{"id":"7d5c1802-5849-42fc-b625-d0059fab4ba4","arxiv_id":"2508.17140","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A two-measurement steering inequality based on robustness of imaginarity detects steerable two-qubit states using only four state parameters and resists more noise than the three-measurement NAQC and NAQI criteria.","lead":"This paper derives a new test for quantum steering, the ability of one party to control another's quantum state, using only two measurement settings per side and four numbers from the shared two-qubit state. The test survives more noise than two earlier partial-information tests, which could make steering detection experiments simpler.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the central steering certificate survives the apparent convexity issue in Eq (10).","rationale":"I read the central claim as Theorem 2: for any bipartite qubit state, I2(ρAB) > √2 certifies that the assemblage is not LHS, using only n1, n2, t11, t22. The proof of this claim is internally consistent. The apparent weakness in Eq (10) is real as written—equality does not hold for mixtures, only subadditivity—but the argument requires only ≤, so the inequality (15) survives unchanged. The witness construction is also sound: I2 is the support function of a finite set of Pauli-pair observables, so a violating state always has some W^k_{i,j} with Tr[Wρ] > √2, giving a negative witness. I found no counterexample or hidden assumption that breaks the steering certificate. The reader's conditional verdict is reasonable because of several fixable presentation and proof-completeness issues, but I do not regard those as load-bearing for the main result; hence no verdict adjustment is needed. My disagreement with the reader's weakest_assumption is therefore deliberate: the identified convexity step is not the point on which the central argument rests.","tokens_in":24421,"tokens_out":33895,"duration_ms":354777,"concrete_test":"Independently re-derive inequalities (13)–(15) without using the equality in Eq (10), starting instead from subadditivity of the extended trace-norm measure IR(Σ_i M_i) ≤ Σ_i IR(M_i) for unnormalized conditional states; verify that the bound Σ_λ p(λ)[I_x(ρ_λ)+I_y(ρ_λ)] ≤ √2 then follows exactly. If the chain holds, the LHS bound underpinning the central criterion is sound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The weakest point usually cited is Eq (10), which writes IR(σ_a|A) = Σ_λ p(λ)p(a|A,λ) IR(ρ_λ) as if robustness of imaginarity were additive under convex mixtures; in general only subadditivity is guaranteed. This is not load-bearing for the central claim, because the subsequent chain (11)–(15) needs only an inequality in that direction. Replacing equality (10) with ≤, we get Σ_a p(a|A) I_basis(ρ_{a|A}) = Σ_a IR(σ_{a|A}) ≤ Σ_λ p(λ) I_basis(ρ_λ), and summing over the two basis choices yields Σ_λ p(λ)[I_x(ρ_λ)+I_y(ρ_λ)] ≤ √2 via the complementarity relation (9). I also checked the four-parameter formula (21)/(24) against the conditional-state Bloch expressions, and the witness family (35)–(40) is complete as a finite set of subgradients: the maximum over the sixteen Pauli-pair expectation values equals max(|n1|,|t11|)+max(|n2|,|t22|)=I2(ρAB). The remaining issues—the equality-versus-subadditivity wording, the numerical constant in Lemma 2's Eq (33), and the omitted algebra in Theorem 4's monogamy proof—are presentation or proof-completeness defects, not defects in the steering certificate itself.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces an imaginarity steering inequality (ISI) for bipartite qubit states. By exploiting the robustness of imaginarity and a complementarity relation between two mutually unbiased bases, the authors derive a steering bound of √2 for local hidden state models. They express the left-hand side I2(ρAB) in terms of only four parameters (n1, n2, t11, t22), construct witness operators whose negative expectation values detect violations, illustrate the criterion on Werner states, X states, and MEMS, prove a monogamy inequality for pure tripartite states, and compare noise robustness and unsharp-measurement tolerance with NAQC