{"id":"9162f359-7763-43f8-bda2-71e5f702ca84","arxiv_id":"2608.04973","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A quantum analogue of Bayes correlated equilibrium is defined and shown to satisfy: more informative quantum information shrinks the equilibrium outcome set in every game, with classical games recovered exactly.","lead":"This paper builds the quantum version of a central tool in game theory, the Bayes correlated equilibrium, where a mediator who sees the real state of the world privately recommends actions. It shows that more informative quantum information shrinks the set of equilibrium outcomes in every game, and that the ordering is computable by semidefinite programming.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the unconstrained realisability premise is load-bearing but is justified by the canonical expansion of Theorem 8.3, so ACCEPT stands.","rationale":"I independently checked the proofs that carry the central claims: the Loewner characterization (Theorem 6.2), the exact classical embedding (Theorem 7.1), the expansion theorem (Theorem 8.3), the non-instrumentality of the mediator (Proposition 8.5), and the soundness of quantum individual sufficiency (Theorem 9.4). The mathematics is internally consistent, and the two theorems named in the reader's strongest claim do follow from the stated definitions. The one genuinely load-bearing assumption is that any positive decomposition of ρ_ω into components ρ^a_ω counts as a decision rule (Definition 5.1, Lemma 5.2). This is what makes the transfer step in Theorem 9.4 possible: the coarser structure can reproduce the finer outcome by applying T_ω to the components, with no feasibility condition beyond positivity and trace preservation. The reader is right that this premise is structural and that Proposition 8.5 shows the mediator is not a quantum instrument on the players' systems. Nevertheless, the premise is not ad hoc. Theorem 8.3 and Corollary 8.4 show that every QBCE rule is realized by the truthful profile of an explicit expansion, where the mediator prepares a state conditional on ω. That is a physical preparation procedure, not an instrument applied to systems that already exist. This is the quantum analogue of the omniscient mediator in Bergemann-Morris, and it is exactly what keeps the theory incentive-based rather than feasibility-based. I therefore find no concrete error or unjustified step that would change the reader's ACCEPT verdict. The open general converse of Theorem 9.4 and the minor spectrahedron/projected-spectrahedron overstatement in the abstract are scope and presentation issues, not correctness risks.","tokens_in":26038,"tokens_out":25874,"duration_ms":329021,"concrete_test":"Instantiate Theorem 8.3 for the Proposition 8.5 example: for the rule ρ^0_ω=1_{ω=1}ρ_ω and ρ^1_ω=1_{ω≠1}ρ_ω, write out the canonical expansion ρ*_ω=∑_a ρ^a_ω⊗|a⟩⟨a| and verify that the truthful profile is optimal against every split deviation, hence against every POVM on H_i⊗E_i by Lemma 8.2. If the expansion is a valid classical-quantum state and the obedience inequalities (5) hold, the unconstrained realisability premise is physically grounded and the reader's concern does not land.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The only load-bearing modelling premise is the unconstrained realisability of decision rules (Definition 5.1 and Lemma 5.2), exactly as the reader notes. Step 1 of Theorem 9.4 uses it to transfer the finer rule to the coarser structure: once T_ω(ρ_ω)=ρ'_ω is imposed, the components ρ'^a_ω:=T_ω(ρ^a_ω) are automatically a valid decision rule, so the outcome is reproduced without any additional feasibility constraint. If one instead required {ρ^a_ω} to be a quantum instrument on the players' systems, Proposition 8.5 shows this would fail, and the soundness proof would not go through. This is not an internal inconsistency, however. A decision rule is not an operation applied to pre-existing systems; it is a mechanism-design object. Theorem 8.3 realises every QBCE rule as the truthful quantum Bayes Nash equilibrium of the canonical expansion ρ*_ω=∑_a ρ^a_ω⊗|a⟩⟨a|, which is an ordinary state preparation conditional on ω, and Corollary 8.4 extends this to arbitrary POVM deviations. This mirrors the omniscient mediator in classical BCE, where join feasibility is deliberately absent. The residual caveats are scope, not correctness: the converse of the comparison theorem is proved only in the two classes of Theorems 9.10-9.11, and the abstract overstates the outcome set as a spectrahedron rather than a projected spectrahedron. Neither affects Theorem 9.4 or Theorem 7.1.