{"id":"a3fa1b04-3ae7-4045-920e-c3b75b1fc1cf","arxiv_id":"2608.04947","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Sharp dimension-dependent continuity moduli for Petz and sandwiched optimized conditional Rényi entropies, attained by isotropic state pairs, for alpha in [1/2,1).","lead":"Two families of quantum Rényi conditional entropies are shown to satisfy a sharp, explicit continuity bound in trace distance for every Rényi order between one half and one, with states that attain the bound for every distance. The proof works through a comparison-state linearization, but a load-bearing inequality in the technical appendix is currently printed with a wrong exponent and must be repaired.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the flagged Eq. (105) concern rests on a misread exponent (1−p, not 1−1/p), and the proof structure is internally consistent.","rationale":"The central claim is a sharp, dimension-dependent continuity modulus for optimized Petz and sandwiched Rényi conditional entropies on α ∈ [1/2,1). The proof architecture is: Schmidt-rank domination builds a comparison point with the correct partial-trace multiplicities; concavity linearizes Q_α via a supporting operator; Theorem A.2 controls the spectrum and anchor calibration of that operator; and an algebraic cancellation converts the resulting inequality into the claimed entropy bound. The only fragile point identified by the reader is Theorem A.2, Eq. (105). Reading Eq. (105) as θ^{1−1/p} makes it false, but the surrounding proof and Eq. (111) unambiguously require θ^{1−p}. In the proof, θZ ≤ M ≤ Z and operator monotonicity give θ^{p−1}Z^{p−1} ≤ M^{p−1}; since D_M(1) = pM^{p−1} and the inverse derivative is positive, one obtains Y_θ ≤ θ^{1−p}. The scalar example cited by the reader then satisfies 10 ≤ 10, so the counterexample disappears. The same exponent is needed in Eq. (111): θ^{1−p} multiplied into Eq. (110)'s (p−1)θ^p factor yields exactly (p−1)θ. Proposition 4.1 applies Theorem A.2 with p=1/α, X=σ^α, Z=r1⊗Tr_A σ^α, and θ=b/a; homogeneity turns Eqs. (105)–(106) into the spectral bounds and anchor calibration. The cancellation in Theorem 2.1 uses only the weight identities a^{1/α}=1−δ and (D−1)b^{1/α}=δ, together with the lower bound Q_α(σ) ≥ r^{−(1−α)/α}; all steps check out. The sandwiched proof in Proposition 5.1 mirrors this structure via the optimizer fixed-point equation and operator monotonicity of t^{−(1−α)/α}; no problematic step was found. The classical-condiitioning recovery and the α↑1 limit are consistent with the stated claims. The AI-assistance disclosure is noted but is not evidence of unsoundness. I therefore see no load-bearing concern; the reader's conditional verdict should be lifted to accept, with at most a typographical clarification of Eq. (105) requested.","tokens_in":19813,"tokens_out":40885,"duration_ms":329829,"concrete_test":"Open the arXiv source and render Eq. (105) and Eq. (111) to check whether the superscript is 1−p or 1−1/p. Re-run the scalar check with p=2, X=0, Z=1, θ=0.1: the correct Y_θ is 10, θ^{1−p} = 10, while θ^{1−1/p} ≈ 0.316. Trace Eq. (111): it requires θ^{1−p}·(p−1)θ^p = (p−1)θ. If the source unambiguously shows 1−p, the reader's objection is resolved and Proposition 4.1 stands.","verdict_should_be":"ACCEPT","load_bearing_attack":"I find no load-bearing mathematical defect. The reader's objection to Theorem A.2 appears to misread Eq. (105): the superscript is 1−p (followed by the identity operator), not 1−1/p. In the scalar case X→0, Z=1, p=2, θ=0.1, Y_θ = 10 and θ^{1−p} = 10, so the displayed bound holds exactly. The proof uses this exponent: from θZ ≤ M ≤ Z and operator monotonicity of x^{p−1}, one gets D_M(1) = pM^{p−1} ≥ pθ^{p−1}Z^{p−1}; applying the positive inverse derivative yields Y_θ ≤ θ^{1−p}1. The same exponent is required in Eq. (111), where θ^{1−p}·TrΔ_θ combines with Eq. (110) to give the (p−1)θ factor; with θ^{1−1/p} the step would fail. Proposition 4.1 then follows from Theorem A.2 by homogeneity with p = 1/α, and the subsequent cancellation in Theorem 2.1 is algebraically consistent. No unsupported step was found in the sandwiched proof beyond the same calibration structure.