{"id":"fcbc6d20-1d6e-4a01-898c-b312e456b6e5","arxiv_id":"2607.25954","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A new map built from reshuffling and partial transposition produces four-partite states with exactly equal bipartition entropies, approaching maximal entanglement in large local dimension.","lead":"This paper introduces a linear map that forces the three balanced bipartitions of a four-partite quantum state to have exactly equal entanglement entropies, without requiring maximal entanglement. The map gives a simple way to construct 'equi-entropic' states and shows that random inputs produce nearly maximal common entropy in high dimensions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equi-entropy theorem is proven, but the claim that typical outputs are near-maximally entangled rests on uncontrolled moment approximations and lacks a concentration bound.","rationale":"The reader's weakest_assumption correctly identifies the central load-bearing concern. I independently checked Theorem 1's proof: the index manipulations using the cyclic invariance Y^{RΓ}=Y and Y^{ΓR}=Y do show YY† = Y^R(Y^R)† = Y^Γ(Y^Γ)†, so the equi-entropy claim is solid. The vulnerability is Proposition 2, where Eq. (46) is unshown and Eq. (47) replaces the expectation of a ratio with the ratio of expectations, potentially introducing bias. Moreover, the lack of a variance bound means the paper's language about 'typical' outputs is not justified. The paper itself flags the omitted fourth-moment expansion, so this is a self-admitted limitation. Since the central theorem stands and the asymptotic claim is plausible but unproven, the conditional verdict is appropriate. The proposed numerical test directly probes whether the mean is representative, which would settle whether the concern actually lands.","tokens_in":11387,"tokens_out":18937,"duration_ms":146363,"concrete_test":"For d ∈ {3,4,5,6,8,10,16}, draw N=10^4 independent Haar-random matrices X ∈ U(d²), compute Y=Ξe(X) via Eq. (20), and record S(Y). Compute the sample mean, sample standard deviation σ_d, and the fraction of samples with S(Y) ≤ 1 - c/d for a fixed c (e.g., c=0.1). If σ_d decays as O(d^{-2}) (or at least faster than the mean deficit) and the distribution is concentrated around the predicted 1 - 1/d², the 'typical' claim holds. If σ_d ∼ O(1/d) or O(1), the average is not representative and Proposition 2's interpretation as typical behavior fails. Also compare the sample mean to 1 - 1/d² to assess the bias from replacing the expectation of the ratio.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The core mathematical result, Theorem 1, is rigorously proven: for any nonzero Y=Ξe(X) or Y=Ξo(X), the identity YY† = Y^R(Y^R)† = Y^Γ(Y^Γ)† ensures S(Y)=S(Y^R)=S(Y^Γ). The load-bearing weakness lies in Proposition 2, which supports the abstract and summary statements that Haar-random inputs produce 'highly entangled outputs' with average entropy deficit O(d^{-2}). The derivation has two uncontrolled steps: (i) Eq. (46) for E[Tr((YY†)^2)] is asserted as a leading-order result with the full fourth-moment Haar expansion omitted (the text says 'The complete fourth-moment expansion is straightforward but lengthy'); (ii) Eq. (47) replaces the expectation of the ratio Tr((YY†)^2)/Tr^2(YY†) by the ratio of expectations. No variance or concentration bound is given. If the fluctuations of S(Ξe(X)) decay more slowly than the mean deficit d^{-2}, then many individual outputs could have entropy substantially below the predicted average, contradicting the 'typical outputs are near-maximal' narrative. Since the summary explicitly states 'typical high-dimensional inputs make the common value close to maximal,' this gap is directly load-bearing for the advertised probabilistic claim. The theorem itself is not threatened, but the quantitative asymptotic claim lacks rigorous support.