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a quantum generalization of Bergemann and Morris's Bayes correlated equilibrium. A quantum information structure is a state-dependent ensemble {ρ_ω} with one subsystem per player; a decision rule splits each ρ_ω into positive operators indexed by action profiles, and obedience requires each player's recommended action to be optimal against arbitrary POVM deviations on their subsystem. The paper proves that obedience is equivalent to Loewner domination of obedience operators (Theorem 6.2), making the set of equilibrium decision rules a spectrahedron and the outcome set a projected spectrahedron (Corollary 6.3), with nonemptiness by state-wise Nash play (Proposition 6.4). It then proves exact classical recovery (Theorem 7.1), the degeneracy of an advice model (Theorem 7.2), an incentive-triviality criterion in terms of local flatness (Theorem 7.4), a quantum analogue of the classical expansion theorem (Theorem 8.3), and a comparison order, quantum individual sufficiency, defined through positive trace-preserving maps (Definition 9.2). The main soundness result (Theorem 9.4) states that quantum individual sufficiency implies inclusion of equilibrium outcome sets in every basic game; converses are proved for simultaneously diagonalizable structures (Theorem 9.10) and for comparisons with the null structure (Theorem 9.11). The paper closes with a depolarizing-chain example computed by semidefinite programming.","tokens_in":26143,"tokens_out":19897,"duration_ms":218219,"significance":"This is a substantial and carefully executed contribution. I checked the central proofs and found them correct: Theorem 6.2's two-directional POVM argument is valid; Theorem 7.1's pinching reduction is sound; Theorem 8.3's expansion argument and Lemma 8.2's block-diagonal reduction are rigorous; Theorem 9.4 transfers obedience through positive maps exactly as claimed; Theorems 9.10 and 9.11 reduce to the classical Bergemann--Morris theorem where applicable. The paper is a conservative extension of the classical BCE theory in the precise sense of Theorem 7.1, and the comparison order is inhabited by local garblings, shared-randomness, and shared-entanglement garblings (Proposition 9.6). I particularly value the transparent treatment of the omniscient mediator: Lemma 5.2 makes realizability free, and Theorem 8.3 justifies this by realizing every QBCE decision rule as a quantum Bayes Nash equilibrium of a canonical expansion, while Proposition 8.5 honestly records that the mediator is not an instrument on the players' systems. The limitation that the converse of Theorem 9.4 is established only for two classes of structures is stated explicitly.","major_comments":[],"minor_comments":[{"comment":"The outcome set QBCE(G,Q) is described as a \"spectrahedron\"; Corollary 6.3 establishes only that the set of decision rules is a spectrahedron and that the outcome set is its linear image, i.e. a projected spectrahedron. Please correct the wording in the abstract and in Section 1.2.","section":"Abstract and Section 1.2"},{"comment":"The sentence \"A quantum decision rule X_i^{a_i,b_i} is a quantum Bayes correlated equilibrium\" misstates the object: the object whose equilibrium status is being defined is the family {ρ^a_ω}, not an individual obedience operator. Please fix this wording.","section":"Definition 5.4"},{"comment":"The text contains the typo \"an decomposition\"; it should read \"a decomposition.\"","section":"Section 2.2, contribution (1)"},{"comment":"The numerical results in Figure 1 are reported without code or a reproducibility statement. Since the paper's analytic claims do not depend on these values, this is a presentation matter, but a code/data supplement would be helpful.","section":"Section 10"},{"comment":"The expression for the deviation gain is printed as \"q(a)/2 √2\"; if the intended value is q(a)/(2√2), please parenthesize the denominator to avoid ambiguity.","section":"Example 7.3"}],"recommendation":"minor_revision","confidential_remarks":"The paper is well within the scope of quantum game theory and information economics, and I see no concerns about citation patterns or novelty disclosure. The only substantive caveat is that the general converse of the comparison theorem is left open, but the paper is explicit about this and the soundness half is the central contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. This is the first quantum analogue of the Bergemann-Morris ordering that actually works, and the paper earns its claims. The core construction - QBCE with a Loewner characterization of obedience - is genuinely new, and the proofs hold up on inspection. I checked the main threads: Theorem 6.2 (POVM quantification reduces to finitely many Loewner inequalities), Theorem 7.1 (exact classical embedding via pinching), Theorem 8.3 (canonical expansion realizes every QBCE rule as a truthful QBNE), and Theorem 9.4 (soundness of the comparison order). The arguments are complete and internally consistent.