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes sharp continuity bounds, in trace distance, for the optimized Petz and sandwiched Rényi conditional entropies for all orders α ∈ [1/2, 1). The main results, Theorems 2.1 and 2.2, state that for two bipartite states within trace distance δ, both conditional entropies differ by at most Γ_{α,D}(δ), where D = d_A min{d_A,d_B} and Γ_{α,D} is the explicit binary modulus in Eq. (7), with a plateau log D beyond δ = 1 − 1/D. The proofs use a comparison point built from the isotropic equality family, Schmidt-rank domination, concavity of the relevant Rényi functionals, trace-distance duality, and a noncommutative calibration estimate. The bound is shown to be attained for every δ by an isotropic pair with a maximally entangled anchor, and taking α ↑ 1 recovers the recent sharp conditional-entropy bound of Berta et al.","tokens_in":20062,"tokens_out":18127,"duration_ms":136935,"significance":"If correct, this resolves the sharp continuity modulus for both optimized Petz and sandwiched Rényi conditional entropies in the full quantum case for α ∈ [1/2, 1), a question that was previously open. The result is parameter-free, the saturation family is explicit, and the classical-conditioning case recovers the Jabbour–Datta theorem. The paper also derives a uniform 1/2-Hölder continuity statement for the random-coding exponent and honestly identifies the limitations below α = 1/2. A notable strength is the transparent proof architecture: the comparison point is constructed to have exactly the target Q_α value, and all remaining work is reduced to the calibration inequality in Proposition 4.1 and its sandwiched analogue, with no fitted constants. The disputed technical estimate in Theorem A.2 is correct as written: Eq. (105) states Y_θ ≤ θ^{1−p} 1, not θ^{1−1/p}, and the proof uses the exponent 1−p consistently, e.g., in Eq. (111).","major_comments":[],"minor_comments":[{"comment":"The display for the one-ray transport inequality would be much less ambiguous if written as 1 = Y_1 ≤ Y_θ ≤ θ^{1−p} 1, with explicit parentheses around the exponent and the identity operator. As typeset, the superscript can be misread as 1 − 1/p; the proof and Eq. (111) rely on the exponent 1 − p, which is correct.","section":"Appendix A.2, Eq. (105)"},{"comment":"The limit expression (G_{σ,δ}^{(α)} − 1_{AB})/(1 − α) → G_AB should be typeset with clear parentheses and a defined notation; the current inline formatting makes it easy to misparse the numerator and denominator.","section":"Section 7.2, Eq. (94)"},{"comment":"The limit identity lim_{α↑1} (1/(1−α)) log[(1−δ)^α + (D−1)^{1−α}δ^α] = h_2(δ) + δ log(D−1) is only valid on the increasing branch δ ≤ 1 − 1/D. Please state that restriction explicitly or replace δ by ε = min{δ, 1 − 1/D} to avoid confusion.","section":"Introduction, Eq. (3)"},{"comment":"The sentence 'At δ = 1 − D^{-1}, its value is log D' should name the quantity being evaluated, namely Γ_{α,D}(1 − D^{-1}) = log D, rather than relying on the pronoun.","section":"Section 4, proof of Theorem 2.1"},{"comment":"The disclosure that the noncommutative estimate was 'assisted by ChatGPT 5.6 Sol' is unusual; the authors should clarify the role of the AI tool and ensure the disclosure conforms to the journal's policy on AI assistance.","section":"Abstract and Acknowledgements"}],"recommendation":"minor_revision","confidential_remarks":"The reader's report's strongest objection concerns Eq. (105) in Appendix A.2, but that objection rests on a misreading of the exponent. The manuscript clearly uses θ^{1−p}, and the scalar check quoted in the reader's report actually confirms the bound: for p = 2 and θ = 0.1, θ^{1−p} = 10, which matches Y_θ ≈ 10. I found no load-bearing mathematical defect in the proof structure, and the algebra in Sections 4, 5, and Appendix A is internally consistent. The remaining issues are presentation-level. The unusual AI-assistance disclosure may warrant editorial attention regarding journal policy."