{"id":"6b2be30d-2262-4be9-8a3f-2594469db687","arxiv_id":"1908.02378","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Explicit symmetric 'gnu' probe states from permutation-invariant codes provably retain Heisenberg scaling of quantum Fisher information under a constant number of erasure errors or a single dephasing error.","lead":"The paper proposes explicit symmetric quantum probe states, built from permutation-invariant quantum error correction codes, that keep a quantum advantage in phase estimation even after a few qubits are lost or dephased. The authors prove analytical lower bounds on the quantum Fisher information showing Heisenberg scaling in the asymptotic limit and a practical advantage for roughly 50 to 200 qubits.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 9–10 does not establish robustness to a constant number of i.i.d. dephasing errors: for t=1 the stated lower bound is negative, and the proof of (V.12) is not justified.","rationale":"Good-faith reading: the paper's erasure results are explicit and detailed. Lemma 1, Theorem 2, and Theorem 4 give a parameter-dependent formula whose asymptotics are evaluated; nothing in the erasure section appears unsound. The dephasing result for D_λ is also a genuine result: Theorem 7 gives an explicit lower bound, and Corollary 8 shows a positive fraction of N^2 for λ=1/2. The paper honestly labels its i.i.d. analysis as an approximation and admits the lack of a full practical proposal. The problem is that the i.i.d. extension is presented as a corollary and used in the abstract and introduction to claim robustness to a constant number of dephasing errors; this is the part that does not hold up. The underived constant in (V.12) and the negative bound at t=1 mean that the proof does not establish the claimed result. This is a correctness risk in a supporting but advertised claim, not fraud or circularity. The natural verdict is CONDITIONAL: the core single-error and erasure claims may stand, but the i.i.d. dephasing claim needs either repair or removal. Since the reader already reached this verdict, no change is needed.","tokens_in":19378,"tokens_out":14569,"duration_ms":157052,"concrete_test":"Compute exactly, for |ϕ2> with n=2, g=50 (N=200) and p=1/N, the quantities A=||[σ_1,H]||_1, B=||[ε_1,H]||_1, and τ_1 using the closed-form expressions from Lemma 5 and Theorem 7. Test whether (A−B)^2 ≥ A^2 − 2N^2τ_1; if this fails, (V.12) is false and the proof of Corollary 9 is invalid. Also evaluate the Corollary 9 lower bound itself: if it is negative at t=1 (equivalently, Corollary 10's RHS is negative), the corollary cannot support a constant-number-of-errors claim regardless of whether (V.12) happens to hold for this instance.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Corollary 9 is the only bridge from the artificial single-error channel D_λ to the standard i.i.d. dephasing model. Its proof replaces σ by σ_1 and invokes (V.12), ||[σ,H]||_1^2 ≥ ||[σ_1,H]||_1^2 − 2N^2τ_1. This inequality does not follow from the bounds stated immediately above it. The triangle and Hölder bounds give only ||[ε_1,H]||_1 ≤ 2Nτ_1 and ||[σ_1,H]||_1 ≤ 2N; combining them yields at best ||[σ,H]||_1^2 ≥ ||[σ_1,H]||_1^2 − 8N^2τ_1, not the advertised 2N^2τ_1. The constant matters because the discarded tail probability τ_1 is O(t), so the penalty is O(tN^2). Even granting (V.12), Corollary 10's lower bound e^{−2t}/(2n) − 2et is negative for every fixed integer t≥1: for n=2, t=1 it is e^{−2}/4 − 2e ≈ −5.40. The bound becomes positive only for t below roughly 0.04 for n=2 and O(1/n) in general, so it proves Heisenberg scaling only for an expected error number that is not a fixed positive constant. Thus the abstract's 'constant number of dephasing errors' is supported by the D_λ single-error results (Theorem 7 and Corollary 8), but not by the i.i.d. dephasing extension; Corollary 9–10 either needs a corrected derivation and a much tighter tail/penalty analysis or should be restated for t→0.