{"id":"ff5429bb-56d6-4daf-b026-49b951f9ca69","arxiv_id":"2601.12475","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors introduce a conditional quantum Fisher information whose average equals the standard QFI, and use it to derive single-trajectory speed limits with a possible negative interference term.","lead":"Researchers define a trajectory-level quantum Fisher information that tracks how sensitive a single monitored run of a quantum experiment is to a parameter, and show it decomposes into classical population changes, quantum basis rotations, and an interference term that can go negative. The work connects quantum metrology to stochastic thermodynamics, aiming at single-shot sensitivity bounds and adaptive feedback.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"CQFI is defined by two non-equivalent expressions (Eq. 9 vs Eq. A7); the Appendix A derivation relies on an invalid Cauchy-Schwarz step, so the central definition and its link to SFI are unsupported.","rationale":"The reader's weakest assumption pinpoints the central flaw: two inequivalent definitions of CQFI are conflated, and the derivation from the classical SFI is invalid because the Cauchy-Schwarz step uses the wrong second factor. I agree that this is the most load-bearing concern. The paper does contain a well-defined object—the state-conditioned form f_{Q,γ}=⟨ψ_γ|L^2|ψ_γ⟩—which satisfies the averaging property E[f]=Tr(ρL^2) for any faithful unravelling, and the decomposition into incoherent and coherent parts plus a cross-term is algebraically consistent for this form. The cross-term negativity is also a real mathematical feature. However, the paper explicitly claims that this state-conditioned form is equivalent to the ratio defined in Eq. (9) and that it arises from the stochastic Fisher information via the Cauchy-Schwarz bound; both of these claims are false. Without the ratio definition or a valid derivation connecting the state-conditioned form to the SFI, the central claim that CQFI is the quantum generalization of SFI is unsupported. The trajectory-level speed limits and geometric interpretations built on this CQFI inherit the ambiguity. The numerical validation of the speed limit is also weak: Fig. 4 tests an observable whose rate of change is zero, so the inequality is trivially satisfied. These issues justify the reader's REJECT verdict; no adjustment is needed.","tokens_in":15617,"tokens_out":15502,"duration_ms":129744,"concrete_test":"Evaluate both candidate definitions for the driven thermal qubit example in Sec. IV: at several times t, compute (a) ⟨ψ_γ(t)|L_t^2|ψ_γ(t)⟩ and (b) Tr(|ψ_γ⟩⟨ψ_γ| L_t^2 ρ_t)/Tr(|ψ_γ⟩⟨ψ_γ| ρ_t) for the same trajectory state |ψ_γ(t)⟩ and ensemble ρ_t. If they differ (as they must whenever ρ_t is not pure and |ψ_γ⟩ is not an eigenstate of ρ_t), the claimed equivalence in Sec. II.B fails. Additionally, re-derive the Cauchy-Schwarz bound in Appendix A with the correct B=√ρ L√Π to confirm that the upper bound is Tr(Π L ρ L)/Tr(Πρ), not the expression in Eq. (9).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central object, the CQFI, is introduced in Eq. (9) as f_{Q,α}=Tr(Π_α L^2 ρ)/Tr(Π_α ρ). In Sec. II.B it is claimed that 'an equivalent representation' is f_{Q,α}=Tr(Π_α L^2). These are not equivalent: for a rank-one projector and mixed ρ, the two expressions generally differ. For example, with ρ=|+⟩⟨+| and Π=|0⟩⟨0|, the ratio form gives ⟨+|L^2|+⟩ while the state-conditioned form gives ⟨0|L^2|0⟩, which are unequal for any L that does not commute with the projector. The derivation in Appendix A purportedly obtains Eq. (9) from the classical SFI via Cauchy-Schwarz, but the choice A=√ρ√Π, B=√ρ L√Π yields Tr(A†A)=Tr(Πρ) and Tr(B†B)=Tr(Π L ρ L), not Tr(Π L^2 ρ). Thus the bound |Tr(ΠLρ)|^2 ≤ Tr(Πρ)Tr(ΠL^2ρ) is unjustified. All trajectory-level simulations in Sec. IV use the state-conditioned form f_{Q,γ}=⟨ψ_γ|L^2|ψ_γ⟩, which is not Eq. (9) for mixed ρ. Consequently the decomposition of Sec. II.B/Appendix B and the trajectory-level speed limits (Eq. 32) apply to a different quantity than the one derived from the SFI. This equivocation is load-bearing: the paper's central claim that CQFI generalizes the classical stochastic Fisher information depends on this derivation and on the equivalence of the two definitions.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a trajectory-level 'Conditional Quantum Fisher Information' (CQFI), intended to generalize classical stochastic Fisher information to quantum trajectories. The CQFI is defined in Eq. (9) as a ratio involving a POVM element and the ensemble SLD operator, and it is claimed to average to the standard QFI and to decompose into incoherent, coherent, and interference contributions. The authors further define stochastic length and action, derive trajectory-level speed limits, and validate the formalism on a driven thermal qubit and on displaced Gaussian states. The central advertised results are the negative cross-term as a single-trajectory interference witness and the recovery of ensemble QFI from trajectory averages.","tokens_in":16070,"tokens_out":9562,"duration_ms":103041,"significance":"If the central construction were sound, the paper would offer a useful bridge between stochastic thermodynamics and quantum metrology, with