{"id":"d62f312f-3cb3-480a-846a-bb6dd3224747","arxiv_id":"2608.10783","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"B can steer A relative to a pure state exactly when a state-preserving conditional expectation maps B' onto A.","lead":"This paper proves that, in quantum systems with infinitely many degrees of freedom, one party can steer the other party's states precisely when a certain algebraic bridge, called a state-preserving conditional expectation, exists between the relevant operator algebras. The result ties a core quantum information concept to subfactor theory and settles an open question about two-way steering.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified. The central equivalence is supported by a standard commutant Radon–Nikodym step and the stated pure-state hypothesis.","rationale":"The reader correctly identifies Corollary 8 and the commutant Radon–Nikodym theorem as the load-bearing bridge between the operational steering condition and the algebraic ensemble-extension property. I agree that this is the step on which the whole theorem depends, but I do not regard it as a weakness: the theorem is standard, the pure-state hypothesis is exactly what is needed, and the mixed-state generalization is handled separately. I also checked the passage from ensemble extensions to conditional expectations: Corollary 12 applies because A is a factor, so the central support of ω_A is 1, and the corner-extension arguments in Lemmas 10 and 11 supply the required state-preserving conditional expectation. The proof of the auxiliary final implication in Theorem B contains a minor wording issue about cyclicity, but the underlying claim is correct once 'cyclic for B' replaces 'cyclic in B′'. Since no counterexample or internally inconsistent step emerged, the reader's ACCEPT verdict should stand unchanged.","tokens_in":28449,"tokens_out":44077,"duration_ms":470349,"concrete_test":"Verify Corollary 8 by direct computation in a finite-dimensional example with an environment: take H=H_A⊗H_B⊗H_E, A=B(H_A)⊗1⊗1, B=1⊗B(H_B)⊗1, and a pure Ω with a correlated A–E marginal. Construct a nontrivial ensemble on A, extend it to B′ via the formula in Corollary 8, run the commutant Radon–Nikodym construction of Lemma 7 to obtain operators in B, and check that they sum to 1 and reproduce the ensemble. This tests the non-faithful vector-state case of the load-bearing step.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the proof chain, I find no load-bearing gap in the central claim. The key step is Corollary 8, which converts steering into the ensemble-extension property. Its converse uses the commutant Radon–Nikodym theorem: for the vector state Ω on B′, every positive functional dominated by ω_B′ is represented by an operator in (B′)′=B. This is standard and remains valid without faithfulness; the pure-state hypothesis is precisely what makes the 'if' direction work, and the mixed-state case is explicitly reduced to the pure case in Proposition 14. The equivalence (b)⇔(c) then follows from Corollary 12, whose hypotheses are met because A is a factor. The only textual caveat I noticed is in the bonus final statement of Theorem B: '[BΩ] is a cyclic projection in B′' should read that p≤[BΩ] is cyclic for B, i.e., [B pΩ]=p, which is true and sufficient. This does not affect the main equivalence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that, for commuting factors A and B on a Hilbert space and a pure state vector Omega, the operational property 'B can steer A relative to Omega' is equivalent to the existence of a normal conditional expectation E: B' -> A preserving the marginal state omega_B'. The proof proceeds by first establishing a general lifting theorem (Theorem 3) for state-preserving normal unital positive maps between JBW*-algebras: ensemble lifts exist iff the map has a state-preserving left inverse, equivalently a factorization through a conditional expectation onto the multiplicative domain. A commutant Radon-Nikodym argument (Corollary 8) converts steering into an ensemble-extension property on B', and Corollary 12 converts the ensemble-extension property into the desired conditional expectation. Additional results characterize two-way steering for tomographically complete systems in terms of finite index and Haag duality (Theorem C), and reduce mixed-state steering to the pure case (Proposition 14). An appendix gives a self-contained proof of the Kadison-Schwarz inequality for JB-algebras.","tokens_in":28568,"tokens_out":33037,"duration_ms":309596,"significance":"If correct, the paper establishes a precise and non-obvious equivalence between a central concept in quantum information (steering) and a central tool in subfactor theory (state-preserving conditional expectations). The general lifting theorem for JBW*-algebras is a strong standalone result, and the paper's organizational structure, including a dependency graph, makes the proof chain transparent. The paper is careful about the pure-state hypothesis and explicitly records the mixed-state reduction. The proof of the main equivalence appears sound; the only flaw I found is in the proof of the additional 'bonus' statement of Theorem B, which is