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes explicit symmetric 'gnu' probe states, drawn from the permutation-invariant quantum code family of Ref. [29], and asks whether they retain a metrological advantage after small, constant numbers of erasure or dephasing errors. For erasures, the authors derive an explicit form for the partially traced density matrix (Lemma 1) and a closed-form lower bound on the QFI (Theorem 2), then evaluate the leading large-g asymptotics for t=1,2 erasures (Theorem 4), obtaining a constant fraction of Heisenberg scaling when n is fixed. For dephasing, they analyze a single-error channel D_lambda using Dicke-state inner products expressed through Krawtchouk polynomials (Lemmas 5-6), obtaining a lower bound in Theorem 7 and its asymptotic form in Corollary 8; they then attempt to extend this to i.i.d. dephasing via an at-most-one-error approximant (Corollaries 9-10). The central claim is that the probe states yield a quantum advantage in the NISQ regime and recover Heisenberg scaling asymptotically under a constant number of erasure errors and a constant number of dephasing errors.","tokens_in":19768,"tokens_out":11841,"duration_ms":123218,"significance":"If the technical results held as stated, the paper would provide one of the first explicit, code-inspired state families that provably retain Heisenberg scaling under a bounded number of erasures, complementing the random-state results of Oszmaniec et al. The erasure section is a genuine calculational contribution: the partial-trace formula and the resulting lower bound are explicit and do not rely on fitting or on unproved error-correction folklore. The dephasing analysis for D_lambda is also a concrete calculation whose finite-N and asymptotic forms can be used directly. The main value is therefore the explicit construction and the transparent lower-bound methodology. The main weakness is that the i.i.d. dephasing extension, which is needed for the abstract's 'constant number of dephasing errors' claim, is not established by the proof as written.","major_comments":[{"comment":"The asserted inequality ||[sigma,H]||_1^2 >= ||[sigma_1,H]||_1^2 - 2N^2 tau_1 is not derived by the bounds preceding it. The triangle inequality gives ||[sigma,H]||_1 >= ||[sigma_1,H]||_1 - ||[epsilon_1,H]||_1, and the correct Holder bound is ||epsilon_1 H||_1 <= N tau_1, not N, so ||[epsilon_1,H]||_1 <= 2N tau_1. Since ||[sigma_1,H]||_1 <= 2N, squaring can only yield, in the worst case, ||[sigma,H]||_1^2 >= ||[sigma_1,H]||_1^2 - 8N^2 tau_1, not the advertised 2N^2 tau_1. This factor is material because tau_1 = O(t), and the threshold in Corollary 10 depends linearly on this coefficient.","section":"Section V, Eq. (V.12)"},{"comment":"As stated, the lower bound e^{-2t}/(2n) - 2te is negative for every integer t >= 1; for n=2, t=1 it equals e^{-2}/4 - 2e, which is approximately -5.40. Thus Corollary 10 does not establish robustness to a constant number of i.i.d. dephasing errors. Even granting Eq. (V.12), the bound becomes positive only for t below roughly 1/(4en), i.e. for an expected number of errors that is not a fixed positive constant. This is inconsistent with the abstract's 'constant number of dephasing errors' and with the introduction's stated regime. The authors should either provide a corrected tail/penalty analysis or explicitly restate the i.i.d. result as a t -> 0 statement.","section":"Section V, Corollary 10"},{"comment":"There is an internal inconsistency in the stated asymptotic regime for i.i.d. dephasing. Section V says 'consider the limit of large N where the average number of phase errors is held constant,' which gives p = t/N, i.e. p decays as 1/N. The introduction, by contrast, says the probability of dephasing per qubit 'approaches zero faster than the reciprocal of the number of qubits,' which would require t -> 0. These are different claims, and the formal result in Corollary 10 corresponds to the first, while the positivity of the bound requires the second. The paper should state unambiguously which regime is being claimed and adjust the abstract accordingly.","section":"Section V and Introduction"}],"minor_comments":[{"comment":"The text says the dephasing bound is 'given explicitly in Theorem 2' and later references an unresolved 'Theorem ??'; the correct statement is Theorem 7, and the placeholder should be fixed.","section":"Section II, dephasing paragraph"},{"comment":"The proof of Lemma 3 takes limits as g -> infinity but repeatedly writes lim_{n->infinity}; this is notationally incorrect, although the intended calculation is clear from the context.","section":"Lemma 3 proof"},{"comment":"There are several typos: 'Explictly' in Eq. (V.1), 'Thereom' in the caption of Figure 2, and 'completees' in the proof of Lemma 5.