{"id":"8891e3be-cc78-4eb2-b7e0-d301a8ca126d","arxiv_id":"2607.01712","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Quantum relative entropy is the unique normalized, additive, Lorenz-continuous divergence monotone under binary guessing games.","lead":"The paper proves that any measure of quantum state distinguishability which respects binary guessing odds, adds under tensor products, and is continuous on classical inputs must be the Umegaki quantum relative entropy. This gives an operational reason why this entropy, rather than one of many alternatives, is the canonical measure of quantum state difference.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The classical MPST representation adaptation in S8 is the load-bearing step and is not proven from the paper's axioms.","rationale":"The reader's weakest assumption correctly identifies the finite-alphabet MPST adaptation in S8 as the place where the proof's external input is least secure. The paper's central claim depends on Eq. (37) in an essential way: without the classical Rényi-mixture representation, the quantum part of the proof cannot even be written, because the one-shot mixture form (38) is obtained by applying Theorem 3 to the classical mixture (37). The concern is not that the MPST theorem is false; it is that the paper substitutes its own Lorenz continuity for the continuity notion in [36] without showing that the substitution preserves the theorem's hypotheses. This is a verifiable, addressable gap rather than a refutation. Because the rest of the proof is structured and the missing verification is plausibly repairable, the appropriate verdict remains conditional, matching the reader's assessment. The independent computational checks in the qubit Lemma 12 are not enough to rescue the argument if the classical representation fails, which is why this concern is load-bearing rather than cosmetic.","tokens_in":27819,"tokens_out":10115,"duration_ms":95994,"concrete_test":"Obtain the exact statement and hypotheses of the main MPST theorem from [36] and check line-by-line whether every hypothesis is satisfied by a divergence that is normalized, additive, monotone under stochastic maps, and uniformly continuous on each sector L_C with respect to d_L. In particular, verify whether the paper's Lorenz continuity implies MPST's continuity notion; if not, construct an explicit functional satisfying the paper's axioms but not the MPST representation, or supply a proof that Lorenz continuity implies the MPST continuity condition. A clean side-check is to repeat the S8 endpoint-removal argument using the counterexample sequences of Lemma 10 to see whether any positive D_0 or D_infinity mass is truly incompatible with the paper's axioms without assuming the full symmetric representation first.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 4 collapses the full one-shot family only after the classical restriction is represented as a positive Rényi mixture, Eq. (37). That representation is imported from Mu, Pomatto, Strack and Tamuz [36] via an 'adaptation' in Supplementary Section S8. The adaptation assumes the original MPST theorem applies to any divergence that is normalized, additive, monotone under stochastic maps, and Lorenz continuous in the paper's sectorwise d_L sense. The paper does not prove that these conditions imply the hypotheses of [36], whose continuity assumption is stated in a different topology and whose derivation also covers both orientations of the pair. If the original theorem requires a stronger form of continuity, then a divergence satisfying Definition 2 need not have the representation (37), and the mixed Rényi form (38), the balance equation (44), and the final collapse to α=1 all lose their foundation. The reductions inside S8 correctly manipulate a symmetric representation, but they are conditional on that representation holding; the paper itself calls this a 'normalized adaptation' rather than a proof. This is the single most load-bearing concern because every later step, including Lemma 12, starts from the Rényi-mixture form forced by (37).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an axiomatic characterization of the Umegaki quantum relative entropy. It defines quantum Lorenz divergences as functionals monotone under quantum Lorenz majorization, which is an operational preorder based on binary guessing games, and proves two main results. Theorem 3 states that a classical divergence that is monotone under classical relative majorization and Lorenz continuous has a unique quantum Lorenz extension. Theorem 4 states that a normalized, additive quantum Lorenz divergence whose classical restriction is Lorenz continuous must equal Tr[rho log rho] - Tr[rho log sigma] for all finite-dimensional pairs with supp(rho) subset of supp(sigma). The proof combines the classical MPST representation of additive monotone divergences as mixtures of Renyi divergences with a qubit-based separation lemma that forces all mass of the mixture to concentrate at alpha = 1. The paper includes a lengthy supplementary information section with detailed proofs of the layer-cake representation, Lorenz continuity properties, the qubit gap