local and correctable.","major_comments":[{"comment":"The proof of the last statement of Theorem B claims that s_B' = s_A e = [B Omega] is a cyclic projection in B', and that therefore every projection p in B' with p <= s_B' is cyclic in B'. This claim is false in general. For example, let H = C^2 tensor C^3, A = B(C^2) tensor 1, B = 1 tensor B(C^3), and let Omega be a vector of full Schmidt rank; then A and B are commuting factors with A vee B = B(H), A = B', and condition (c) holds with E = id. Here s_B' = [B Omega] = 1, but 1 is not cyclic in B': for every xi in H, [B' xi] = H_A tensor S_B(xi) with dim S_B(xi) <= 2 < 3, so [B' xi] is never the identity. The desired conclusion of the bonus statement can be recovered by working with cyclicity for B = (B')' (any p <= [B Omega] is cyclic for B with generating vector p Omega, and Murray-von Neumann equivalence preserves this), but the argument as written needs to be corrected. Since the final statement is part of Theorem B, this requires a revision.","section":"Section 3.1, proof of Theorem B (final statement)"}],"minor_comments":[{"comment":"In the definition of beta = alpha-hat^{-1} composed with E composed with P_supp omega plus (1 - supp phi) * omega, please spell out that the second term is the map x maps to (1 - supp phi) omega(x); the notation can otherwise be misread as a product of elements.","section":"Section 2.2, proof of Theorem 3"},{"comment":"After Eq. (47), the step from s_A' = s_B in B to A = B' uses that s_B = 1 because B is a factor; this should be stated explicitly.","section":"Section 3.2, proof of Proposition 19"},{"comment":"The dependency diagram in Fig. 2 is very useful, but the font is small; consider enlarging it for readability.","section":"Introduction and Figure 2"}],"recommendation":"major_revision","confidential_remarks":"The central equivalence is sound and the paper is a good fit for the journal. The faulty cyclicity claim in the proof of the final statement of Theorem B does not affect the main equivalence, but it is a genuine error in the proof of a stated theorem and should be fixed before publication; the correction appears straightforward. No concerns about attribution or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: the main theorem is new and the proof structure holds up. The paper shows that for commuting factors A, B and a pure state Ω, B steering A is equivalent to the existence of an Ω-preserving normal conditional expectation from B' onto A. That is a real and interesting bridge between quantum steering and subfactor theory. Theorem C adds a clean two-way characterization: finite index for two possibly different reference states, Haag duality only for a single pure state. The lifting theorem (Theorem D) for positive maps between JBW*-algebras looks like a solid generalized tool, and the appendix proves the Kadison-Schwarz inequality for JB-algebras from scratch, which the authors correctly note they could not find in the literature.\n\nThe proofs are readable for people comfortable with standard operator-algebraic machinery. I checked the dependency chain from Theorem B down: steering ⇔ ensemble extension (Lemma A/Corollary 8) uses the commutant Radon-Nikodym theorem in the standard way, and the factor hypothesis gives the central support condition for Proposition E/Corollary 12. The stress-test note about the cyclic-projection phrase is a harmless typo, not a gap.\n\nSoft spots are minor. The abstract states the pure-state equivalence without the factor hypothesis, but Theorem B needs A and B to be factors; that should be fixed in a revision. Several standard theorems (Takesaki, Haagerup's operator-valued weights, Wright's theorem) are invoked without proof, which is normal in this field. The Kadison-Schwarz appendix was generated with LLM assistance, and although the argument looks elementary and correct, it deserves independent checking before publication. Nothing here is load-bearing.\n\nThis paper is for people working on entanglement in algebraic quantum field theory or on subfactor theory; it also gives a nice state-based complement to Takesaki's theorem. I would send it to a serious referee.","headline":"A genuine operator-algebraic characterization of steering; solid proofs, minor presentational gaps.","tokens_in":29129,"tokens_out":2783,"would_cite":true,"duration_ms":27230,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L10","46L37","46L60","81P40","17C65"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that, for pure global states, quantum steering by a commuting subsystem is equivalent to the existence of a state-preserving conditional expectation from the commutant onto the subsystem algebra.","keywords":["quantum steering","conditional expectation","von Neumann algebras","subfactor theory","Haag duality","ensemble lifting","JBW*-algebras","Jones index"],"falsifier":"Exhibit commuting factors $A, B \\subset B(H)$ and a state vector $\\Omega$ such that $B$ can steer $A$ relative to $\\Omega$ but no $\\Omega$-preserving normal conditional expectation $B' \\to A$ exists; Theorem B declares this combination impossible. The paper itself flags the irreducible inclusion $R_\\infty \\subset R_\\infty \\rtimes \\mathbb{R}$ as admitting no normal conditional expectation, so a representation of that inclusion with a steering state vector would refute the equivalence.","tokens_in":28224,"feed_emoji":"🔗","tokens_out":11401,"duration_ms":104041,"temperature":0.7,"pith_summary":"The paper proves that, in infinite-dimensional quantum systems, the operational ability of one subsystem to steer another is a structural fact about the pair of operator algebras, not a detail of the state's tomography. For two commuting subsystems $A$ and $B$ in a joint pure state, $B$ can steer $A$—can realize every statistical mixture of $A$'s states by some measurement on $B$—exactly when there is a state-preserving conditional expectation from the commutant $B'$ back onto $A$. The result connects quantum steering to subfactor theory, and it shows that tomographic completeness no longer guarantees steering once the two algebras stop being exact commutants of each other (failure of Haag duality). The authors also characterize two-way steering: both sides steering each other relative to the same state is equivalent to Haag duality plus purity, while steering in both directions with different reference states is equivalent to finite subfactor index. The core argument is a general lifting theorem for ensembles under positive maps between Jordan algebras, which gives a state-based analogue of Takesaki's classical criterion for conditional expectations.","feed_headline":"Quantum steering equals state-preserving conditional expectations","feed_subtitle":"Algebra inclusions decide which side can steer which—not tomography.","key_machinery":"The load-bearing object is the $\\Omega$-preserving conditional expectation $E : B' \\to A$, a normal unital positive map that fixes $A$ pointwise and reproduces the global state's marginal on $B'$ from the marginal on $A$ ($\\omega_{B'} = \\omega_A \\circ E$); this is precisely a state-preserving left inverse of the inclusion $A \\hookrightarrow B'$. The technical engine that turns steering into such a map is the ensemble-lifting theorem (Theorem D): for any state-preserving normal unital positive map $\\alpha : (N,\\varphi) \\to (M,\\omega)$ between JBW*-algebras, every ensemble on $N$ with average $\\varphi$ lifts to one on $M$ with average $\\omega$ if and only if $\\alpha$ has a state-preserving left inverse, shown via Petz dual maps, self-polar forms, and multiplicative domains; applied to an inclusion this yields Proposition E, and applied to $A \\hookrightarrow B'$ together with the commutant Radon–Nikodym realization of steering (Corollary 8) it yields Theorem B.","core_discovery":"For commuting factors $A, B \\subset B(H)$ with a state vector $\\Omega$, the paper's main theorem states that three conditions are equivalent: (a) $B$ can steer $A$ relative to $\\Omega$—every ensemble of states of $A$ with average $\\omega_A$ can be produced by a measurement (POVM) in $B$; (b) every ensemble on $A$ with average $\\omega_A$ extends to an ensemble on $B'$ with average $\\omega_{B'}$; (c) there is an $\\Omega$-preserving normal conditional expectation $E : B' \\to A$, i.e., $\\omega_{B'} = \\omega_A \\circ E$. Steering is thus identical, for pure global states, to the existence of a state-preserving conditional expectation, an object of subfactor theory. The same circle of ideas yields a characterization of two-way steering under tomographic completeness: both sides can steer each other relative to a single state exactly when $A = B'$ (Haag duality) and the state is pure, and relative to two possibly different states exactly when the inclusion $A \\subset B'$ has finite Jones–Kosaki–Longo index. Along the way the authors prove a general ensemble-lifting theorem for state-preserving positive maps between Jordan (JBW*-) algebras, which for an inclusion of von Neumann algebras says that the ensemble-extension property is equivalent to the existence of a state-preserving conditional expectation.","pith_inferences":["If the equivalence holds, the Jones index gains a direct operational meaning: the ability of two tomographically complete regions to steer each other with different reference states is exactly finite index, so index could in principle be probed through steering experiments in lattice models.","For the surface-code ground state, the paper's framework predicts full steering for disjoint-cone regions because the ground state is preserved by the conditional expectation; a finite-size numerical steering test in the toric code would be a ready check.","Because the lifting theorem works for positive maps between JBW*-algebras rather than completely positive maps between von Neumann algebras, the same steering characterization plausibly extends to probabilistic theories with Jordan-algebraic state spaces.","In rational conformal field theories, the Reeh–Schlieder property prevents the vacuum from being preserved by the conditional expectation, so the paper predicts a qualitative contrast: interval regions in the vacuum cannot be steered, whereas topologically ordered ground states can be."],"forward_implications":["Even when the two subsystems are tomographically complete ($A \\vee B = B(H)$), one-way steering can fail; the obstruction