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this paper settles the sharp fully quantum continuity modulus for optimized Petz and sandwiched conditional Rényi entropies for all α in [1/2,1), and the proof is in better shape than the reader's report suggests. The flagged flaw in Theorem A.2 is not a flaw: the bound in Eq (105) is Y_θ ≤ θ^{1−p} 1, not θ^{1−1/p}. In the scalar counterexample p=2, θ=0.1, both sides equal 10. The proof of that lemma is coherent: θZ ≤ X+θ(Z−X) ≤ Z, operator monotonicity of x^{p−1}, then the positive inverse derivative gives exactly θ^{1−p}. So the central calibration estimate stands.\n\nWhat is genuinely new: prior sharp bounds only covered classical or classical-quantum conditioning, and the fully quantum bounds that existed were not sharp. The comparison-point construction, based on the isotropic saturating family and Schmidt-rank domination, is the right idea and it works. The bound is explicit, attained for every δ, and taking α↑1 recovers the order-one result. The classical recovery of Jabbour–Datta and the clean discussion of why α<1/2 is outside the current technique are useful context.\n\nSoft spots are minor. The proof is dense—Appendix A carries real weight, and a referee should go through Theorem A.2 and Proposition 4.1 line by line, but no circularity or fitted constants appear. The AI-assistance disclosure is transparent; it doesn't by itself weaken the math, and the matrix-analysis lemma is stated and proved in the text. The discussion section honestly notes that orders below 1/2 and above 1 remain open, including the interesting point that the binary modulus does not extend to the min-entropy endpoint.\n\nWho is this for? Anyone working on quantum information measures, continuity bounds, or Rényi entropies. It's worth citing and it deserves a serious referee. I would send it to peer review with a request for a careful check of the technical appendix.","headline":"The paper solves a real open problem and the proof holds up; the reader's Eq (105) worry is a misread exponent.","tokens_in":20608,"tokens_out":5819,"would_cite":true,"duration_ms":50627,"reading_group":"yes","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 for every order $\\alpha\\in[1/2,1)$, two bipartite states within trace distance $\\delta$ have optimized Petz and sandwiched R\\'enyi conditional entropies differing by at most $\\Gamma_{\\alpha,D}(\\delta)$, and that this…","keywords":["Rényi conditional entropy","Petz divergence","sandwiched Rényi divergence","continuity bound","trace distance","Schmidt-rank domination","quantum information theory","conditional entropy"],"falsifier":"A reader can test Theorem A.2 with scalars $p=2$, $X\\approx0$, $Z=1$, $\\theta=0.1$: the quantity $Y_\\theta$ defined in Eq. (104) is about $10$, while $\\theta^{1-1/p}=\\theta^{1/2}\\approx0.316$, so Eq. (105) is false; Eq. (111) only compiles with the exponent $1-p$. This scalar check settles whether the printed transport estimate is usable, and without a repaired estimate the sharp continuity proof does not go through.","tokens_in":19600,"feed_emoji":"⚛️","tokens_out":8550,"duration_ms":70098,"temperature":0.7,"pith_summary":"The paper establishes a sharp modulus of continuity for two quantum information quantities: the optimized Petz and sandwiched R\\'enyi conditional entropies, for every order $\\alpha\\in[\\frac12,1)$. If two bipartite states are within trace distance $\\delta$, then both entropies differ by at most $\\Gamma_{\\alpha,D}(\\delta)=\\frac{1}{1-\\alpha}\\log[(1-\\varepsilon)^{\\alpha}+(D-1)^{1-\\alpha}\\varepsilon^{\\alpha}]$, where $\\varepsilon=\\min\\{\\delta,1-1/D\\}$ and $D=d_A\\min\\{d_A,d_B\\}$ is the effective dimension. The bound is optimal for every distance constraint, attained by an isotropic pair anchored at a maximally entangled state, and its $\\alpha\\uparrow1$ limit recovers the sharp order-one conditional-entropy bound. A classical conditioning system reduces the same construction to scalar power inequalities, recovering the known sharp quantum-classical bound for all $0<\\alpha<1$.","feed_headline":"One