computations, and the MPST adaptation.","tokens_in":28151,"tokens_out":15555,"duration_ms":135647,"significance":"If the main theorem is correct, it is a substantial result: it eliminates the infinite family of DPI-monotone quantum divergences at the single-shot level, under an operational distinguishability ordering and additivity, without assuming super-additivity. The proof is not circular: the Umegaki form is not assumed, and the conclusion follows from the axioms plus the imported MPST classification. The paper ships a detailed supplementary with explicit formulas, including the qubit perturbative expansion that underlies the collapse to alpha = 1. The claimed rigidity is also falsifiable, since any normalized additive Lorenz-continuous divergence violating Eq. (39) would constitute a counterexample. The main risk is that the classical classification step is imported from Mu, Pomatto, Strack, and Tamuz via an 'adaptation' rather than proved from the paper's own assumptions.","major_comments":[{"comment":"The finite-alphabet MPST representation is load-bearing and is not proved from the paper's axioms. Section S8 states, without proof, that every normalized, additive, monotone-under-stochastic-maps, Lorenz-continuous classical divergence admits the one-sided Renyi mixture (37). The original MPST theorem has different technical hypotheses, involving a symmetric representation, finiteness on bounded pairs, and a continuity/topology condition that is not obviously identical to the sectorwise Lorenz continuity of Definition 2. The paper rewrites the symmetric MPST formula and removes endpoint terms, but it never verifies that a divergence satisfying the paper's conditions satisfies all hypotheses of [36]. If the MPST theorem requires stronger regularity, Eq. (37) fails, and with it Eqs. (38), (42)-(44), and Theorem 4 lose their foundation. This needs to be either proved as a self-contained lemma or explicitly added as an axiom.","section":"S8 / Eq. (37)"},{"comment":"The abstract claims the result 'requires neither a thermodynamic limit of infinitely many copies', but the proof of Theorem 4 explicitly uses the infinite-copy regularization limit D(rho||sigma) = lim_{n->infinity} (1/n) D(rho^{otimes n}||sigma^{otimes n}) in Eq. (40), and Section S9 justifies interchanging this limit with the MPST integral. It is true that exact additivity makes Eq. (40) redundant for the value of D, but the derivation of Eq. (42) relies on the asymptotic regularization of HT Renyi divergences in Eq. (41). The abstract and Section V should be rephrased to say that no thermodynamic limit is assumed as an axiom, while acknowledging that an n-to-infinity regularization step is used in the proof. As written, the claim is misleading.","section":"Abstract and Eq. (40)"},{"comment":"The bracket construction in Lemma 11 states 'fix C > 2 Dmax(rho||sigma)' for the bounded sector. If this is literally the number 2 times the max-divergence, the condition is insufficient: the slopes of the quantum Lorenz curve are governed by the exponential of Dmax, e.g. C should be larger than a quantity like exp(Dmax) or 2^{Dmax}. For Dmax large, a literal C > 2 Dmax may fall below the maximal slope, invalidating the claim that the approximating classical pairs lie in L_C. Please clarify whether this is a typesetting error and correct the bound in the final version.","section":"Lemma 11 (SI S6)"}],"minor_comments":[{"comment":"There is a stray punctuation artifact in the abstract: 'quantum noncommutativity. collapses' should read 'quantum noncommutativity collapses', and the accented 'R\\'enyi' appears with broken markups.","section":"Abstract"},{"comment":"The notation D_alpha is used for both the one-shot HT Renyi divergence and the Petz Renyi divergence after regularization. In Eqs. (42)-(44) the two objects have different meanings; please introduce separate symbols (e.g., D_alpha^{HT}, D_alpha^{Petz}) to avoid confusion in the cancellation step.","section":"Eqs. (42)-(44)"},{"comment":"The supplementary material contains internal numbering remnants: '2 The General Case' appears as a subsection header inside S3, and some figure labels appear to come from an earlier draft. Please clean up the section and figure numbering.","section":"SI S3"},{"comment":"The Lorenz continuity condition is stated for the classical restriction on sectors L_C. It would help to state explicitly that D_cl is assumed to take finite values on each L_C, since uniform continuity with respect to d_L is otherwise ambiguous when infinite values are allowed.","section":"Definition 2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious mathematical contribution and the supplementary material is unusually detailed, with explicit computations. My main reservation is the unproved adaptation of the MPST theorem: it is the single load-bearing step on which the classical reduction rests, and the paper's own text describes it as an 'adaptation' rather than a derivation. I would ask the editor to require the author to provide either a self-contained proof of the S8 representation from Definition 2's continuity assumption or an explicit statement that this is an additional axiom. The abstract's 'no thermodynamic limit' phrasing also needs correction even if the mathematical result is sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the main theorem — additive quantum Lorenz divergences with a Lorenz-continuous classical restriction collapse to Umegaki relative entropy — is genuinely new. It is not a repackaging of Matsumoto or of Wilming–Gallego–Eisert; the qubit gap separation lemma is the real work, and the proof is technically serious. Second, the paper has a load-bearing dependence on an adaptation of the Mu–Pomatto–Strack–Tamuz classification in Supplementary Section S8. Everything downstream rests on Eq. (37), and that representation is not proven from the paper's axioms.