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces two linear maps Ξ_e and Ξ_o on d²×d² matrices, obtained as the even and odd subsequence limits of iterating X_{k+1} = (X_k^R + X_k^Γ)/2, where R and Γ are reshuffling and partial transposition. Theorem 1 proves that every nonzero output has equal normalized linear entropy across the three balanced bipartitions, and Proposition 1 establishes convergence with explicit formulas. The authors characterize fixed points, relate two-unitary fixed points to orthogonal Latin squares, propose Conjecture 1 about centrosymmetric two-unitary permutation fixed points when d ≢ 2 (mod 3), and state Proposition 2 that for Haar-random unitary inputs the average entropy deficit is O(d^{-2}), supported by a leading-moment calculation and numerical simulations.","tokens_in":11636,"tokens_out":8507,"duration_ms":84244,"significance":"The core construction is clean, explicit, and parameter-free: Ξ_e and Ξ_o are group-theoretic projections onto cyclic and coset subspaces, and the stronger identity YY† = Y^R(Y^R)† = Y^Γ(Y^Γ)† established in the proof of Theorem 1 is a useful structural result. The equi-entropy property is exact and fully proven, and the combinatorial examples with MOLS and two-unitary permutation fixed points are valuable. The main weakness is that the advertised probabilistic claim — that typical high-dimensional Haar-random inputs yield near-maximal common entropy — rests on an uncontrolled moment approximation and lacks a concentration statement.","major_comments":[{"comment":"The asymptotic claim 1−E[S(Ξ_e(X))] ∼ d^{-2} is not proven as stated. Eq. (46) asserts the leading fourth-moment term with the full Haar expansion omitted ('The complete fourth-moment expansion is straightforward but lengthy'), and Eq. (47) replaces E[ratio] by E[numerator]/E[denominator] without justification. No variance or concentration bound is given, so neither 'typical' nor even high-probability near-maximality follows from the average. Since the abstract and Sec. IV explicitly say that 'typical high-dimensional inputs make the common value close to maximal', this gap is load-bearing. Either provide a complete Weingarten-based moment computation including a variance/concentration estimate, or explicitly demote Proposition 2 to a heuristic and soften the 'typical' statements. Fig. 2 (256 samples, standard errors only) does not resolve the fluctuation issue.","section":"III.B.3, Proposition 2 (Eqs. 46–50)"}],"minor_comments":[{"comment":"The sentence 'The condition d ≢ 2 (mod 3) is necessary. Otherwise, no such permutation can exist. This follows from ... analysis of invariants' gives no proof. Since the statement is used to delimit the conjecture, either supply the argument or label the necessity as part of the conjecture rather than as an established fact.","section":"III.B.2, Conjecture 1"},{"comment":"The displayed equation for ξ^{2k} contains a corrupted arrow/glyph ('/leftr⫯g⊸tl⫯ne→'); the mathematical meaning is clear, but the typesetting needs correction.","section":"III.A, Proposition 1 proof"},{"comment":"The figure caption reports sample means over 256 independent Haar-random unitaries with standard error bars, but no individual variances or ranges are shown. Since the claim concerns 'typical' behavior, reporting a measure of the spread of S(Ξ_e(X)) across samples would be informative.","section":"III.B.3, Fig. 2"},{"comment":"The replacement of the expectation of a ratio by the ratio of expectations is introduced with the symbol ≃ but without any discussion of its accuracy. Even if the fourth-moment term were exact, this step would remain uncontrolled.","section":"III.B.3, Eq. (47)"}],"recommendation":"major_revision","confidential_remarks":"I agree with the conditional verdict: Theorem 1 is solid and the construction is interesting, but the advertised 'typical near-maximal' claim is not supported rigorously. The fix is within scope — either a complete moment computation with concentration or an honest reframing as a heuristic asymptotic prediction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Core take: the paper does what it says. The map Ξ and its two parity branches are new, and Theorem 1 is a real theorem: every nonzero output has exactly equal normalized linear entropies across the three balanced bipartitions, because the Gram matrices YY†, Y^R(Y^R)†, and Y^Γ(Y^Γ)† coincide. The proof is short and correct. The group-theoretic interpretation and the fixed-point analysis are also clean. The connection to orthogonal Latin squares is a nice bonus, and Conjecture 1 is plausible given the supporting data.\n\nWhere it gets softer: Proposition 2. The claim that Haar-random inputs yield average entropy deficit ~d^{-2} rests on two uncontrolled steps. Equation (46) asserts the leading fourth moment without showing the expansion—the paper says it is \"straightforward but lengthy\"—and Eq. (47) replaces the expectation of a ratio by the ratio of expectations. There is no concentration bound, so the statement that \"typical\" high-dimensional inputs are close to maximal is not established beyond the mean. That said, the paper is transparent: it calls this a leading-moment approximation, and the numerics (256 samples, standard errors) support the mean. So this is a soft spot, not a fatal one. The theorem itself is unaffected.