formula sets the sharp bound for Rényi conditional entropy","feed_subtitle":"δ-close states shift both Rényi conditional entropies by the same sharp formula; an isotropic pair attains it.","key_machinery":"The engine is the supporting-hyperplane linearization of the concave, positively homogeneous functional $Q_\\alpha(X)=\\mathrm{Tr}[(\\mathrm{Tr}_A X^{\\alpha})^{1/\\alpha}]$ (and its sandwiched analogue $\\widetilde{Q}_\\alpha$). At a comparison point $\\tau^{(\\alpha)}_{\\sigma,\\delta}$, defined by $\\tau^{\\alpha}=a\\sigma^{\\alpha}+b(r\\,1_A\\otimes\\mathrm{Tr}_A[\\sigma^{\\alpha}]-\\sigma^{\\alpha})$ with $a=(1-\\delta)^{\\alpha}$ and $b=(\\delta/(D-1))^{\\alpha}$, the gradient $\\nabla Q_\\alpha(\\tau)$ is a supporting operator: $Q_\\alpha(X)\\le\\mathrm{Tr}[\\nabla Q_\\alpha(\\tau)X]$. Schmidt-rank domination, $X\\le r\\,1_A\\otimes\\mathrm{Tr}_A[X]$, supplies the positive complement with the required $1:(D-1)$ partial-trace relation, and a so-called one-ray transport theorem yields the spectral bounds and anchor calibration on $\\nabla Q_\\alpha(\\tau)$. Together these estimates reproduce the isotropic equality calculation exactly. For the sandwiched functional, the supporting operator is identified from the optimizer fixed-point equation and Danskin's theorem.","core_discovery":"The central claim is that continuity of both the optimized Petz conditional entropy $H^\\uparrow_\\alpha(A|B)$ and its sandwiched counterpart is governed by one scalar modulus: whenever $T(\\rho,\\sigma)\\le\\delta$, the entropy difference is at most $\\Gamma_{\\alpha,D}(\\delta)$, with $\\Gamma_{\\alpha,D}(\\delta)=\\frac{1}{1-\\alpha}\\log[(1-\\varepsilon)^{\\alpha}+(D-1)^{1-\\alpha}\\varepsilon^{\\alpha}]$, $\\varepsilon=\\min\\{\\delta,1-1/D\\}$, and $D=d_A\\min\\{d_A,d_B\\}$. The bound is attained for every $\\delta$ by the isotropic pair $\\rho_\\delta=(1-\\delta)\\Phi_r+\\frac{\\delta}{D-1}(1_A\\otimes P_{B0}-\\Phi_r)$ with $\\sigma=\\Phi_r$, where $\\Phi_r$ is a maximally entangled state of Schmidt rank $r=\\min\\{d_A,d_B\\}$. For a general anchor $\\sigma$, the proof constructs a comparison point whose $\\alpha$-power has exactly the same partial-trace multiplicities as the isotropic equality model, then linearizes the concave R\\'enyi functional $Q_\\alpha$ at that point; two supporting-operator estimates control the change-of-measure term and the anchor term without weakening the constant. Taking $\\alpha\\uparrow1$ recovers the sharp von Neumann conditional-entropy bound, and classical conditioning recovers the sharp quantum-classical bound for all $0<\\alpha<1$.","pith_inferences":["The comparison-point construction with weights $a,b$ and a Schmidt-rank complement is geometrically independent of $\\alpha$, so it likely transfers to the $(\\alpha,z)$-family of R\\'enyi divergences, provided the required operator-convexity and operator-monotonicity ranges are rechecked.","Because the printed one-ray transport estimate fails already in scalars, the sharp modulus may still be true but needs a repaired transport lemma; if the correct exponent is $1-p$ rather than $1-1/p$, the overall proof structure can probably be preserved.","The open range $0<\\alpha<1/2$ may be approachable by a different operator inequality, since the construction itself does not need $\\alpha\\ge1/2$ except for the operator-convexity and monotonicity properties of the power maps.","Above order one the extremal geometry genuinely changes: the min-entropy endpoint has modulus $\\min\\{\\log D,\\log(1+D\\delta)\\}$, strictly larger than the formal $\\alpha\\to\\infty$ limit of $\\Gamma_{\\alpha,D}$, so the binary formula should not be extrapolated past $\\alpha=1$."],"forward_implications":["If Theorem 2.1 is correct, the random-coding exponent $E_r(R)_\\rho$ is uniformly $1/2$-H\\\"older continuous in trace distance, with worst-case sensitivity governed by $\\Gamma_{1/2,D}(\\delta)\\le 2\\sqrt{(D-1)\\delta}$.","Taking $\\alpha\\uparrow1$ yields the sharp order-one continuity bound $|H(A|B)_\\rho-H(A|B)_\\sigma|\\le h_2(\\varepsilon)+\\varepsilon\\log(D-1)$, so