension of the face, then compare with the bound $\\frac{1}{2}m_A(m_A-1)$; if such a face exists, Theorem 1 fails.","tokens_in":14769,"feed_emoji":"⚛️","tokens_out":9101,"duration_ms":86879,"temperature":0.7,"pith_summary":"The paper answers a long-open question: must every tight Bell inequality be violated by quantum mechanics? For two-party, binary-outcome correlation inequalities, the answer is yes. A tight inequality is one that defines a facet of the polytope of local classical correlations, so it belongs to any minimal description of the classical boundary. The proof shows geometrically that if an XOR game has no quantum advantage, the face on which classical and quantum maxima coincide has codimension at least two, and so cannot be a facet. The same mechanism shows the set of quantum correlations has no nontrivial facets, leaving only the trivial non-negative probability bounds as facet-defining.","feed_headline":"Every tight correlation Bell inequality has a quantum violation","feed_subtitle":"A facet is a minimal classical bound; this proof shows entanglement beats every one in the correlation case.","key_machinery":"The central device is the semidefinite-programming dual of the quantum bias of an XOR game. The game matrix $\\Phi$ is embedded into a block matrix $\\tilde\\Phi = \\frac{1}{2}\\begin{pmatrix}0&\\Phi\\\\ \\Phi^T&0\\end{pmatrix}$; the quantum bias is the maximum of $\\operatorname{tr}(\\tilde Q\\tilde\\Phi)$ over Gram matrices $\\tilde Q$ with all diagonal entries equal to 1. The dual introduces diagonal multipliers $t_i$, and strong duality plus complementary slackness imply that at an optimum with equal classical and quantum values, the matrix $\\Gamma-\\tilde\\Phi$ annihilates the optimal classical vector $|s\\rangle=|\\alpha_c\\rangle\\oplus|\\beta_c\\rangle$, where $\\Gamma=\\sum_i t_i |i\\rangle\\langle i|$ is strictly positive for exhaustive games. This splits into $\\Sigma|\\alpha_c\\rangle=\\Phi|\\beta_c\\rangle$ and $\\Lambda|\\beta_c\\rangle=\\Phi^T|\\alpha_c\\rangle$, hence $|\\beta_c\\rangle=F|\\alpha_c\\rangle$ with $F=\\Lambda^{-1}\\Phi^T$, a fixed matrix independent of the strategy. The linear constraint leaves only Alice's $m_A$ sign vectors free, and the correlator block contributes at most $\\frac{1}{2}m_A(m_A-1)$ dimensions, giving the codimension bound that rules out facets. The same machinery, with the same $F$, applies to optimal quantum strategies and yields the facetlessness of the quantum set.","core_discovery":"On the paper's own terms, the central claim is Theorem 2: every nontrivial tight correlation Bell inequality for bipartite systems with binary outcomes admits a quantum violation. Equivalently, no XOR game whose classical and quantum maximum biases coincide can define a facet of the Bell polytope. Writing the game matrix as $\\Phi$, the classical bias is $\\xi_c = \\max_{\\alpha_c,\\beta_c} \\langle \\alpha_c|\\Phi|\\beta_c\\rangle$, and the quantum bias is the optimum of a semidefinite program over Gram matrices with unit diagonal. The proof assumes $\\xi_Q=\\xi_c$ and uses complementary slackness in the SDP dual to force a linear relation $|\\beta_c\\rangle = F|\\alpha_c\\rangle$ between Alice's and Bob's optimal deterministic strategies, with $F=\\Lambda^{-1}\\Phi^T$ fixed by the dual multipliers. Consequently the face of classical behaviours maximising the inequality is contained in the affine span of Alice's strategy space, whose dimension is at most $m_A + \\frac{1}{2}m_A(m_A-1)$; this is far below the facet dimension $D-1$. The same $F$-relation, applied to optimal quantum strategies, bounds the dimension of faces of the quantum correlation set, proving that it has no facets except the trivial inequalities $|c_{xy}|\\le 1$.","pith_inferences":["Editorial inference: The same complementary-slackness mechanism is likely to transfer to MOD-$q$ XOR-like games with $q$-ary outputs, which the paper names as an open direction; the strict-positivity condition would be the main thing to re-check.","Editorial inference: The facetlessness of the quantum correlation set suggests that no single linear inequality can cleanly separate classical from quantum correlations at a maximal face; separations, if any, must be witnessed by lower-dimensional faces or by families of inequalities.","Editorial inference: A direct numerical scan of small XOR games with no quantum advantage could test the tightness of the codimension formula and reveal exactly which games saturate the bound, providing a finite catalogue of extremal no-advantage faces."],"forward_implications":["Every facet-defining correlation Bell inequality for two parties with binary outcomes is violated by some entangled strategy; there is no exception among tight inequalities.","Any XOR game with no quantum advantage defines a face whose codimension is bounded below by an explicit formula in the numbers of inputs, so the classical and quantum regions of the boundary are separated by at least one dimension.","Non-local computation games, the canonical examples of no-quantum-advantage games, are never facet-defining; the dimension bound generalises the earlier proof and shows the face is far from being a