\n\nA minor issue: the paper gives no application, and the authors are upfront about that. The significance is moderate—a useful tool for constructing equi-entropic states, not a resolution of a major open problem.\n\nOverall, the math claimed as proven is proven. The asymptotic is a well-labeled estimate. I would send it to a referee, asking either for the full fourth-moment calculation or for the asymptotic claim to be tempered to a numerical observation. The referee should not reject on the strength of the gap, but should require that the authors close it or label it more carefully.","headline":"A clean new map that provably equalizes bipartition entropies, with a d^{-2} average deficit that is a well-labeled leading-moment estimate rather than a proven theorem.","tokens_in":12152,"tokens_out":3084,"would_cite":true,"duration_ms":29718,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper introduces a linear map built from reshuffling and partial transposition that forces the three balanced bipartitions of any four-party state to have exactly equal linear entropies, and shows that for random inputs this common ent","keywords":["equi-entropic map","linear entropy","reshuffling","partial transposition","absolutely maximally entangled states","two-unitary matrices","orthogonal Latin squares","Haar-random unitary"],"falsifier":"Compute the exact Haar fourth moment for Y = Ξ_e(X) for a small fixed dimension such as d = 3 or 4 using the full Weingarten expansion; if the predicted deficit 1 − E[S(Y)] deviates from 1/d² at order larger than O(d⁻⁴), or if a numerical sampling over many Haar-random unitaries shows that the variance of S(Y) decays slower than d⁻⁴, then the claim that typical outputs are near-maximal for large d would be undermined. A separate falsifier for Conjecture 1 is to search for a centrosymmetric two-unitary permutation fixed point of Ξ_e at d = 5 or d = 8; if none exists, the conjecture is false.","tokens_in":11229,"feed_emoji":"⚛️","tokens_out":3807,"duration_ms":34662,"temperature":0.7,"pith_summary":"The paper asks whether entanglement can be made uniformly distributed across the three balanced bipartitions of a four-party system without requiring it to be maximal. It answers with a linear map Ξ that, applied to any nonzero d²×d² matrix, guarantees exact equality of the normalized linear entropies of the three bipartitions. The map arises as the limit of an iterative averaging procedure alternating reshuffling and partial transposition. For Haar-random unitary inputs, the paper's leading-moment analysis predicts that the common entropy approaches its maximal value as d grows, with an average deficit of order d⁻². The paper also shows that special fixed points of Ξ are two-unitary matrices corresponding to exact AME(4,d) states, and connects permutation fixed points to orthogonal Latin squares.","feed_headline":"Averaging reshuffles equalizes all three bipartition entropies","feed_subtitle":"The common entropy is close to maximal for random inputs, approaching 1 as local dimension grows.","key_machinery":"The central object is the linear map Ξ, defined by averaging over a subgroup of the group G generated by the two index rearrangements reshuffling (R) and partial transposition (Γ). The group has six elements, decomposing into the cyclic subgroup G_e = {I, RΓ, ΓR} and its complementary coset G_o = {R, Γ, RΓR}. The key mechanism is that Y = Ξ_e(X) is invariant under cyclic index permutations (Y^{RΓ} = Y, Y^{ΓR} = Y), which forces the Gram matrices YY†, Y^R(Y^R)†, and Y^Γ(Y^Γ)† to coincide. Since the normalized linear entropy depends only on the squared singular values of the matrix, this spectral identity yields exact entropy equality across all three balanced bipartitions.","core_discovery":"Theorem 1: For any matrix X ∈ C^{d²×d²} with d ≥ 2, if Y = Ξ_e(X) or Y = Ξ_o(X) is nonzero, then the three balanced bipartition entropies are equal: S(Y) = S(Y^R) = S(Y^Γ). The proof establishes a stronger identity: YY† = Y^R(Y^R)† = Y^Γ(Y^Γ)†, so the three reduced states have identical spectra. The even and odd limiting maps are defined as Ξ_e(X) = (X + X^{RΓ} + X^{ΓR})/3 and Ξ_o(X) = (X^R + X^Γ + X^{RΓR})/3, obtained as the convergent subsequences of the iteration X_{k+1} = (X_k^R + X_k^Γ)/2. Proposition 2 states that for Haar-random unitary X, the leading-moment approximation gives 1 − E[S(Ξ_e(X))] ∼ 