and NAQI criteria.","tokens_in":24535,"tokens_out":10439,"duration_ms":109140,"significance":"The central steering certificate is attractive because it uses only two measurement settings per party and a partial tomographic description of the state. I verified the key algebraic reductions: the LHS bound in Eq. (15), the four-parameter formula in Eqs. (21)-(24), the Werner and X-state evaluations, and the equivalence I2 = 2λ under unsharp measurements are internally consistent. The witness family in Eqs. (35)-(40) is also a valid subgradient construction: its maximum over the Pauli-pair expectations equals I2(ρAB). If the proof gaps identified below are repaired, this would be a useful addition to the steering literature.","major_comments":[{"comment":"The equality I_R(σ_{a|A}) = Σ_λ p(λ)p(a|A,λ) I_R(ρ_λ) is not valid for the robustness of imaginarity: the trace norm is subadditive, not additive, so only ≤ holds in general. The subsequent derivation needs only the ≤ direction, so the steering bound in Eq. (15) survives, but the manuscript must replace the equality by an inequality and adjust the accompanying text, which currently attributes additivity to homogeneity of the norm.","section":"Section III B, Eq. (10)"},{"comment":"The norm estimate ∥Σ_{i,j}[(σ_i⊗I)+(I⊗σ_i)+(σ_i⊗σ_j)]∥ = √15/2 is not justified and appears numerically incorrect. The continuity of I2 can be proven by a simpler direct Lipschitz bound because n1, n2, t11, t22 are linear functions of ρ and | |a|-|b| | ≤ |a-b|, but the current proof should be corrected before the compactness argument for the separating witness is accepted.","section":"Section IV, Lemma 2, Eqs. (33)-(34)"},{"comment":"The proof jumps from the explicit expressions in Eqs. (62)-(65) to the claimed inequality I2(ρAB)+I2(ρAC) ≤ 2√2 without supplying the required algebra or optimization. The assertion that the maximum is attained at η0 = 1/√2, η1 = η4 = 0, η2 = η3 = 1/2 is stated but not demonstrated. Since monogamy is a headline result, the derivation must be completed or replaced by a valid argument.","section":"Section V, Theorem 4, Eq. (66)"}],"minor_comments":[{"comment":"There is a typo: 'unhsarp' should be 'unsharp'.","section":"Abstract"},{"comment":"The notation I_R(ρ_{λ_max}) is used before λ_max is defined; please define it explicitly.","section":"Section III B, Eq. (11)"},{"comment":"The kets such as |0x0x⟩ are not defined; please add a sentence explaining the tensor-product shorthand, e.g. |ab⟩ = |a⟩⊗|b⟩.","section":"Section IV C, Eqs. (54)-(57)"},{"comment":"The notation I2(ρAB) = { ±n1±n2, ... } is nonstandard; it should be written as the maximum over the displayed expressions, since the witness construction relies on I2 = max(|n1|,|t11|) + max(|n2|,|t22|).","section":"Section IV A, Eq. (35)"},{"comment":"The sentence 'The same bounds are obtained if Alice performs unsharp measurements with parameter v' is confusing because v denotes visibility in the table; please clarify whether the table's entries apply to the sharpness parameter λ.","section":"Section VI A, Table I"},{"comment":"Theorem 1 correctly uses subadditivity and homogeneity, which is inconsistent with the equality claimed in Eq. (10); harmonizing these derivations would improve the paper.","section":"Section III B, Eq. (10) versus Theorem 1, Eq. (18)"}],"recommendation":"major_revision","confidential_remarks":"The core steering inequality and its four-parameter certificate appear sound, and I found no fundamental obstacle to the main claim. However, the false equality in Eq. (10), the unsupported norm estimate in Lemma 2, and the missing algebra in Theorem 4 are all located in main results. These are fixable, but they require substantive revision rather than copyediting."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is better than its modest packaging. The central object is a two-measurement steering inequality built from robustness of imaginarity, with a closed four-parameter form I2 = 1/2(|n1−t11|+|n1+t11|+|n2−t22|+|n2+t22|) > √2. That is genuinely not in the cited literature: NAQC and NAQI are three-measurement criteria, and CFFW needs the full correlation matrix. I checked the chain from the LHS bound to the Werner and X-state evaluations; it is internally consistent. The witness construction is also a real addition, and the monogamy bound is plausible. The claim that real states give I2 ≤ 1 is correct, so the criterion genuinely probes imaginarity.