facet.","The convex set of quantum correlations has no nontrivial facets: every exposed face is low-dimensional except for the trivial inequalities $|c_{xy}|\\le 1$.","Because the codimension bound is strict and explicit, not only facets but also all faces of dimension above $D-\\Delta$ must carry a quantum violation, so the result applies to a quantitative neighbourhood of tightness."],"supporting_citations":[{"why":"Supplies the necessary and sufficient condition for an XOR game to have no quantum advantage, which the proof of Theorem 1 builds on.","marker":"[36]"},{"why":"Supplies the semidefinite-programming formulation and dual used to write the quantum bias and its complementary slackness conditions.","marker":"[37]"},{"why":"Establishes that quantum correlators are characterised by inner products of unit vectors, the representation used throughout the proof.","marker":"[12]"},{"why":"Introduces XOR games and their bias expressions, the operational setting of all the theorems.","marker":"[18]"},{"why":"Provides the first class of non-local computation games with no quantum advantage, whose tightness question the paper generalises.","marker":"[21]"},{"why":"Proves that non-local computation Bell inequalities are not facet-defining, a result the paper improves with explicit dimension bounds.","marker":"[25]"},{"why":"Gives the dimension of the no-signalling and Bell polytopes used to compute the facet codimension.","marker":"[26]"},{"why":"Provides the local-hidden-variable characterisation under which the Bell polytope is the convex hull of deterministic strategies.","marker":"[28]"}],"fun_headline_variants":["Tight Bell inequalities always yield quantum advantage","Every tight Bell test has a quantum win","Entanglement beats all tight classical correlations","No tight Bell inequality is classically optimal","Quantum advantage proven for all tight Bell games"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on every question in the game being asked with positive probability; if a question could be ignored, the argument that Bob's strategy is a fixed linear function of Alice's no longer goes through.","fun_headline_variants_meta":{"raw":{"variants":["Tight Bell inequalities always yield quantum advantage","Every tight Bell test has a quantum win","Entanglement beats all tight classical correlations","No tight Bell inequality is classically optimal","Quantum advantage proven for all tight Bell games"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000255,"raw_usage":{"total_tokens":1594,"prompt_tokens":992,"completion_tokens":602,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":608,"completion_tokens_details":{"reasoning_tokens":537}},"tokens_in":608,"tokens_out":602,"duration_ms":6937,"temperature":1.0,"reasoning_tokens":537,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:38:27.889506+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search for an exhaustive XOR game with $\\xi_Q=\\xi_c$ whose optimal classical face has dimension $D-1$ in the full Bell polytope, or dimension $m_A m_B$ in the correlation polytope; the theorem says neither can happen. A direct numerical check on small cases, say $m_A=2$, $m_B=3$, would enumerate all optimal sign vectors and compute the affine dimension of the face, then compare with the bound $\\frac{1}{2}m_A(m_A-1)$; if such a face exists, Theorem 1 fails.","supporting_citations":[{"cited_title":"Characterizing the Per- formance of XOR Games and the Shannon Capacity of Graphs","cited_arxiv_id":null,"evidence_quote":"Supplies the necessary and sufficient condition for an XOR game to have no quantum advantage, which the proof of Theorem 1 builds on."},{"cited_title":"Junge, M","cited_arxiv_id":null,"evidence_quote":"Supplies the semidefinite-programming formulation and dual used to write the quantum bias and its complementary slackness conditions."},{"cited_title":"Cirel’son","cited_arxiv_id":null,"evidence_quote":"Establishes that quantum correlators are characterised by inner products of unit vectors, the representation used throughout the proof."},{"cited_title":"Consequences and limits of nonlocal strate- gies","cited_arxiv_id":null,"evidence_quote":"Introduces XOR games and their bias expressions, the operational setting of all the theorems."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the first class of non-local computation games with no quantum advantage, whose tightness question the paper generalises."},{"cited_title":"Tightness of correlation inequalities with no quantum violation","cited_arxiv_id":null,"evidence_quote":"Proves that non-local computation Bell inequalities are not facet-defining, a result the paper improves with explicit dimension bounds."},{"cited_title":"Short, and Andreas Winter","cited_arxiv_id":null,"evidence_quote":"Gives the dimension of the no-signalling and Bell polytopes used to compute the facet codimension."},{"cited_title":"Almeida, Jean-Daniel Bancal, Nicolas Brunner, Antonio Ac´ ın, Nicolas Gisin, and Stefano Pironio","cited_arxiv_id":null,"evidence_quote":"Provides the local-hidden-variable characterisation under which the Bell polytope is the convex hull of deterministic strategies."}],"review_version":1}