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper's core erasure calculations and the D_lambda dephasing calculation are worth publishing, but the abstract overclaims the i.i.d. dephasing robustness. The self-citation to Ref. [29] is justified because the probe states are literally taken from that code family, and the QFI bounds are direct computations rather than appeals to error correction. The authors should be asked to either repair the i.i.d. dephasing proof or explicitly demote that claim to a small-t asymptotic statement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know up front. The explicit symmetric probes for erasure metrology are solid and new; the i.i.d. dephasing extension is not. I agree with the stress-test note: Corollaries 9 and 10 do not establish robustness to a constant number of dephasing errors.\n\nWhat the paper actually contributes: it takes the permutation-invariant “gnu” code states from Ref. [29], defines the logical-plus probe |ϕ_u⟩, and proves analytic lower bounds on the QFI under t erasures. Lemma 1 and Theorems 2 and 4 are detailed and internally consistent. The asymptotic erasure bounds recover Heisenberg scaling up to a constant, and the numerical plots show advantage in the 50–200 qubit range. That part deserves credit. The Krawtchouk/Dicke machinery in Section IV is also explicit and checkable. The self-citation to [29] is not a problem: the QFI bounds are computed directly from (II.4)–(II.6), not from the code’s error-correction capability.\n\nThe soft spot is Section V. Theorem 7 and Corollary 8 are fine for the artificial channel D_λ that allows at most one phase error. But the step to i.i.d. dephasing rests on (V.12), which is asserted without proof. From the inequalities displayed just above it, the best you can get is ||[σ,H]||_1^2 ≥ ||[σ_1,H]||_1^2 − 8N²τ_1, unless you have an unstated bound on ||[σ_1,H]||_1 that caps it at N/2. Even taking the claimed 2N²τ_1, Corollary 10 gives e^{−2t}/(2n) − 2et, which is negative for every fixed integer t ≥ 1. So the lower bound proves Heisenberg scaling only for t below about 0.04 in the n=2 case. That is not a “constant number of dephasing errors” in any standard sense, and the abstract overstates the result. The single-error D_λ result is still interesting, but the i.i.d. corollaries need either a corrected derivation with a much tighter tail analysis or a substantially more modest claim.\n\nThe paper is otherwise honest about its limitations: no measurement protocol, no state-preparation/readout details, and no symmetric logarithmic derivative calculation. No code or data ships, but the calculations are reproducible from the text.\n\nBottom line: worth a serious referee. I would expect major revision of the i.i.d. dephasing section, and I would not let the abstract claim “constant number of dephasing errors” until that is fixed. The erasure half should survive.","headline":"Erasure results are solid; the i.i.d. dephasing extension is overclaimed and (V.12) doesn't follow as written.","tokens_in":20303,"tokens_out":7671,"would_cite":true,"duration_ms":82542,"reading_group":"yes","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 proves that explicit symmetric probe states drawn from permutation-invariant quantum codes retain Heisenberg scaling of the quantum Fisher information after constant erasure or dephasing errors, giving a quantum advantage in…","keywords":["quantum metrology","quantum Fisher information","permutation-invariant codes","Dicke states","Heisenberg scaling","erasure errors","dephasing noise","NISQ"],"falsifier":"Compute the full quantum Fisher information of the dephased state $\\sigma$ in Eq. (V.1) for a small but finite $t=pN$, say $t=1$ with $n=2,g$ large, and compare it with the claimed lower bound $(e^{-2t}/(2n)-2te)N^2$; if the exact value divided by $N^2$ falls below the claimed coefficient, or if inequality (V.12) fails