\n\nThe best practical feature is efficiency: two settings per side and four measured parameters, with better Werner-state noise tolerance than the three-measurement rivals. That should make it useful for steering certification and semi-device-independent tasks.\n\nNow the soft spots, in order of severity. First, Eq (10) states an equality for robustness of imaginarity under convex mixtures, where only subadditivity is guaranteed. The stress-test is right that this is not load-bearing: the derivation needs only ≤, and the subsequent cascade gives the √2 bound. But the text should be corrected, because as written it is a false statement.\n\nSecond, Lemma 2's continuity proof has a wrong numerical constant: the bound on ||ρ−ρ0|| is not √15/2 c. This is a small error, but it is visible and easily fixable.\n\nThird, Theorem 4's monogamy proof omits the actual algebra. The parameterized state is written, the conditional imaginarities are listed, and then the sum is asserted to be ≤ 2√2. The omitted step is exactly the part that needs to be checked; as it stands the proof is a claim, not a demonstration.\n\nFourth, the noise comparison is slanted. They compare against NAQC and NAQI but leave out CFFW, which for Werner states detects standard steering up to v > 0.5. Their v > 0.707 bound is worse than the standard steering threshold, and they should say so. The paper does note at the end that general steerability does not imply imaginarity steerability, which is the right caveat, but the comparison table should include CFFW or at least mention the standard steering threshold.\n\nFinally, no experiment or numerical simulation is provided, which is fine for a theory paper, but it means the claimed experimental advantage is asserted rather than demonstrated.\n\nVerdict: conditional acceptance with minor-to-moderate revision. The central certificate survives scrutiny; the weaknesses are presentation and completeness, not load-bearing math. I would send this to a serious referee. It is the kind of paper that gets better with one careful round of revision, and the four-parameter criterion could well become a standard tool.","headline":"Candid take: the 2-measurement imaginarity steering inequality is a real and useful variant of NAQC/NAQI, the central algebra checks out, but the paper has proof-presentation gaps and an incomplete noise comparison.","tokens_in":25309,"tokens_out":2350,"would_cite":true,"duration_ms":22425,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.65.Ud","03.67.Mn","03.65.Ta"],"model":"deepseek-v4-flash","headline":"Imaginarity can certify quantum steering from just four real parameters of a two-qubit state, via a new steering inequality that any local hidden state model must satisfy, with violation detected by simple Pauli witness operators.","keywords":["quantum steering","imaginarity","robustness of imaginarity","steering inequality","witness operator","monogamy","two-qubit states","mutually unbiased bases"],"falsifier":"Run a semidefinite-programming feasibility search for a local-hidden-state model on the four-parameter family with $I_2(\\rho_{AB})>\\sqrt{2}$; the theorem is settled by whether any such model reproduces the $x/y$ measurement statistics, and a positive example for the visibility-mixed singlet at $v=0.75$ would directly contradict the paper's prediction.","tokens_in":24013,"feed_emoji":"⚛️","tokens_out":13654,"duration_ms":131401,"temperature":0.7,"pith_summary":"This paper claims that the part of a two-qubit density matrix that changes under transposition — its imaginarity — is enough to certify quantum steering, and that detecting this requires less information than earlier steering tests. The authors derive an imaginarity steering inequality, $I_2(\\rho_{AB}) \\le \\sqrt{2}$, that must hold for every local-hidden-state model, so a bipartite qubit state violating it lets Alice steer Bob's local imaginarity. The