{"id":"3522d029-7bca-45fe-a904-5de9fcd630f4","arxiv_id":"2607.28961","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The Holevo information of a multi-task quantum system is bounded by K log(1 + √TrA/(2√K)), where A is a prior-weighted global quantum Fisher information matrix.","lead":"This paper derives upper bounds on how much information a single quantum carrier can carry about several simultaneous tasks, using a matrix built from quantum Fisher information. It also shows—in a two-mode photonic model—that adding more light does not always increase the usable information for independent tasks, because past a certain resource level the information concentrates into a single collective mode.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unproved Lemma 1 is the load-bearing foundation of Theorems 1–2; it is not shown to apply to Holevo chi as opposed to accessible information, so the main bound is conditional.","rationale":"The central claim is the non-asymptotic, measurement-independent Holevo bound in terms of the g-QFIM. The proof structure is: Lemma 1 -> Theorem 1 -> Theorem 2 -> K-task extension. I checked the internal steps after Lemma 1: the chain rule for cq mutual information, QFI monotonicity under partial trace, Jensen and Cauchy–Schwarz steps, and the PSD construction of A all appear correct. Thus the truth of the paper's main theorem is exactly as secure as Lemma 1. The reader's weakest assumption identifies this lemma, and I agree that it is the load-bearing point. The additional worry — that Lemma 1 might be a statement about accessible information rather than Holevo information — is legitimate because the paper does not show how the Holevo quantity itself is controlled. I am not asserting the lemma is false; rather, this is the one place where the central claim is least secure. The phase-transition claim is indeed supported only by a specific toy model, but it is secondary: even if the 'phase transition' language is overclaimed, the theorems could still be correct. Since the missing lemma proof and the phase-transition overclaim both warrant the reader's CONDITIONAL verdict, I recommend no change in verdict. The proposed test — checking the exact source theorem and/or searching for a counterexample — would settle whether the main bound has a valid proof.","tokens_in":16490,"tokens_out":26116,"duration_ms":229390,"concrete_test":"Check Lemma 1 against the original source and by independent derivation. (1) Obtain from [29] the exact theorem invoked and verify that it bounds the Holevo quantity chi (not merely accessible information sup_M I(Θ;Y)) and that it applies to non-smooth priors on a compact interval. (2) Independently prove or disprove the scalar inequality for differentiable one-parameter cq states, paying attention to the factor 1/2. (3) As a numerical cross-check, run a random search over qubit/qutrit families theta -> rho_theta and absolutely continuous priors (uniform, triangular, smoothed two-point), computing chi = S(∫ p rho) - ∫ p S(rho) and L = ∫ sqrt(F); if any instance violates chi <= log(1 + L/2), Lemma 1 is false and Theorems 1–2 are unproven; if all pass, the missing proof should still be written out explicitly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Appendix B's Lemma 1 (Holevo chi <= log(1 + (1/2) ∫ sqrt(F) dtheta)) is the only external input carrying Theorems 1 and 2, and it is quoted without proof or a precise derivation. This is not merely a formality. The standard route from QFI to a mutual-information bound uses data processing for a fixed POVM, which controls sup_M I(Θ;Y) — the accessible information. The Holevo information chi(Θ:B) can be strictly larger than single-copy accessible information and is generally achieved only by block codes/collective measurements; bounding chi therefore requires a separate quantum argument, not a corollary of the classical Fisher bound. The paper does not supply that argument, nor does it specify the exact theorem in [29] that is being used. The stated hypothesis 'prior support contained in a finite interval' is also too weak: for a prior with atoms, ∫ sqrt(F) dtheta contributes zero on atoms and the bound would fail for a two-point prior. If Lemma 1 holds only for absolutely continuous priors, or only for accessible information, then the chain in Appendix B (marginal + conditional applications for Theorem 1, then Jensen/Cauchy–Schwarz for Theorem 2) collapses, and the K-task extension in Appendix E falls with it. The numerical 'phase transition' is downstream of the same bound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces a 'global quantum Fisher information matrix' (g-QFIM) and uses it to bound the Holevo information of a multi-task quantum ensemble. For K independent scalar tasks with product priors, it derives a separable capacity bound (Theorem 1) and a trace-based capacity bound (Theorem 2), with extensions to correlated priors and