\n\nThe literature division in Section 2 is sharp: it correctly identifies that the belief-invariant convention in quantum game theory forecloses the comparison question, and the embedding theorem shows the quantum theory is a conservative extension of BM. The transposition result (Theorem 9.1) is a genuinely nice insight, and the decision to use positive maps rather than channels is well-justified.\n\nThe main caveat is scope rather than correctness. The converse of the comparison theorem is proved only for simultaneously diagonalizable structures and for comparisons with the null structure; the general converse is open. The paper is explicit about this, but it means the \"quantum analogue\" is not yet a full equivalence theorem. Two minor issues: the abstract calls the equilibrium set a spectrahedron, while the paper proves it is a projected spectrahedron (the outcome set is a linear image of a spectrahedron of decision rules). And the numerical example reports SDP values without shipping code or data - reproducible in principle, but not in practice from the text alone.\n\nThe load-bearing modeling assumption - that any POVM decomposition is a valid decision rule - is exactly what the paper claims it is. The expansion theorem (8.3) gives it a clean foundation: the mediator is not a physical operation on the players' systems, but the equilibria are realized as ordinary state-preparation protocols conditional on omega in the canonical expansion. The stress-test note is right that this is the quantum mirror of the omniscient classical mediator, and it is not an internal inconsistency. It does mean QBCE is an incentive concept, not a feasibility concept, which is the right scope for a comparison theorem.\n\nThis deserves a serious referee and will likely be accepted after minor revision. I would cite it in my own work on quantum information economics.","headline":"A rigorous, mostly complete quantum analogue of the Bergemann-Morris ordering; the math is sound, and the open converse is a scope limitation, not a flaw.","tokens_in":26845,"tokens_out":2488,"would_cite":true,"duration_ms":28789,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","91A27","91A80","90C22","94A15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs quantum Bayes correlated equilibrium, proves obedience is a Loewner operator inequality, and shows that under quantum individual sufficiency a more informative structure always induces a weakly smaller equilibrium…","keywords":["quantum Bayes correlated equilibrium","quantum information structures","games of incomplete information","Loewner order","semidefinite programming","quantum individual sufficiency","comparison of information structures","obedience constraints"],"falsifier":"Search numerically over two-player, two-state, two-action games with random qubit structures $Q,Q'$ and explicitly chosen maps $T_\\omega,\\Phi_i$ satisfying Definition 9.2; if any such pair has an outcome in $\\operatorname{QBCE}(G,Q)$ that is not in $\\operatorname{QBCE}(G,Q')$, then Theorem 9.4 is false. The same search can also test the weaker form of $T_\\omega$ positivity on the interval $[0,\\rho_\\omega]$, which the paper claims suffices; a failure there would localize the problem in the transfer step.","tokens_in":25637,"feed_emoji":"🎲","tokens_out":9264,"duration_ms":102155,"temperature":0.7,"pith_summary":"Games of incomplete information are governed by what the players learn; this paper claims that when the learning is quantum, the same logic as the classical Bayes correlated equilibrium theory works, with one change: obedience becomes an operator inequality instead of a scalar one. The paper builds a quantum information structure as a family of density operators indexed by the payoff state, lets an omniscient mediator split each state into action recommendations, and defines quantum Bayes correlated equilibrium by requiring that no player can gain from any measurement on their own subsystem. Its central theorems say that the equilibrium set is a nonempty convex spectrahedron computable by semidefinite programming, that classical information structures are recovered exactly, and that a more informative quantum structure always induces a weakly smaller equilibrium set in every basic game. A reader should care because this turns a comparison of information structures into a computable incentive question and shows that the classical comparison extends to quantum information without generating spurious outcomes.","feed_headline":"Quantum info shrinks equilibrium sets in every game","feed_subtitle":"One operator inequality makes quantum Bayes correlated equilibrium computable, and classical games recover exactly.","key_machinery":"The carrying object is the obedience operator\n$$$X_i^{{a_i,b_i}}$ = \\sum_{\\omega,a_{-i}} \\psi(\\omega) u_i(b_i,a_{-i},\\omega)\\,\\operatorname{Tr}_{-i}\\!