for that explicit $\\sigma_1$, the Heisenberg-scaling claim for i.i.d. dephasing collapses.","tokens_in":19179,"feed_emoji":"⚛️","tokens_out":11388,"duration_ms":107213,"temperature":0.7,"pith_summary":"The paper claims that explicitly constructed symmetric probe states—gnu probe states, formed as the equal superposition of the two logical codewords of a permutation-invariant quantum code—remain useful for quantum metrology after a constant number of erasure errors or after a single dephasing error. For erasures, the authors prove an analytical lower bound on the quantum Fisher information of the partially traced state, and in the limit of many qubits this bound grows as a constant times $N^2$, recovering Heisenberg scaling. For dephasing, they obtain a similar lower bound for a channel that applies at most one phase error, and for i.i.d. dephasing with a fixed expected number of errors they derive an asymptotic bound with a positive Heisenberg coefficient when that expected number is small. If these claims are right, near-term noisy quantum sensors could achieve a provable quantum advantage without active error correction, and the paper gives an explicit family of states to prepare.","feed_headline":"Noisy symmetric probes still reach Heisenberg scaling","feed_subtitle":"Explicit states from permutation-invariant codes keep a quantum advantage after erasure and dephasing errors.","key_machinery":"The load-bearing object is the gnu probe state $|\\phi_u\\rangle$, an equal superposition of the two logical codewords of a permutation-invariant code with parameters $g,n,u$, expanded as a superposition of Dicke states $|D^N_w\\rangle$. Two structural ingredients carry the argument: the partial trace of $|\\phi_1\\rangle$ over $t$ erased qubits has an explicit sparse representation in terms of orthogonal vectors $|\\theta_u\\rangle$, and inner products of Dicke states with products of Pauli $Z$ operators are expressed through Krawtchouk polynomials. These reduce the commutator lower bound on the QFI to binomial sums of the form $\\sum_{j=0}^n \\binom{n}{j} j^x$, which are evaluated in closed form.","core_discovery":"The central claim is that the gnu probe state $|\\phi_u\\rangle = (|0_L\\rangle+|1_L\\rangle)/\\sqrt{2}$ from Eq. (II.6), which lies inside a permutation-invariant quantum code with parameters $g,n,u$ and $N=gnu$ qubits, remains a good metrological resource after noise. With $u=1$ and $t$ erased qubits, the QFI is lower-bounded by the explicit expression in Theorem 2; for one erasure and large $g$ this becomes $\\frac{n-1}{n^2}(N^2+(n-2)N-(n-1))$, and for two erasures a similar quadratic, so the asymptotic scaling is Heisenberg, $O(N^2)$. With $u=2$, the dephasing channel $D_\\lambda$ (which applies zero or one phase error) yields the closed-form lower bound of Theorem 7, and for large $g$ the QFI divided by $N^2$ is at least $\\frac{\\lambda^2}{2n}+\\frac{\\lambda(1-\\lambda)(n-1)}{4n^2}+\\frac{(1-\\lambda)^2(n^3+n-2)}{32n^4}$; at $\\lambda=1/2$ this coefficient is $25/(128n)-1/(16n^2)+1/(128n^3)-1/(64n^4)$, whereas a GHZ state would become classically useless. A further approximation for i.i.d. dephasing with expected error count $t=pN$ gives an asymptotic lower bound of $(e^{-2t}/(2n)-2te)N^2$. All bounds come from evaluating the generator lower bound $\\|[\\rho,H]\\|_1^2\\ge\\|[\\rho,H]\\|_2^2$ through the sparse structure of the partially traced state.","pith_inferences":["The dephasing result is more delicate than the erasure result: the bound $(e^{-2t}/(2n)-2te)N^2$ is positive only for $t$ below roughly $0.03$, so in the i.i.d. case 'constant number of dephasing errors' means a very small constant; a direct check of inequality Eq. (V.12) on explicit small states would settle whether the stated coefficient is reliable.","The same Dicke/Krawtchouk machinery should extend to other Pauli noise models such as amplitude damping or depolarizing noise, at the cost of evaluating inner products with non-diagonal Pauli strings; the paper lists these as open directions.","Because the probe states are symmetric and explicit, they can in principle be generated by Dicke-state preparation circuits, but whether the scheme is practical depends on the overhead