\n\nWhat is good. The testing-region/Lorenz framing is clean and operationally motivated. The extension uniqueness theorem (Theorem 3) is a nice structural result in its own right and would stand even if the main theorem fell. The supplementary material is detailed, with explicit perturbation expansions and domination bounds. The paper is also honest in places: the discussion admits that only regularized additivity is needed at the quantum step, and S9 carefully justifies the limit-integral interchange with a uniform D_max bound.\n\nSoft spots, in proportion. The S8 adaptation of MPST is the biggest concern. The original theorem gives a symmetric representation with both orientations and requires finiteness on bounded pairs; the paper rewrites it as a one-sided Rényi mixture and argues endpoint masses vanish by Lorenz continuity. That reduction is plausible but not proven at the same level of rigor as the surrounding text. If the MPST hypotheses require stronger regularity than Definition 2 supplies, the representation (37) — and with it the mixed Rényi form (38), the balance equation (44), and the final collapse to α=1 — loses its foundation. This is exactly where a serious referee should push.\n\nThere is also a smaller inconsistency: the abstract claims no thermodynamic limit is needed, yet Eq. (40) is an infinite-copy regularization limit. The discussion does acknowledge that only regularized additivity is used, so this is an abstract-level overstatement rather than a hidden fatal flaw. It should be fixed.\n\nI have not verified every coefficient in Lemma 14, but the proof skeleton is coherent, the domination bounds are present, and the perturbative mechanism is credible. If Lemma 12 and the S8 adaptation both hold, the theorem is almost certainly correct.\n\nBottom line: this paper deserves peer review, not a desk reject. Send it to a referee who knows the MPST theorem and is willing to check the algebra in S7–S8. My expectation is a conditional accept after substantial revision, with the S8 gap closed.","headline":"Genuinely new uniqueness theorem for Umegaki relative entropy, with a serious proof and one load-bearing adaptation of MPST that needs independent verification before the result can be trusted.","tokens_in":28516,"tokens_out":1695,"would_cite":false,"duration_ms":16787,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","94A17"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that any normalized, additive measure of quantum distinguishability that is monotone under binary testing orders and continuous on classical inputs must be the Umegaki relative entropy.","keywords":["quantum relative entropy","Umegaki relative entropy","Lorenz majorization","binary state discrimination","Rényi divergence","additivity","data-processing inequality","uniqueness theorem"],"falsifier":"Take the two-parameter qubit family $\\rho_r=\\frac12\\begin{pmatrix}1&r\\\\r&1\\end{pmatrix}$ and $\\sigma_s=\\operatorname{diag}(\\frac{1+s}{2},\\frac{1-s}{2})$ from Supplementary Section S7, pick an order $\\alpha\\neq 1$, and numerically evaluate the one-shot layer-cake divergence (38) on tensor powers: if $D_\\alpha(\\rho_r^{\\otimes 2}\\|\\sigma_s^{\\otimes 2})=2D_\\alpha(\\rho_r\\|\\sigma_s)$ holds for some $\\alpha\\neq 1$ and some $r,s$, then additivity does not single out order one and Theorem 4 is false. Equivalently, a direct search for any normalized additive quantum Lorenz divergence with Lorenz-continuous classical restriction that differs from the Umegaki value on this family would refute the claim.","tokens_in":27614,"feed_emoji":"🎯","tokens_out":15673,"duration_ms":133936,"temperature":0.7,"pith_summary":"Quantum relative entropy is usually justified by the data-processing inequality, but that inequality alone permits an infinite family of distinguishability measures. The paper proposes a sharper operational order: one pair of states is more distinguishable than another only if it offers at least as good winning odds in every binary guessing game, for every prior bias. Its main theorem states that any measure which is monotone under this order (a quantum Lorenz divergence), additive under tensor products, normalized, and has a continuous classical restriction must be exactly the Umegaki relative entropy, $D(\\rho\\|\\sigma)=\\operatorname{Tr}[\\rho\\log\\rho]-\\operatorname{Tr}[\\rho\\log\\sigma]$. Classically the same axioms still allow a full simplex of Rényi mixtures; the collapse to a single measure is driven