certificate uses only four of the state's fifteen real parameters, and the violation is read off from witness operators built from Pauli measurements in two bases per side. The paper also proves a monogamy trade-off and shows greater noise tolerance than two existing three-measurement partial-information steering criteria.","feed_headline":"Imaginarity certifies steering with only four state parameters","feed_subtitle":"Imaginarity inequality uses two settings per side and tolerates more noise than earlier partial-information tests.","key_machinery":"The engine is the imaginarity-steering functional $I_2(\\rho_{AB})$ together with the qubit complementarity relation $I^x_R(\\rho)+I^y_R(\\rho)\\le\\sqrt{2}$ for the robustness of imaginarity $I_R(\\rho)=\\tfrac12\\|\\rho-\\rho^T\\|_1$. The complementarity bound is applied to Bob's conditional states under Alice's $x$- and $y$-measurements; averaging over the hidden states in a local-hidden-state model converts it into the steering inequality. Convexity and compactness of the free set of states then turn the inequality into witness operators built from the Pauli terms $\\mathbb{1}\\otimes\\sigma_x$, $\\mathbb{1}\\otimes\\sigma_y$, $\\sigma_x\\otimes\\sigma_x$, and $\\sigma_y\\otimes\\sigma_y$, so a violation is read off as a negative expectation value.","core_discovery":"The paper's central claim is Theorem 2: for any bipartite qubit state, Alice can steer Bob's local imaginarity whenever $$I_2(\\rho_{AB})=\\tfrac12\\big(|n_1-t_{11}|+|n_1+t_{11}|+|n_2-t_{22}|+|n_2+t_{22}|\\big)>\\sqrt{2}.$$ The quantity is derived from a complementarity relation for the robustness of imaginarity in two mutually unbiased bases, $I^x_R(\\rho)+I^y_R(\\rho)\\le\\sqrt{2}$, applied to Bob's conditional states. The proof shows every separable state satisfies the inequality, and it constructs witness operators $\\widetilde{W}^k_{i,j}=\\sqrt{2}\\,\\mathbb{1}_4-W^k_{i,j}$ whose negative expectation values flag the violation, establishing that no local hidden state model can reproduce the two-measurement statistics. The paper also proves the monogamy trade-off $I_2(\\rho_{AB})+I_2(\\rho_{AC})\\le 2\\sqrt{2}$ for tripartite pure states, and compares the inequality with two earlier partial-information steering criteria, finding a larger noise tolerance for the visibility-mixed singlet family.","pith_inferences":["Because the witness needs only two bases per side and four parameters, a direct estimator that extracts these marginals from raw coincidence counts without reconstructing the full density matrix is a natural next step; the paper does not construct such an estimator.","The inequality is sufficient but not necessary for steering, so there is a gap between the region detected by $I_2$ and the full steerable region; mapping that gap could show how much steering is missed by partial-information tests.","The monogamy trade-off suggests a quantitative bound for one-sided device-independent tasks such as key distribution or self-testing, since all maximally entangled states saturate the maximal violation; quantifying how the violation magnitude converts into a security parameter is not done here."],"forward_implications":["Any two-qubit state with $I_2(\\rho_{AB})>\\sqrt{2}$ is steerable from Alice to Bob, so steering can be certified from four parameters instead of full state tomography.","The witness operators require only spin measurements in the $x$- and $y$-bases, with eight local projectors in the decomposition given in the paper.","In a tripartite pure state, Alice cannot simultaneously steer the imaginarity of both Bob and Charlie; one of the two reduced states must satisfy the inequality.","For white-noise-mixed singlet states the violation starts at $v>1/\\sqrt{2}$, below the $v>0.815$ and $v>0.745$ thresholds quoted for the coherence-based and imaginarity-based three-measurement tests; the same $1/\\sqrt{2}$ threshold applies to unsharp measurement sharpness."],"supporting_citations":[{"why":"Introduces the robustness of imaginarity, the resource measure used to quantify Bob's local imaginarity and to build the complementarity relation.","marker":"[15]"},{"why":"Gives the