of that preparation, which the paper does not quantify.","The generator lower bound $\\|[\\rho,H]\\|_2^2$ is looser than the full quantum Fisher information, so the stated constants are conservative; the true advantage could be larger than reported."],"forward_implications":["With one erasure, the lower bound $O(N^2)$ beats the classical shot-noise limit $O(N)$ once $N$ is sufficiently large, so a fixed erasure rate leaves a quantum advantage in the NISQ regime.","The same probe state with $u=2$ remains useful for dephasing even in the worst case $\\lambda=1/2$, where a GHZ state would be reduced to a classically useless mixture.","For i.i.d. dephasing with an expected number $t$ of errors held constant, Heisenberg scaling survives whenever $t$ is small enough that $e^{-2t}/(2n)-2te>0$.","The authors conjecture that any permutation-invariant code detecting at least one error is good for metrology under erasures, which would place the construction inside a broad family of codes."],"supporting_citations":[{"why":"Supplies the family of permutation-invariant quantum codes whose logical codewords define the gnu probe states.","marker":"[29]"},{"why":"Motivates explicit symmetric probes by showing random symmetric states are almost surely useful under finite particle loss, and provides the commutator lower-bound technique.","marker":"[19]"},{"why":"Establishes that i.i.d. noise destroys Heisenberg scaling, defining the asymptotic regime that the probe states must evade.","marker":"[18]"},{"why":"Shows how permutation-invariant codes can be initialized with superconducting charge qubits, supporting the state-preparation premise.","marker":"[33]"}],"fun_headline_variants":["Symmetric probe states resist erasure and dephasing","Explicit states achieve Heisenberg scaling under noise","Permutation-invariant codes for robust metrology","Noise-robust quantum metrology with explicit states","Quantum advantage persists after constant errors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step is the approximation in Corollary 9, where the fully dephased state is replaced by the state with at most one phase error, and an asserted inequality bounds the damage done by the discarded noise tail; that inequality is not derived in full and may be too optimistic.","fun_headline_variants_meta":{"raw":{"variants":["Symmetric probe states resist erasure and dephasing","Explicit states achieve Heisenberg scaling under noise","Permutation-invariant codes for robust metrology","Noise-robust quantum metrology with explicit states","Quantum advantage persists after constant errors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000625,"raw_usage":{"total_tokens":2944,"prompt_tokens":1050,"completion_tokens":1894,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":666,"completion_tokens_details":{"reasoning_tokens":1822}},"tokens_in":666,"tokens_out":1894,"duration_ms":12847,"temperature":1.0,"reasoning_tokens":1822,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:47:53.575471+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the full quantum Fisher information of the dephased state $\\sigma$ in Eq. (V.1) for a small but finite $t=pN$, say $t=1$ with $n=2,g$ large, and compare it with the claimed lower bound $(e^{-2t}/(2n)-2te)N^2$; if the exact value divided by $N^2$ falls below the claimed coefficient, or if inequality (V.12) fails for that explicit $\\sigma_1$, the Heisenberg-scaling claim for i.i.d. dephasing collapses.","supporting_citations":[{"cited_title":"Pauli Exchange Errors in Quantum Compu- tation,","cited_arxiv_id":null,"evidence_quote":"Supplies the family of permutation-invariant quantum codes whose logical codewords define the gnu probe states."},{"cited_title":"Ancilla- free quantum error correction codes for quantum metrology,","cited_arxiv_id":null,"evidence_quote":"Establishes that i.i.d. noise destroys Heisenberg scaling, defining the asymptotic regime that the probe states must evade."},{"cited_title":"Permutation-invariant qudit codes from polyno- mials,","cited_arxiv_id":null,"evidence_quote":"Shows how permutation-invariant codes can be initialized with superconducting charge qubits, supporting the state-preparation premise."}],"review_version":1}