\\left[\\rho_\\$omega^{{(a_i,a_{-i}}$)}\\right],$$\nwith Theorem 6.2 saying that the rule $\\{\\rho_\\omega^a\\}$ is a quantum Bayes correlated equilibrium iff $X_i^{a_i,a_i} - X_i^{a_i,b_i} \\succeq 0$ for all $i,a_i,b_i$ in the Loewner order. This replaces the continuum of POVMs in the obedience definition with finitely many linear matrix inequalities, and it is the source of the spectrahedron structure and semidefinite-programming computability. The comparison result is carried by positive trace-preserving maps $T_\\omega$ between full ensembles and one $\\omega$-independent positive map $\\Phi_i$ per player satisfying the identity $\\operatorname{Tr}_{-i}[T_\\omega(\\xi)] = \\Phi_i(\\operatorname{Tr}_{-i}[\\xi])$ for $\\xi$ supported on $\\operatorname{supp} \\rho_\\omega$; positivity is exactly what transports the Loewner inequalities from the finer to the coarser structure.","core_discovery":"This paper's central claim is that a quantum version of Bayes correlated equilibrium is well-defined and behaves like the classical object. A quantum information structure is a family of density operators $\\rho_\\omega$, one per payoff state, and a mediator decision rule is a splitting $\\rho_\\omega = \\sum_a \\rho_\\omega^a$ into positive operators indexed by action profiles. The paper proves that obedience to the mediator's recommendation against arbitrary POVM deviations is equivalent to the Loewner domination $X_i^{a_i,a_i} \\succeq X_i^{a_i,b_i}$ for every player and every alternative action, where the Hermitian operators $X_i^{a_i,b_i}$ are built from partial traces of the decision rule weighted by payoffs. From this single inequality the equilibrium set becomes a nonempty compact spectrahedron; classical structures embed exactly; the advice model is degenerate; and quantum individual sufficiency, defined through $\\omega$-dependent positive trace-preserving maps plus $\\omega$-independent local positive maps satisfying a partial-trace compatibility identity, implies inclusion of equilibrium sets in every game.","pith_inferences":["The paper does not draw out the following consequence: if mediator realizability were later required to be a physically implementable quantum instrument, the soundness transfer in Theorem 9.4 would likely collapse, because its proof uses splittings that no physical operation can realize.","The same state-correlated mediator construction suggests a neighbouring problem: extending this comparison to games with continuous action spaces or infinite-dimensional systems, where the Loewner characterization may require spectral side conditions beyond compactness.","A testable extension is to compare the QBCE welfare frontier against classical BCE welfare for the same basic game and the same classical signal accuracy; the paper's worked depolarizing chain already exhibits monotone decline, and systematic comparison would show when quantum coherence genuinely changes the set of robust predictions."],"forward_implications":["If $Q \\succeq_{QIS} Q'$, then $\\operatorname{QBCE}(G,Q) \\subseteq \\operatorname{QBCE}(G,Q')$ for every basic game and every number of players, so more quantum information never enlarges the set of equilibrium outcomes.","Computing equilibrium outcomes, membership, and optimal welfare under a linear objective is a semidefinite program, so information design in this setting inherits the full algorithmic toolbox of SDP.","Classical signals embed exactly: $\\operatorname{QBCE}(G,\\iota(S)) = \\operatorname{BCE}(G,S)$, meaning the quantum extension produces no new equilibrium outcomes when the information structure is purely classical.","The model in which players only receive classical advice extracted from their quantum systems is degenerate: its equilibrium set is the null-structure Bayes correlated equilibrium set and does not depend on the quantum structure at all.","For a single decision-maker, the maximum equilibrium payoff equals the full-information optimum for every structure; additional information never improves the best outcome, it only deletes outcomes elsewhere."],"supporting_citations":[{"why":"Supplies the classical Bayes correlated equilibrium concept and the classical comparison