the R\\'enyi result contains the von Neumann conditional-entropy bound as a limit.","For classical conditioning systems, where $r=1$ and $D=d_A$, the same construction recovers the sharp Jabbour\\,{}-\\,Datta bound for all $0<\\alpha<1$.","The explicit saturation family means every distance constraint $\\delta\\in[0,1]$ has a witnessed pair, so the modulus $\\Gamma_{\\alpha,D}(\\delta)$ cannot be improved.","The decoupling, privacy-amplification, and communication exponents that the paper cites inherit a quantitative robustness estimate under state perturbations of size $\\delta$."],"supporting_citations":[{"why":"Supplies the Schmidt-rank domination inequality $X\\le r\\,1_A\\otimes\\mathrm{Tr}_A[X]$, which constructs the positive complement in the comparison point.","marker":"[TH00]"},{"why":"Defines the Petz\\,{}-\\,R\\'enyi divergence whose optimized conditional entropy is the object of Theorem 2.1.","marker":"[Pet86]"},{"why":"Defines the sandwiched R\\'enyi divergence and supplies the direct-sum property used for concavity of $\\widetilde{Q}_\\alpha$.","marker":"[MLDS+13]"},{"why":"Provides the foundational sandwiched R\\'enyi divergence framework used in the definition and proof for Theorem 2.2.","marker":"[WWY14]"},{"why":"Gives the quantum Sibson identity $H^\\uparrow_\\alpha=\\alpha/(1-\\alpha)\\log Q_\\alpha$, converting entropy continuity into the $Q_\\alpha$ bound.","marker":"[SW13]"},{"why":"Supplies Epstein's trace-concavity theorem, which yields concavity of $Q_\\alpha$ and hence the supporting-operator inequality.","marker":"[Eps73]"},{"why":"Gives the optimizer fixed-point equation and uniqueness theorem used to identify and compute the sandwiched supporting operator.","marker":"[HT16]"},{"why":"Danskin's theorem justifies differentiating the maximization in the sandwiched functional $\\widetilde{Q}_\\alpha$ at its optimizer.","marker":"[Dan67]"},{"why":"Provides the sharp quantum-classical continuity bound that Corollary 6.1 recovers for all $0<\\alpha<1$.","marker":"[JD22]"},{"why":"Supplies the sharp order-one conditional-entropy bound recovered in the limit $\\alpha\\uparrow1$.","marker":"[BCRK+26]"}],"fun_headline_variants":["Sharp modulus unifies Petz and sandwiched Rényi entropies","One bound governs both Rényi conditional entropies","Isotropic pair attains sharp continuity for Rényi entropies","Exact continuity modulus for Rényi conditional entropies","Tight bound for Petz and sandwiched Rényi entropies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the one-ray transport estimate in Appendix A.2, which claims that a certain operator built from the derivative of the power function stays below $\\theta^{1-1/p}$; as printed it fails even for scalar $p=2$, so the central theorem is not yet demonstrated without a corrected version of that estimate.","fun_headline_variants_meta":{"raw":{"variants":["Sharp modulus unifies Petz and sandwiched Rényi entropies","One bound governs both Rényi conditional entropies","Isotropic pair attains sharp continuity for Rényi entropies","Exact continuity modulus for Rényi conditional entropies","Tight bound for Petz and sandwiched Rényi entropies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000364,"raw_usage":{"total_tokens":2044,"prompt_tokens":1113,"completion_tokens":931,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":729,"completion_tokens_details":{"reasoning_tokens":838}},"tokens_in":729,"tokens_out":931,"duration_ms":6263,"temperature":1.0,"reasoning_tokens":838,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:37:24.552518+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader can test Theorem A.2 with scalars $p=2$, $X\\approx0$, $Z=1$, $\\theta=0.1$: the quantity $Y_\\theta$ defined in Eq. (104) is about $10$, while $\\theta^{1-1/p}=\\theta^{1/2}\\approx0.316$, so Eq. (105) is false; Eq. (111) only compiles with the exponent $1-p$. This scalar check settles whether the printed transport estimate is usable, and without a repaired estimate the sharp continuity proof does not go through.","supporting_citations":[],"review_version":2}