is the absence of a state-preserving conditional expectation $B' \\to A$.","If steering works relative to a single state vector with faithful marginal in an irreducible subfactor inclusion, then every state of $A$ admits a purification relative to which $B$ steers $A$.","One-way steering does not imply two-way steering; steering in both directions relative to the same state is equivalent to Haag duality $A = B'$ together with purity of the state.","Allowing different reference states for the two directions, two-way steering is equivalent to the subfactor inclusion $A \\subset B'$ having finite Jones–Kosaki–Longo index.","Every state-preserving conditional expectation is characterized in the Schrödinger picture by the ensemble-extension property on dominated functionals, complementing Takesaki's modular-flow criterion."],"supporting_citations":[{"why":"Establishes that non-uniqueness of purifications is equivalent to failure of Haag duality, the phenomenon this paper extends from purification to steering.","marker":"[8]"},{"why":"Introduces the subfactor index whose finiteness appears in the two-way steering characterization.","marker":"[22]"},{"why":"Provides Takesaki's modular-flow criterion for state-preserving conditional expectations, the classical theorem the paper's Schrödinger-picture characterization complements.","marker":"[39]"},{"why":"Posed the open question about the relation between steering and Haag duality that Theorem C answers.","marker":"[47]"},{"why":"Supplies the multiplicative-domain and Schwarz-inequality results for positive maps that structure the lifting proof.","marker":"[60]"},{"why":"Introduces self-polar forms, used to construct Petz dual maps for general (Jordan) algebras.","marker":"[63]"},{"why":"Develops self-polar forms and Tomita–Takesaki theory for JBW*-algebras, grounding the Jordan-generalized lifting theorem.","marker":"[64]"},{"why":"Gives Petz dual maps and sufficiency for channels over von Neumann algebras, the duality mechanism of Theorem D.","marker":"[67]"},{"why":"Supplies the Radon–Nikodym and cyclicity facts that convert ensembles on $B'$ into POVMs in $B$ for pure states.","marker":"[74]"},{"why":"Haagerup's operator-valued-weight theory links finite index to the existence of conditional expectations and their duals, used in Theorem C.","marker":"[77]"}],"fun_headline_variants":["Steering equals conditional expectations for pure states","Pure-state steering is a conditional expectation","Quantum steering: the state-preserving expectation link","Steering and subfactor theory: one equivalence"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The equivalence assumes the global state is pure: only then does the Radon–Nikodym theorem guarantee that every ensemble on $B'$ is realized by an actual measurement in $B$, while for mixed states an extra tensor-product decomposition is required.","fun_headline_variants_meta":{"raw":{"variants":["Steering equals conditional expectations for pure states","Pure-state steering is a conditional expectation","Quantum steering: the state-preserving expectation link","Steering and subfactor theory: one equivalence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000511,"raw_usage":{"total_tokens":2548,"prompt_tokens":1069,"completion_tokens":1479,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":685,"completion_tokens_details":{"reasoning_tokens":1423}},"tokens_in":685,"tokens_out":1479,"duration_ms":15435,"temperature":1.0,"reasoning_tokens":1423,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:26:40.564754+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit commuting factors $A, B \\subset B(H)$ and a state vector $\\Omega$ such that $B$ can steer $A$ relative to $\\Omega$ but no $\\Omega$-preserving normal conditional expectation $B' \\to A$ exists; Theorem B declares this combination impossible. The paper itself flags the irreducible inclusion $R_\\infty \\subset R_\\infty \\rtimes \\mathbb{R}$ as admitting no normal conditional expectation, so a representation of that inclusion with a steering state vector would refute the equivalence.","supporting_citations":[{"cited_title":"The Schmidt Rank for the Commuting OperatorFramework","cited_arxiv_id":null,"evidence_quote":"Posed the open question about the relation between steering and Haag duality that Theorem C answers."},{"cited_title":"Selfpolar forms and their applications to the C*-algebra theory","cited_arxiv_id":null,"evidence_quote":"Introduces self-polar forms, used to construct Petz dual maps for general (Jordan) algebras."},{"cited_title":"Tomita-Takesaki Theory for Jordan Algebras","cited_arxiv_id":null,"evidence_quote":"Develops self-polar forms and Tomita–Takesaki theory for JBW*-algebras, grounding the Jordan-generalized lifting theorem."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Radon–Nikodym and cyclicity facts that convert ensembles on $B'$ into POVMs in $B$ for pure states."},{"cited_title":"Operator valued weights in von Neumann algebras, II","cited_arxiv_id":null,"evidence_quote":"Haagerup's operator-valued-weight theory links finite index to the existence of conditional expectations and their duals, used in Theorem C."}],"review_version":1}