by quantum noncommutativity in tensor powers. If correct, relative entropy is not merely an asymptotic quantity but the unique additive single-shot distinguishability measure compatible with binary state discrimination.","feed_headline":"Binary guessing games pick out quantum relative entropy","feed_subtitle":"Additivity plus a sharper data-processing rule collapses all Rényi divergences to Umegaki's D(ρ∥σ).","key_machinery":"The central object is the quantum Lorenz preorder on pairs of states: $(\\rho,\\sigma)$ Lorenz-dominates $(\\rho',\\sigma')$ exactly when its testing region $T(\\rho,\\sigma)=\\{(\\operatorname{Tr}[\\Lambda\\sigma],\\operatorname{Tr}[\\Lambda\\rho]):0\\le\\Lambda\\le I\\}$ contains the target's, equivalently when the hockey-stick divergences satisfy $E_\\gamma(\\rho\\|\\sigma)\\ge E_\\gamma(\\rho'\\|\\sigma')$ for every $\\gamma\\ge 0$; these inequalities are also read as winning-odds dominance in every binary guessing game. The workhorse representation is the layer-cake Stieltjes measure $d\\mu_{\\rho,\\sigma}$, whose stop-loss transforms are exactly the hockey-stick divergences, so Lorenz monotone functionals become convex-order monotone functionals on measures. The proof then runs through three load-bearing components: Theorem 3's bracketing construction, which sandwiches any bounded quantum Lorenz curve between finite classical Lorenz curves and shows Lorenz-continuous classical data fix the quantum value; the classical simplex representation of admissible additive divergences as positive mixtures of Rényi divergences (adapted in Supplementary Section S8); and the qubit gap-separation lemma, which computes, on an explicit two-parameter qubit family, the second-order differences between the standard quantum Rényi regularizations below and above order one and the one-shot layer-cake divergences, proving that no positive mixture of orders below one can balance a positive mixture of orders above one in the additivity identity.","core_discovery":"In the paper's own terms, the discovery is Theorem 4: if $D$ is a normalized, additive quantum Lorenz divergence—monotone under the Lorenz preorder generated by binary testing regions—and its restriction to commuting (classical) pairs is Lorenz continuous, then for every finite-dimensional pair with $\\operatorname{supp}(\\rho)\\subseteq\\operatorname{supp}(\\sigma)$, $D(\\rho\\|\\sigma)=\\operatorname{Tr}[\\rho\\log\\rho]-\\operatorname{Tr}[\\rho\\log\\sigma]$. The theorem is exact and single-shot; it does not assume a thermodynamic limit or super-additivity for correlated states. The route is: (i) a structural result (Theorem 3) that any classical divergence satisfying a mild continuity condition has at most one quantum Lorenz extension, with the extension constructed by bracketing the quantum Lorenz curve between classical Lorenz curves; (ii) the classical result that normalized, additive, data-processing monotone, Lorenz-continuous classical divergences form a simplex of Rényi mixtures; and (iii) the observation that additivity on noncommuting pairs forces every Rényi weight to vanish except the order-one one, because the one-shot layer-cake Rényi divergences regularize differently below and above order one and a qubit gap-separation lemma shows the two sides cannot balance. The paper frames the collapse as a purely quantum phenomenon: classically the same axioms leave a continuous family of admissible measures, and it is noncommutativity in tensor powers that removes all freedom.","pith_inferences":["The same mechanism suggests a template for other uniqueness questions: any operational preorder strictly finer than channel convertibility, combined with tensor-product additivity, may collapse admissible measures to the order-one entropy; multi-hypothesis testing or conditional entropies are natural next targets (the paper lists them as open directions, without proof).","The bracketing construction of Theorem 3 carries a quantitative by-product: any additive quantum Lorenz divergence is squeezed between classical envelopes, so its deviation from the Umegaki value is controlled by the Lorenz distance of the pair to the classical domain; this could be tested numerically on the qubit families used in the proof.","The threat to the theorem is the imported finite-alphabet classification used in Supplementary Section S8: if the classical representation (Eq. 37) requires stronger regularity than the conditions imposed here, the classical reduction breaks; a reader verifying the theorem's hypotheses against the original classification statement would settle this.","Without Lorenz continuity, normalized additive QLDs other than relative entropy do exist—the paper's own example $D(\\rho\\|\\sigma)+D_{\\min}(\\sigma\\|\\rho)$ shows boundary-sensitive, orientation-reversing terms are otherwise admissible—so the continuity assumption, not additivity alone, delimits the theorem's reach."],"forward_implications":["The infinite family of data-processing-monotone divergences collapses at the single-shot level: once binary-testing comparison and additivity are imposed, only the Umegaki relative entropy survives, with no asymptotic limit needed.","Classical