{"id":"2bdb3b57-e1f4-4561-bced-a1c29906a78b","arxiv_id":"1909.01755","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For classical-quantum states, the conditional entropy can change by at most epsilon log2(d_B-1) + h2(epsilon) under a trace-distance perturbation epsilon, and this bound cannot be improved.","lead":"This note proves the exact limit on how much the conditional entropy of a classical-quantum state can change when the state is perturbed. The result also improves the best-known bound for entanglement of formation, a central measure of quantum entanglement.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 1 is a correct reduction to the cited Alhejji-Smith classical bound (1), but the epsilon-range and the claimed optimality are inherited wholesale from that quoted bound; this external dependency is the single load-bearing point.","rationale":"The paper is a short, honest reduction note. I traced the proof of Proposition 1 in full. The unitality of the conditional dephasing channel (Eq. (8)) is correct: sum_{x,y} |x><x| tensor |phi^{y,x}><phi^{y,x}| = I_XB because for each x the set {|phi^{y,x}>}_y is an orthonormal basis of B. The entropy monotonicity step (Eqs. (12)-(16)) is valid: H(B|X)_sigma = H(BX)_sigma - H(X)_sigma = H(BX)_sigma - H(X)_{Delta(sigma)} <= H(BX)_{Delta(sigma)} - H(X)_{Delta(sigma)} = H(Y|X)_s. The trace distance step (Eqs. (19)-(21)) is valid by data processing and co-diagonality. The saturation of the cq bound follows because the classical saturating example embeds isometrically as classical-classical (hence cq) states, preserving both the trace distance and the conditional entropy difference. The countable generalization (Corollary 3) uses a standard projection-and-replacement channel and a known limit; it is a corollary and not the central claim. Corollary 2 delegates the entanglement-of-formation argument to Winter's method with the parameter substitution delta = sqrt(epsilon(2-epsilon)); the mapping delta in (0, 1-1/d] to epsilon in (0, 1 - sqrt(2d-1)/d] checks out algebraically. The one point where the result stands or falls without internal demonstration is the quoted classical bound (1) and its tightness example. The reader identified exactly this as the weakest assumption, and I agree. It does not change the verdict: the reduction is mathematically clean, the dependency is stated openly, and Eq. (1) is a separately verifiable theorem in the literature. The recommended outcome is therefore UNCHANGED (ACCEPT), with the concrete check above as a prudent verification of the inherited tightness claim.","tokens_in":6487,"tokens_out":30097,"duration_ms":259440,"concrete_test":"Verify the inherited optimality/range claim at the smallest case that exercises the interior and the endpoint of the range: take d_B = |Y| = 3 and epsilon = 1/2 and epsilon = 2/3. Write out the distributions p_XY and q_XY from Eqs. (27)-(28) of [1] explicitly for these parameters; check that (1/2)||p_XY - q_XY||_1 = epsilon exactly and |H(Y|X)_p - H(Y|X)_q| = epsilon log2(2) + h2(epsilon). Then embed them as classical-classical states rho_XB and sigma_XB diagonal in |x>_X tensor |y>_B, recompute the trace distance and the cq conditional entropies, and confirm both equalities are reproduced exactly. If the example only saturates at the endpoint or only approximately, then the optimality statement of Proposition 1 needs qualification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, Proposition 1 (Eq. (4)), is obtained by reducing the classical-quantum problem to the classical equivocation bound of Alhejji and Smith, Eq. (1), via the conditional dephasing channel in Eq. (8). I checked the reduction step by step: the channel is unital, so H(BX) does not decrease under it; the X marginal is preserved by the partial trace, so H(X) is unchanged; rho_XB and Delta(sigma_XB) are co-diagonal, so the trace distance in Eqs. (19)-(21) is exactly the total variation distance between the induced classical distributions r_XY and s_XY; and the spectral coefficients r(y|x), s(y|x) = <phi^{y,x}|sigma_B^x|phi^{y,x}> are valid conditional distributions. The argument is internally sound. The single load-bearing input is Eq. (1) together with its claimed saturation by the example in Eqs. (27)-(28) of [1]; neither is proved in this manuscript. If Eq. (1) were valid only on a narrower range, or if the saturation example achieved the bound only up to a constant or only at the endpoint, then the range epsilon in (0, 1-1/d_B] and the 'optimal' claim of Proposition 1 would fail. This is a dependency concern, routed through correctness risk, not an inconsistency; the paper is explicit that it quotes [1], but anyone relying on the optimality claim must also rely on that concurrent preprint.