closed trace-norm form $I_R(\\rho)=\\tfrac12\\|\\rho-\\rho^T\\|_1$ and the operational properties used to bound hidden-state averages.","marker":"[17]"},{"why":"Establishes imaginarity as a convex resource monotone, justifying its treatment as a quantifiable resource in the steering argument.","marker":"[16]"},{"why":"Supplies the 2-2-2 necessary and sufficient steering criterion that the imaginarity inequality extends and is compared with.","marker":"[26]"},{"why":"Provides the three-measurement nonlocal-advantage-of-coherence inequalities used as the main comparison for noise and unsharpness robustness.","marker":"[40]"},{"why":"Provides the three-measurement nonlocal-advantage-of-imaginarity inequalities whose noise threshold is compared in Table I.","marker":"[41]"},{"why":"Defines the mutually unbiased bases in $C^2$ on which the imaginarity complementarity relations rest.","marker":"[42]"},{"why":"Supplies the visibility-mixed singlet family on which the white-noise and unsharpness thresholds are computed.","marker":"[43]"},{"why":"Provides the general two-qubit parametrization from which the four-parameter expression for $I_2(\\rho_{AB})$ is extracted.","marker":"[51]"}],"fun_headline_variants":["Imaginarity steering beats prior tests with fewer settings","Two measurements per side certify imaginarity steering","Monogamous imaginarity steering tolerates higher noise","Complex numbers drive an efficient steering inequality"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that when Alice's cheating strategy averages hidden states, the imaginarity of the averaged state is no larger than the weighted average of the imaginarities of the individual hidden states; if mixing could concentrate imaginarity, the $\\sqrt{2}$ ceiling on $I_2$ would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Imaginarity steering beats prior tests with fewer settings","Two measurements per side certify imaginarity steering","Monogamous imaginarity steering tolerates higher noise","Complex numbers drive an efficient steering inequality"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000647,"raw_usage":{"total_tokens":2969,"prompt_tokens":939,"completion_tokens":2030,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":555,"completion_tokens_details":{"reasoning_tokens":1969}},"tokens_in":555,"tokens_out":2030,"duration_ms":16806,"temperature":1.0,"reasoning_tokens":1969,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:10:37.691801+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a semidefinite-programming feasibility search for a local-hidden-state model on the four-parameter family with $I_2(\\rho_{AB})>\\sqrt{2}$; the theorem is settled by whether any such model reproduces the $x/y$ measurement statistics, and a positive example for the visibility-mixed singlet at $v=0.75$ would directly contradict the paper's prediction.","supporting_citations":[{"cited_title":"Experimental Masking of Real Quantum States","cited_arxiv_id":"2107.01589","evidence_quote":"Gives the closed trace-norm form $I_R(\\rho)=\\tfrac12\\|\\rho-\\rho^T\\|_1$ and the operational properties used to bound hidden-state averages."},{"cited_title":"Mondal, T","cited_arxiv_id":null,"evidence_quote":"Provides the three-measurement nonlocal-advantage-of-coherence inequalities used as the main comparison for noise and unsharpness robustness."},{"cited_title":"Oppenheim and S","cited_arxiv_id":null,"evidence_quote":"Provides the three-measurement nonlocal-advantage-of-imaginarity inequalities whose noise threshold is compared in Table I."},{"cited_title":"Pati and P","cited_arxiv_id":null,"evidence_quote":"Defines the mutually unbiased bases in $C^2$ on which the imaginarity complementarity relations rest."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the visibility-mixed singlet family on which the white-noise and unsharpness thresholds are computed."},{"cited_title":"Girolami, Observable measure of quantum coherence in finite dimensional systems, Phys","cited_arxiv_id":null,"evidence_quote":"Provides the general two-qubit parametrization from which the four-parameter expression for $I_2(\\rho_{AB})$ is extracted."}],"review_version":2}