theorem that the paper quantizes.","marker":"[4]"},{"why":"Gives the single-agent quantum statistical model comparison that sets the boundary condition for the quantum ordering.","marker":"[10]"},{"why":"Establishes the one-player experiment-comparison order that the multi-player relation must reduce to.","marker":"[7]"},{"why":"Provides the belief-invariant quantum equilibrium framework with state-independent shared advice that Definition 5.1 deliberately removes.","marker":"[2]"},{"why":"Draws the distinction between direct quantum access and classical advice that reappears as QBCE versus QBCE_pp.","marker":"[1]"},{"why":"Documents the local-hidden-variable shared-advice form used in prior Bayesian quantum games, which Section 4.1 contrasts with state-correlated structures.","marker":"[9]"}],"fun_headline_variants":["Quantum information structures shrink equilibrium sets in all games","One Loewner inequality captures quantum correlated equilibrium","Quantum Bayes correlated equilibrium is computable via SDP","Quantum mediator obedience reduces to a single operator domination","Quantum individual sufficiency shrinks equilibrium sets in every game"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that every family of positive operators summing to the state operator is a legitimate mediator decision rule, even though the paper itself shows such a rule need not correspond to a physically realizable quantum operation on the players' systems; without that freedom the soundness transfer from one structure to another would fail.","fun_headline_variants_meta":{"raw":{"variants":["Quantum information structures shrink equilibrium sets in all games","One Loewner inequality captures quantum correlated equilibrium","Quantum Bayes correlated equilibrium is computable via SDP","Quantum mediator obedience reduces to a single operator domination","Quantum individual sufficiency shrinks equilibrium sets in every game"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000158,"raw_usage":{"total_tokens":1176,"prompt_tokens":850,"completion_tokens":326,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":466,"completion_tokens_details":{"reasoning_tokens":252}},"tokens_in":466,"tokens_out":326,"duration_ms":4300,"temperature":1.0,"reasoning_tokens":252,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T12:34:19.221589+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search numerically over two-player, two-state, two-action games with random qubit structures $Q,Q'$ and explicitly chosen maps $T_\\omega,\\Phi_i$ satisfying Definition 9.2; if any such pair has an outcome in $\\operatorname{QBCE}(G,Q)$ that is not in $\\operatorname{QBCE}(G,Q')$, then Theorem 9.4 is false. The same search can also test the weaker form of $T_\\omega$ positivity on the interval $[0,\\rho_\\omega]$, which the paper claims suffices; a failure there would localize the problem in the transfer step.","supporting_citations":[{"cited_title":"Bayes correlated equilibrium and the comparison of information structures in games.Theoretical Economics, 11(2):487–522, 2016","cited_arxiv_id":null,"evidence_quote":"Supplies the classical Bayes correlated equilibrium concept and the classical comparison theorem that the paper quantizes."},{"cited_title":"Comparison of quantum statistical models: equivalent conditions for sufficiency.Communi- cations in Mathematical Physics, 310(3):625–647, 2012","cited_arxiv_id":null,"evidence_quote":"Gives the single-agent quantum statistical model comparison that sets the boundary condition for the quantum ordering."},{"cited_title":"Equivalent comparisons of experiments.The Annals of Mathematical Statistics, 24(2):265–272, 1953","cited_arxiv_id":null,"evidence_quote":"Establishes the one-player experiment-comparison order that the multi-player relation must reduce to."},{"cited_title":"Belief-invariant and quantum equilibria in games of incomplete information.Theoretical Computer Science, 895:151–177, 2021","cited_arxiv_id":null,"evidence_quote":"Provides the belief-invariant quantum equilibrium framework with state-independent shared advice that Definition 5.1 deliberately removes."},{"cited_title":"Abbott, Mehdi Mhalla, and Pierre Pocreau","cited_arxiv_id":null,"evidence_quote":"Draws the distinction between direct quantum access and classical advice that reappears as QBCE versus QBCE_pp."},{"cited_title":"Connection between Bell nonlocality and Bayesian game theory.Nature Communications, 4:2057, 2013","cited_arxiv_id":null,"evidence_quote":"Documents the local-hidden-variable shared-advice form used in prior Bayesian quantum games, which Section 4.1 contrasts with state-correlated structures."}],"review_version":1}