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves an optimal uniform continuity bound for the conditional entropy of finite-dimensional classical–quantum (cq) states. Proposition 1 (Eq. (4)) states that for ε ∈ (0, 1 − 1/d_B], if ρ_XB and σ_XB are cq states with half trace distance ≤ ε, then |H(B|X)_ρ − H(B|X)_σ| ≤ ε log2(d_B − 1) + h2(ε), and that the bound is tight for every d_B and ε in the range. The proof uses a conditional dephasing channel in the eigenbasis of ρ's B-conditional states, exploits unitality to bound σ's conditional entropy by a classical conditional entropy, and invokes the Alhejji–Smith classical equivocation bound (Eq. (1)). Corollary 2 gives a Winter-style uniform continuity bound for entanglement of formation, and Corollary 3 extends the main bound to countably infinite classical conditioning alphabets.","tokens_in":6766,"tokens_out":22058,"duration_ms":219511,"significance":"If the Alhejji–Smith classical bound is valid, Proposition 1 is optimal and improves the corresponding case in Winter's Lemma 2. The reduction is elegant and the trace-distance bookkeeping is exact; there are no free parameters or fitted constants. The paper is explicit that both the range and the saturation example are inherited from [1], so the internal derivation is not circular. The entanglement-of-formation application and the countable-X extension are useful additions, although they rely on standard methods and cited results rather than new techniques.","major_comments":[],"minor_comments":[{"comment":"The one-sentence reference to the saturation example in Eqs. (27)–(28) of [1] would be clearer if it explicitly said that the classical alphabet Y is encoded as diagonal states on system B, so the classical pair of distributions converts to a pair of cq states with the same trace distance and the same conditional entropies.","section":"Proposition 1, tightness paragraph"},{"comment":"Because the proof is delegated to Winter's Corollary 4, please add one sentence explaining how δ = sqrt(ε(2 − ε)) arises from Uhlmann's theorem and that both directions of the entanglement-of-formation inequality follow from the same argument; as written the reader must reconstruct this step.","section":"Corollary 2"},{"comment":"Please define ρ_B = Σ_x r(x)ρ_B^x explicitly before Eq. (49), where the notation first appears, so that the final identity H(B)_ρ − I(X;B)_ρ = Σ_x r(x)H(ρ_B^x) is unambiguous.","section":"Corollary 3"},{"comment":"Reference [1] should be updated to its published version, if one exists, rather than cited as a September 2019 preprint, because the optimality claim of Proposition 1 depends on its correctness.","section":"References"}],"recommendation":"accept","confidential_remarks":"This is a short, correct note. The main result is a clean reduction to the concurrent classical result of Alhejji and Smith, and the derivation is internally sound. The editor may wish to confirm that [1] has been through the appropriate review process, but this does not affect the validity of the reduction presented here. The paper fits a short-format quantum information journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper does exactly what the title promises. It proves the optimal uniform continuity bound for the conditional entropy of classical–quantum states, with the correct constant ε log(d_B − 1) + h₂(ε), by reducing the quantum problem to the recent classical equivocation bound of Alhejji and Smith. I checked the reduction line by line and it holds up.\n\nWhat is new: Winter's earlier bound for this case was suboptimal; here the bound is tight for every ε in (0, 1 − 1/d_B] and every d_B, and the saturation transfer works because the trace distance between the two constructed cq states equals the total variation distance of the induced classical distributions. The entanglement of formation corollary is a real improvement over Winter, and the countable-alphabet extension answers an open question from Alhejji–Smith. The conditional dephasing trick itself is not new (it goes back to Petz, via Winter), but the application is clean and the optimality statement is the genuinely new content.\n\nWhere it is soft: the single load-bearing input is the quoted Alhejji–Smith bound; the paper does not prove it, and if that bound were wrong or its ε-range narrower than claimed, Proposition 1 would collapse. That is an honest dependency — the paper flags it explicitly — and the classical bound has since held up in the literature, so I do not treat it as a real flaw. Corollary 2 delegates its proof to Winter's method, and Corollary 3 leans on Kuznetsova's limit theorem; both are minor because the cited arguments are standard and the central claim, Proposition 1, is proved in full. One small note: in the upper half of the range, ε ∈ (1/2, 1−1/d_B], the right-hand side is still increasing in ε because (d_B−1)(1−ε)/ε ≥ 1, so applying the classical bound with the looser ε is legitimate.