1/d² as d → ∞, supported by numerical simulations.","pith_inferences":["Since the proof establishes equality of the full Gram spectra, not just a scalar entropy, the map actually enforces equal entanglement spectra across the three bipartitions; this stronger 'spectrum bundling' could be exploited in tasks sensitive to the entire reduction spectrum, such as certain quantum error-correction or teleportation protocols.","The asymptotic claim about near-maximal entropy is proven only at the level of averaged leading moments; whether typical individual outputs are near-maximal hinges on concentration of the entropy around its mean, which the paper does not address. A natural extension is to compute the variance of S(Ξ_e(X)) over Haar-random unitaries to verify that fluctuations decay faster than d⁻⁴.","The group-theoretic construction generalizes naturally: choosing other subgroups of index permutations for systems with more than four parties could produce maps that equalize entropies across arbitrary selected bipartitions, offering a systematic tool for designing multipartite entanglement with prescribed symmetry.","The authors' suggestion of a continuous-variable extension could be tested by defining the analog of index reshuffling on four-mode Gaussian states and checking whether balancing symplectic purities leads to a similar 'equi-purity' property under energy constraints."],"forward_implications":["Any nonzero output of Ξ provides a four-party state with exactly equal bipartition entropies, offering a way to construct states with symmetric entanglement without solving the two-unitary existence problem.","Two-unitary fixed points of Ξ_e correspond to AME(4,d) states, so the fixed-point equation gives a systematic search strategy for absolutely maximally entangled states.","For permutation matrices, the fixed-point condition imposes a cyclic symmetry on associated orthogonal Latin squares; Conjecture 1 predicts such centrosymmetric two-unitary permutations exist for all d not ≡ 2 (mod 3) and d ≠ 6.","For Haar-random inputs, the average entropy deficit of order d⁻² implies the reduced states are O(d⁻¹) close to maximally mixed in Hilbert-Schmidt distance, giving a quantitative sense of near-AME behavior in high dimensions.","The iterative procedure converges exponentially fast, so the limiting equi-entropic map can be approximated in practice by a few dozen averaging steps."],"fun_headline_variants":["Averaging trick forces equal entanglement across all three splits","New map makes four-partite states perfectly balanced","Random inputs yield near-maximal entanglement via simple averaging","Entropy equalization achieved for any four-partite state","Almost maximal entanglement from averaging bipartitions"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claim that Haar-random inputs produce near-maximal common entropy rests on approximating the average of a ratio by the ratio of averages and keeping only the leading Gaussian term in the Haar fourth-moment expansion, with no bound on how much individual outputs fluctuate around the mean.","fun_headline_variants_meta":{"raw":{"variants":["Averaging trick forces equal entanglement across all three splits","New map makes four-partite states perfectly balanced","Random inputs yield near-maximal entanglement via simple averaging","Entropy equalization achieved for any four-partite state","Almost maximal entanglement from averaging bipartitions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000182,"raw_usage":{"total_tokens":1159,"prompt_tokens":769,"completion_tokens":390,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":513,"completion_tokens_details":{"reasoning_tokens":326}},"tokens_in":513,"tokens_out":390,"duration_ms":3730,"temperature":1.0,"reasoning_tokens":326,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T01:00:19.908636+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the exact Haar fourth moment for Y = Ξ_e(X) for a small fixed dimension such as d = 3 or 4 using the full Weingarten expansion; if the predicted deficit 1 − E[S(Y)] deviates from 1/d² at order larger than O(d⁻⁴), or if a numerical sampling over many Haar-random unitaries shows that the variance of S(Y) decays slower than d⁻⁴, then the claim that typical outputs are near-maximal for large d would be undermined. A separate falsifier for Conjecture 1 is to search for a centrosymmetric two-unitary permutation fixed point of Ξ_e at d = 5 or d = 8; if none exists, the conjecture is false.","supporting_citations":[],"review_version":1}