Lorenz-continuous divergences have unique quantum extensions; the known one-shot quantum $f$-divergences are therefore forced by their classical restrictions rather than being a matter of quantization choice.","In quantum resource theories and single-shot thermodynamics, relative entropy is the unique additive distinguishability measure, so quantities such as free energy acquire a single-shot justification directly from binary discrimination.","The failure of additivity for every Rényi order other than one is itself a quantitative phenomenon: the regularization gaps $\\Delta^-_\\alpha$ and $\\Delta^+_\\alpha$ measure how far each order deviates from legitimate additivity, and the separation is provable already on qubits."],"supporting_citations":[{"why":"Defines quantum Lorenz majorization via testing regions and the Hilbert α-divergences; the QLD framework is built on this preorder.","marker":"[20]"},{"why":"Supplies the integral formula for the Umegaki relative entropy and the asymptotic regularization limits of the quantum Rényi divergences used in the additivity argument.","marker":"[21]"},{"why":"Introduces the one-shot layer-cake Rényi divergences whose regularization split below and above order one drives the collapse to α=1.","marker":"[30]"},{"why":"Provides the layer-cake integral representation of quantum divergences used to write all QLD candidates as stop-loss integrals.","marker":"[32]"},{"why":"The classical classification theorem whose finite-alphabet adaptation (Eq. 37, Supplementary Section S8) yields the Rényi-mixture form of every admissible classical divergence.","marker":"[36]"},{"why":"Supplies the convex-order/stop-loss characterization that identifies Lorenz majorization with comparison of layer-cake measures.","marker":"[43]"},{"why":"Provides the minimal/maximal extension method that Theorem 3 uses to prove uniqueness of the Lorenz extension.","marker":"[44]"}],"fun_headline_variants":["Noncommutativity collapses divergence freedom to one measure","Additivity plus optimal guessing forces quantum relative entropy","Single-shot discrimination pins down unique quantum entropy","Quantum noncommutativity forces uniqueness of relative entropy","Binary games prove Umegaki entropy is the only additive one"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the finite-alphabet classification theorem imported in Supplementary Section S8 (Eq. 37): every normalized, additive, data-processing monotone, Lorenz-continuous classical divergence is a positive mixture of Rényi divergences, an assumption the paper adapts to its one-sided normalization without proving that the theorem's own hypotheses follow from exactly the conditions used here.","fun_headline_variants_meta":{"raw":{"variants":["Noncommutativity collapses divergence freedom to one measure","Additivity plus optimal guessing forces quantum relative entropy","Single-shot discrimination pins down unique quantum entropy","Quantum noncommutativity forces uniqueness of relative entropy","Binary games prove Umegaki entropy is the only additive one"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000824,"raw_usage":{"total_tokens":3643,"prompt_tokens":1026,"completion_tokens":2617,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":642,"completion_tokens_details":{"reasoning_tokens":2542}},"tokens_in":642,"tokens_out":2617,"duration_ms":19095,"temperature":1.0,"reasoning_tokens":2542,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:36:48.490235+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the two-parameter qubit family $\\rho_r=\\frac12\\begin{pmatrix}1&r\\\\r&1\\end{pmatrix}$ and $\\sigma_s=\\operatorname{diag}(\\frac{1+s}{2},\\frac{1-s}{2})$ from Supplementary Section S7, pick an order $\\alpha\\neq 1$, and numerically evaluate the one-shot layer-cake divergence (38) on tensor powers: if $D_\\alpha(\\rho_r^{\\otimes 2}\\|\\sigma_s^{\\otimes 2})=2D_\\alpha(\\rho_r\\|\\sigma_s)$ holds for some $\\alpha\\neq 1$ and some $r,s$, then additivity does not single out order one and Theorem 4 is false. Equivalently, a direct search for any normalized additive quantum Lorenz divergence with Lorenz-continuous classical restriction that differs from the Umegaki value on this family would refute the claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the integral formula for the Umegaki relative entropy and the asymptotic regularization limits of the quantum Rényi divergences used in the additivity argument."},{"cited_title":"Hirche and M","cited_arxiv_id":null,"evidence_quote":"Introduces the one-shot layer-cake Rényi divergences whose regularization split below and above order one drives the collapse to α=1."},{"cited_title":"Sharp estimates of quantum covering problems via a novel trace inequality","cited_arxiv_id":"2507.07961","evidence_quote":"Provides the layer-cake integral representation of quantum divergences used to write all QLD candidates as stop-loss integrals."},{"cited_title":"Arthur F","cited_arxiv_id":null,"evidence_quote":"The classical classification theorem whose finite-alphabet adaptation (Eq. 37, Supplementary Section S8) yields the Rényi-mixture form of every admissible classical divergence."}],"review_version":3}