\n\nWho this is for: anyone who needs uniform continuity bounds for capacities, entropy estimates, or resource-theoretic quantities in classical–quantum settings. This is a short, checkable, useful note, not a paradigm shift. I would send it to peer review — the optimality of a fundamental bound merits referee time, and a single careful referee can verify the whole proof in one sitting.","headline":"A clean reduction to Alhejji–Smith proves the optimal cq conditional entropy bound; the proof checks out, the external dependency is honest, and the corollaries are useful — worth refereeing.","tokens_in":7257,"tokens_out":7313,"would_cite":true,"duration_ms":65196,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","94A17"],"pacs":["03.67.-a"],"model":"deepseek-v4-flash","headline":"This paper proves that a classical–quantum state's conditional entropy satisfies the optimal uniform continuity bound $\\varepsilon\\log_2(d_B-1)+h_2(\\varepsilon)$ whenever the two states are within trace distance $\\varepsilon$, and that…","keywords":["conditional entropy","classical-quantum states","uniform continuity bound","trace distance","binary entropy","entanglement of formation","data processing","optimal bound"],"falsifier":"For $d_B=3$ and $\\varepsilon=0.3$, the claimed bound is $0.3\\log_2(2)+h_2(0.3)\\approx1.181$ bits; a numerical search over classical–quantum states with $\\frac12\\|\\rho-\\sigma\\|_1=0.3$ that produces any pair with a larger conditional-entropy gap would refute Proposition 1, while recovering exactly this gap from the classical saturating construction would confirm it.","tokens_in":6285,"feed_emoji":"⚛️","tokens_out":15618,"duration_ms":128831,"temperature":0.7,"pith_summary":"This paper proves an optimal uniform continuity bound for the conditional entropy of classical–quantum states: whenever two such states are within trace distance $\\varepsilon$ and $\\varepsilon$ lies in $(0,1-1/d_B]$, their conditional entropies differ by at most $\\varepsilon\\log_2(d_B-1)+h_2(\\varepsilon)$, where $d_B$ is the dimension of the quantum system and $h_2$ is the binary entropy. The proof converts the problem into a classical one by dephasing each conditional quantum state in the eigenbasis of one of the states; this does not decrease the conditional entropy of the comparison state and does not increase the trace distance, so the optimal classical bound applies. The quantum bound inherits optimality from the classical one: for every $d_B$ and every allowed $\\varepsilon$, there is a pair of classical–quantum states that satisfies the inequality with equality. An immediate corollary improves the known uniform continuity bound for entanglement of formation, and the argument also covers countably infinite classical conditioning alphabets.","feed_headline":"Optimal continuity bound for classical–quantum conditional entropy","feed_subtitle":"The exact trade-off between trace distance and conditional entropy is now known, sharpening entanglement bounds.","key_machinery":"The key device is the conditional dephasing channel $\\Delta^{\\mathrm{cd}}_{XB}(\\omega_{XB})=\\sum_{x,y}(|x\\rangle\\langle x|_X\\otimes|\\varphi^{y,x}\\rangle\\langle\\varphi^{y,x}|_B)\\omega_{XB}(|x\\rangle\\langle x|_X\\otimes|\\varphi^{y,x}\\rangle\\langle\\varphi^{y,x}|_B)$, which dephases the quantum system $B$ in the eigenbasis $\\{|\\varphi^{y,x}\\rangle\\}$ of each $\\rho_x^B$ after reading the classical label $x$. This channel leaves $\\rho_{XB}$ unchanged, maps $\\sigma_{XB}$ to a commuting state, and by data processing never increases the normalized trace distance; being unital, it also never decreases the entropy of the full state, so the conditional entropy of the dephased state dominates the original. The reduction turns the quantum inequality into the classical bound $|H(Y|X)_r-H(Y|X)_s|\\leq\\varepsilon\\log_2(|Y|-1)+h_2(\\varepsilon)$, and the classical example that saturates that bound provides the saturating classical–quantum pair.","core_discovery":"The central claim is Proposition 1: for finite-dimensional classical–quantum states $\\rho_{XB}=\\sum_x r(x)|x\\rangle\\langle x|_X\\otimes\\rho_x^B$ and $\\sigma_{XB}=\\sum_x s(x)|x\\rangle\\langle x|_X\\otimes\\sigma_x^B$, with $\\varepsilon\\geq \\frac12\\|\\rho_{XB}-\\sigma_{XB}\\|_1$ and $\\varepsilon\\in(0,1-1/d_B]$, one has $|H(B|X)_\\rho-H(B|X)_\\sigma|\\leq \\varepsilon\\log_2(d_B-1)+h_2(\\varepsilon)$. The bound is uniform in that the right-hand side depends only on $\\varepsilon$ and $d_B$, and it is optimal in the strongest sense: for every $d_B$ and every $\\varepsilon$ in the stated range there exists a pair of states attaining equality. The paper also derives the corresponding uniform continuity bound for entanglement of formation and extends the classical–quantum bound to countable alphabets.","pith_inferences":["The proof's reliance on eigenbasis dephasing suggests that the open quantum–classical and fully quantum analogues will need a different mechanism, since no single dephasing channel can preserve the entropy structure of both states in those cases.","A testable