arbitrary K in the appendices. The authors illustrate the framework with a two-mode vacuum-single-photon phase-encoding model, claiming a structural phase transition at N=2/3, and they simulate attenuation, phase diffusion, and mode crosstalk to show how noise reshapes the geometry. The main theoretical results are presented as non-asymptotic, measurement-independent upper bounds on the Holevo information.","tokens_in":16835,"tokens_out":12139,"duration_ms":123880,"significance":"If the main theorems are correct, the manuscript provides compact, calculation-friendly, measurement-independent ceilings for multi-task quantum information and an interpretable geometric diagnostic (det A and κ(A)) for information allocation. The appendices give transparent chain-rule and Jensen-based derivations with no fitted parameters, which is a strength. However, the entire proof rests on the unproved one-dimensional Lemma 1, and the phase-transition claim goes beyond what the theorems establish. The numerical 'confirmation' also currently compares only QFIM-derived quantities, not the actual Holevo information under noise. With a proof or precise citation for Lemma 1, the framework could be a useful contribution to quantum ISAC and multi-task quantum networks.","major_comments":[{"comment":"Lemma 1 is the load-bearing foundation for Theorems 1, 2, 5, and 6, but it is stated without proof and without a precise theorem number or statement from [29]. The hypothesis 'prior support contained in a finite interval' is too weak as written: for a prior with atoms, the integral ∫√F dθ does not see the point masses, and the bound needs explicit regularity conditions. More importantly, the standard data-processing route controls only the accessible information sup_M I(Θ;Y), not the Holevo information χ of the ensemble. Since Lemma 1 is applied directly to χ in Eqs. (27), (32), (58), and (95), the authors must either supply a self-contained proof or cite the exact theorem from [29] and state explicitly that it applies to Holevo information, not merely to accessible information. If the lemma holds only for the accessible information, the main theorems collapse.","section":"Appendix B, Lemma 1"},{"comment":"The claimed 'structural phase transition' is not a consequence of Theorems 1–2. The criterion ∂det A/∂N=0, ∂²det A/∂N²<0, ∂κ(A)/∂N>0 is introduced ad hoc, and the specific value N=2/3 is an exact property of the symmetric toy model (η=1/2, W=2π), as can be seen from χ_true=H(1−N,N/2,N/2). The theorems only give upper bounds; they do not imply that det(A) or κ(A) determines the true capacity. The abstract's language 'reveal a structural phase transition' overstates the generality. Please either derive the criterion from the bounds or explicitly present it as a model-based observation with clear limitations.","section":"Results and 'General systems' criterion"},{"comment":"The numerical 'confirmation' under noise does not directly validate the theorems. The figures plot the task-budget curves (L1,L2) and the QFIM diagnostics det(A) and κ(A), but they do not compute or compare the actual Holevo information χ for the noisy channels. No error bars, confidence intervals, or Monte Carlo averaging are reported, and the 'realistic noise channels' are simple state-level ansätze. The statement that simulations 'confirm these predictions' is therefore not supported. Please either provide a direct comparison of the bound to χ under noise, or rephrase the claim as an illustrative demonstration.","section":"Non-ideal effect models and Figures 3–5"}],"minor_comments":[{"comment":"Figure 4 and Figure 5 are both captioned 'Tradeoff curve family under attenuation,' but the text describes phase diffusion and crosstalk, respectively. The captions should be corrected.","section":"Figure captions"},{"comment":"The main-text Theorem 1 states K=2, while the general-K result is relegated to Appendix E. Please clarify in the main text that the general case is in the appendix, and unify the notation for K=2 and K>2.","section":"Theorem 1 statement"},{"comment":"Several axis labels and quantities in Figures 2–5 are garbled or missing units (e.g., 'Information (nats)'), and the extracted figure text contains encoding artifacts. Please provide clean, legible figures.","section":"Figures and axis labels"}],"recommendation":"major_revision","confidential_remarks":"The decisive question is Lemma 1. If the authors cannot supply a proof or an exact citation showing that the bound applies to Holevo information (not just accessible information), I would not support acceptance. The phase-transition claim should also be substantially tempered. The manuscript's core idea is promising, but the current version is too conditional."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe paper introduces the g-QFIM, a prior-weighted integral of the quantum Fisher matrix, and uses it to bound the Holevo information of a multi-parameter ensemble. That is genuinely new: nobody has packaged the QFIM this way for multi-task capacity. Theorem 2's bound, χ ≤ K log(1 + (1/(2√K))√Tr A), is a clean expression, and the interpretation of det(A) as capability and κ(A) as allocability is a nice way to think about resource allocation. The toy-model simulations are sensible and the noise cases show distinct behavior. Credit where due: the internal derivations in Appendices C and E are correct elementary steps.\n\nThe soft spots are real, and one is load-bearing. Appendix B's Lemma 1 — the one-dimensional finite-support Holevo–QFI bound — is quoted without proof or a precise reference to the exact theorem in [29]. Everything else is built on it. The concern is not a formality: the usual Fisher-information bound of this type controls the accessible information for a fixed POVM, not the Holevo quantity, which can be larger. If [29] only proves the accessible version, Theorems 1 and 2 collapse. The lemma also needs an absolutely continuous prior; the text says 'support contained in a finite interval,' which is too weak if the prior is atomic. A referee should demand a complete proof or a page-and-line pointer to [29].\n\nThe 'structural phase transition' is oversold. It is demonstrated in one two-mode model with a uniform 2π prior, and the critical point N=2/3 comes from that model. The paper's own conclusion admits tightness and achievability are open, so the right frame is 'initial evidence in a toy model,' not a general phenomenon.\n\nMinor points: no code or data is released, and the figures have rendering artifacts in the text extract, though the numbers seem meaningful. The authors do not hide the limitations — they say the bound is not tight and achievability is open — which makes the overclaim in the abstract more annoying than dishonest.\n\nWho should read it: people working on quantum ISAC or multi-parameter estimation will find the g-QFIM perspective a useful tool, even if the bound is loose. It deserves a serious referee, specifically to verify Lemma 1 and the prior assumptions. I would not cite it in my own work until that lemma is nailed down.\n\nRecommendation: send to peer review, but the report should require a proof or a precise citation for Lemma 1, and a softening of the phase-transition language.","headline":"New g-QFIM bound for multi-task quantum systems is clean but rests on an unproved one-dimensional lemma; needs referee scrutiny.","tokens_in":17199,"tokens_out":6980,"would_cite":false,"duration_ms":64530,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45"],"pacs":[],"model":"deepseek-v4-flash","headline":"One matrix sets the ceiling on multi-task quantum capacity.","keywords":["quantum Fisher information matrix","Holevo information","multi-task quantum systems","integrated sensing and communication","quantum capacity bounds","information geometry","phase encoding","task tradeoff"],"falsifier":"Compute the exact Holevo information for the two-mode phase-encoding ensemble of Eq. (11) at resource values around N = 2/3 by optimizing over POVMs, and compare with the trace bound: any violation of χ ≤ 2 log(1 + (1/(2√2))√TrA) would refute Theorem 2. Similarly, testing Lemma 1 on a one-parameter state family with a discontinuous prior (for instance uniform on [0, 2π]) would settle whether the unproved one-dimensional bound actually holds.","tokens_in":16396,"feed_emoji":"⚛️","tokens_out":6128,"duration_ms":59657,"temperature":0.7,"pith_summary":"The paper claims that every multi-task quantum system—one physical carrier used for sensing and communication at once—has a single ceiling on how much classical information it can deliver, measured by the Holevo information (the information available in the quantum ensemble before any measurement is chosen). The ceiling is a Shannon-capacity-like bound written in terms of a single matrix, the global quantum Fisher information matrix, which aggregates the carrier's sensitivity to the task variables. The paper also claims that as physical resources grow, the system crosses a 'structural phase transition': before the crossover, extra resources add independent task capacity; after it, they concentrate information into a collective single-task mode. If true, this gives a unified way to compare sensing, communication, and computing tasks and to decide how to allocate a quantum resource among them.","feed_headline":"One matrix sets the ceiling on multi-task quantum capacity","feed_subtitle":"Measurement-independent and Shannon-like, the bound exposes a crossover where extra resources stop adding independent task capacity.","key_machinery":"The global quantum Fisher information matrix (g-QFIM), A = ∫ R(φ) J̃_Q(φ) R(φ) dφ over the normalized task-variable cube, where R is a diagonal matrix of prior weights. It aggregates the local quantum Fisher information of the received state family into a single positive-semidefinite matrix