extension is to apply the same conditional-dephasing reduction to conditional mutual information or other one-sided information measures; if a matching tight classical inequality exists, the same argument would likely produce the optimal quantum bound.","The bound's dependence only on $d_B$, and not on the classical alphabet size, indicates that the conditioning side enters only through probability weights; related one-sided measures may exhibit the same collapse of dimension dependence."],"forward_implications":["For every $d_B$ and every $\\varepsilon\\in(0,1-1/d_B]$, there exist states meeting the bound exactly, so no uniform bound of the same form can be improved.","The entanglement-of-formation version bounds $|E_F(\\rho_{AB})-E_F(\\sigma_{AB})|$ by $\\delta\\log_2(d-1)+h_2(\\delta)$ with $\\delta=\\sqrt{\\varepsilon(2-\\varepsilon)}$ and $d=\\min\\{d_A,d_B\\}$, valid up to $\\varepsilon=1-\\sqrt{(2d-1)/d}$, improving the previous bound.","The bound extends without change to a countably infinite classical alphabet $X$ as long as the quantum system remains finite-dimensional.","Tight conditional-entropy estimates of this kind are the standard ingredient for converting approximate closeness of quantum channels into estimates of their communication capacities, so the optimal value sharpens those estimates whenever classical–quantum output states are involved."],"supporting_citations":[{"why":"Supplies the tight classical conditional-entropy bound and the saturating pair of distributions that the quantum proof reduces to.","marker":"[1]"},{"why":"Gives the previous uniform continuity bound for quantum conditional entropy that Proposition 1 improves, and the proof pattern used for the entanglement-of-formation corollary.","marker":"[14]"},{"why":"Attributed source of the conditional dephasing reduction idea that turns classical–quantum states into commuting classical distributions.","marker":"[19]"},{"why":"Defines entanglement of formation through a minimization over pure-state decompositions, which the corollary bounds.","marker":"[20]"},{"why":"Provides the conditional-entropy formula and limiting argument needed to extend the bound to countable classical alphabets.","marker":"[21]"}],"fun_headline_variants":["Exact entropy-distance trade-off for cq states","Optimal continuity bound for cq conditional entropy","Sharpest possible bound for cq entropy","Tightest possible bound for classical-quantum entropy"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantum proof inherits everything from a quoted classical bound; if that classical bound had a gap or a narrower valid range, the claimed quantum result would lose both its range and its tightness.","fun_headline_variants_meta":{"raw":{"variants":["Exact entropy-distance trade-off for cq states","Optimal continuity bound for cq conditional entropy","Sharpest possible bound for cq entropy","Tightest possible bound for classical-quantum entropy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00158,"raw_usage":{"total_tokens":6267,"prompt_tokens":875,"completion_tokens":5392,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":491,"completion_tokens_details":{"reasoning_tokens":5331}},"tokens_in":491,"tokens_out":5392,"duration_ms":36108,"temperature":1.0,"reasoning_tokens":5331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:08:56.688580+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For $d_B=3$ and $\\varepsilon=0.3$, the claimed bound is $0.3\\log_2(2)+h_2(0.3)\\approx1.181$ bits; a numerical search over classical–quantum states with $\\frac12\\|\\rho-\\sigma\\|_1=0.3$ that produces any pair with a larger conditional-entropy gap would refute Proposition 1, while recovering exactly this gap from the classical saturating construction would confirm it.","supporting_citations":[{"cited_title":"A Tight Uniform Continuity Bound for Equivocation","cited_arxiv_id":"1909.00787","evidence_quote":"Supplies the tight classical conditional-entropy bound and the saturating pair of distributions that the quantum proof reduces to."},{"cited_title":"Tight uniform continuity bounds for quantum entropies: conditional entropy, relative entropy distance and energy constraints","cited_arxiv_id":"1507.07775","evidence_quote":"Gives the previous uniform continuity bound for quantum conditional entropy that Proposition 1 improves, and the proof pattern used for the entanglement-of-formation corollary."},{"cited_title":"Quantum Information Theory and Quantum Statistics","cited_arxiv_id":null,"evidence_quote":"Attributed source of the conditional dephasing reduction idea that turns classical–quantum states into commuting classical distributions."},{"cited_title":"Quantum conditional entropy for infinite-dimensional systems","cited_arxiv_id":"1004.4519","evidence_quote":"Provides the conditional-entropy formula and limiting argument needed to extend the bound to countable classical alphabets."}],"review_version":1}