whose trace bounds total capacity and whose spectrum determines how information is distributed: det(A) measures the independent information volume and κ(A) measures anisotropy between task directions. The proof machinery also uses one-dimensional directional Fisher lengths L_i, Jensen inequalities, and a classical-register extension to reduce the multi-task Holevo information to single-parameter bounds.","core_discovery":"The central result is Theorem 2: for K independent tasks encoded in a quantum state, the joint Holevo information satisfies χ(Z1,...,ZK:B) ≤ K log(1 + (1/(2√K))√TrA), where A is the prior-weighted integral of the quantum Fisher information matrix over the task-variable space. Theorem 1 gives a sharper, separable bound in terms of task-directional Fisher lengths L_i. The same object A is then interpreted through det(A) as the system's independent multi-task capability and through κ(A) as its allocability; when extra resources make det(A) decrease while κ(A) grows, the information geometry becomes anisotropic and information concentrates into fewer collective modes. The paper demonstrates this","pith_inferences":["A testable extension: for Gaussian or coherent-state ensembles the same trace bound should be comparable with exact Holevo capacities, and one could search for the extremal state family that saturates the inequality.","The 'structural phase transition' is exhibited in a two-mode toy model; whether it sharpens into a genuine critical phenomenon as the number of modes and the resource budget grow is a question the paper leaves open.","The bound suggests a design rule the paper does not state explicitly: allocate physical resources so that the g-QFIM spectrum is as isotropic as possible, because anisotropy κ(A) is what wastes independent task capacity.","The same geometry likely extends to classical multi-antenna or classical ISAC systems, where an analogous Fisher-information matrix would play the role of A, connecting the quantum result to a broader capacity-region picture."],"forward_implications":["Multi-task capacity is not additive: sharing one carrier across K tasks tightens the ceiling because the bound scales as K log(1 + (1/(2√K))√TrA).","There is an optimal resource point where det(A) is maximal; beyond it, extra physical resources stop increasing independent task capacity and instead flow into a collective mode (the paper identifies N = 2/3 for its photonic model).","Different noise mechanisms leave distinct fingerprints: attenuation primarily shrinks total capacity, phase diffusion shrinks capacity while leaving allocability relatively intact, and mode crosstalk mainly destroys allocability by making the information geometry anisotropic.","The framework gives a measurement-independent criterion for designing integrated sensing-and-communication systems, including the possibility of deliberately hiding certain parameter directions for privacy or selective visibility."],"fun_headline_variants":["One matrix sets quantum multi-task capacity limit","Quantum multi-task ceiling from a single Fisher matrix","Information geometry caps multi-task quantum capacity","Phase transition in multi-task quantum capacity","Extra resources shift quantum info into fewer modes"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The main proofs reduce to a one-dimensional finite-support Holevo–QFI inequality stated without proof in Appendix B (Lemma 1); if that lemma fails for Holevo information or for the non-smooth finite-support priors the paper allows, both core bounds collapse, and the phase-transition claim is demonstrated only for a specific two-mode model with a uniform 2π prior.","fun_headline_variants_meta":{"raw":{"variants":["One matrix sets quantum multi-task capacity limit","Quantum multi-task ceiling from a single Fisher matrix","Information geometry caps multi-task quantum capacity","Phase transition in multi-task quantum capacity","Extra resources shift quantum info into fewer modes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000606,"raw_usage":{"total_tokens":2649,"prompt_tokens":717,"completion_tokens":1932,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":461,"completion_tokens_details":{"reasoning_tokens":1867}},"tokens_in":461,"tokens_out":1932,"duration_ms":13139,"temperature":1.0,"reasoning_tokens":1867,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T16:27:05.953418+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the exact Holevo information for the two-mode phase-encoding ensemble of Eq. (11) at resource values around N = 2/3 by optimizing over POVMs, and compare with the trace bound: any violation of χ ≤ 2 log(1 + (1/(2√2))√TrA) would refute Theorem 2. Similarly, testing Lemma 1 on a one-parameter state family with a discontinuous prior (for instance uniform on [0, 2π]) would settle whether the unproved one-dimensional bound actually holds.","supporting_citations":[],"review_version":1}