potential applications in adaptive metrology and single-shot speed limits. The authors provide explicit analytic decompositions, numerical simulations with 5×10^4 trajectories, and no fitted parameters, and the averaged state-conditioned quantity does indeed converge to the QFI in their examples. However, the central definition is internally inconsistent: two inequivalent expressions for the CQFI are used interchangeably, and the derivation connecting the CQFI to the classical stochastic Fisher information relies on an invalid Cauchy-Schwarz step. The numerical and speed-limit results therefore concern a different object from the one formally defined and derived, undermining the paper's main claim as stated.","major_comments":[{"comment":"The paper defines the CQFI in Eq. (9) as f_{Q,α}=Tr(Π_α L² ρ)/Tr(Π_α ρ), then states that 'an equivalent representation' is f_{Q,α}=Tr(Π_α L²)=⟨α|L²|α⟩. These are not equivalent for a generic ρ. For example, with ρ=|+⟩⟨+| and Π=|0⟩⟨0|, Eq. (9) evaluates to √2 ⟨0|L²|+⟩, while Eq. (A7) is ⟨0|L²|0⟩. The full spectral decomposition in Appendix B and the trajectory-level expressions in Eqs. (35)–(37) are built on the state-conditioned form, and the simulations use f_{Q,γ}=⟨ψ_γ|L²|ψ_γ⟩. Thus the quantity whose average is shown in Fig. 2(a) and which enters the speed limits is not the quantity defined in Eq. (9). This equivocation is load-bearing for the central claim.","section":"§II.B, Eq. (9) vs Eq. (A7)"},{"comment":"The derivation of Eq. (9) from the classical stochastic Fisher information is invalid. With A=√ρ√Π and B=√ρ L√Π, the Cauchy-Schwarz inequality gives |Tr(ΠLρ)|² ≤ Tr(Πρ)Tr(Π L ρ L), not |Tr(ΠLρ)|² ≤ Tr(Πρ)Tr(Π L² ρ). The latter requires additional commutation assumptions that are neither stated nor satisfied in general. Consequently the bound and the 'saturation' claim leading to Eqs. (A4)–(A5) are unjustified. The asserted link between the CQFI and the classical SFI therefore rests on an incorrect inequality.","section":"Appendix A"},{"comment":"The single-trajectory speed limit is asserted without a proof. The standard observable-speed bound uses the QFI of the state being evolved; here f_Q[ρ_γ] is constructed from the ensemble SLD, not from the QFI of the instantaneous pure trajectory state. Since the stochastic Schrödinger equation (24) contains jump terms, a derivation is needed for |Tr(O dρ_γ/dt)| ≤ Δ_{ρ_γ}O sqrt(⟨ψ_γ|L²|ψ_γ⟩). The numerical test in Fig. 4 is trivial because the chosen observable H_RWA has ∂_t ⟨H_RWA⟩=0 identically, so it does not validate the inequality.","section":"§IV, Eqs. (31)–(32)"}],"minor_comments":[{"comment":"The detailed-balance condition L_k^- = L_k^+ e^{-Δs_k/2} is introduced but never used in the subsequent derivation or simulation. Please either connect it to the jump operators in Eq. (33) or remove it.","section":"§III.B, Eq. (24)"},{"comment":"The text itself notes that the speed-limit inequality is trivially satisfied for O=H_RWA because the rate of change is zero. A nontrivial observable should be used to provide a meaningful numerical check of Eq. (32).","section":"§IV, Fig. 4"},{"comment":"The Gaussian CQFI in Eq. (C7) uses the ratio definition, whereas the main trajectory formalism and Appendix B use the state-conditioned definition. This further illustrates the need to fix a single consistent definition before the results can be evaluated.","section":"Appendix C.2"}],"recommendation":"reject","confidential_remarks":"The central construction conflates two inequivalent definitions and the appendix derivation contains a load-bearing error. These are not merely presentation issues: the trajectory-level results, the speed limits, and the claimed generalization of the classical SFI all depend on the invalid equivalence. I would not invite a revision unless the authors commit to a single definition and rework the derivations and numerical analysis accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the manuscript's central object, the CQFI, is defined two incompatible ways, and the derivation that connects it to classical stochastic Fisher information relies on a false Cauchy-Schwarz step. The numerics use one of the definitions, so the simulations are likely correct but about a different quantity than the one derived. The idea is worth rescuing, but this version isn't.\n\nWhat's new: the decomposition of a trajectory-level QFI into incoherent and coherent terms plus a cross-term that can be negative is a nice idea. For the state-conditioned quantity f_Q = ⟨ψ|L²|ψ⟩, the average over an unraveling does recover Tr(ρL²) = QFI, and the three-term decomposition follows directly from splitting L into diagonal and off-diagonal parts. The negative cross-term as a witness of destructive interference is an interesting observation. The trajectory-level speed limits are straightforward applications of Cauchy-Schwarz to this object, and they appear correct.\n\nWhere it falls apart: Eq. (9) defines f_Q,α = Tr(Π_α L² ρ)/Tr(Π_α ρ). The text says this is equivalently ⟨α|L²|α⟩ (Eq. A7). That is not true: the two expressions differ generically for mixed states and even for pure states when the projector doesn't align with the state. So the paper equivocates between a conditional average and a state-conditioned expectation.\n\nWorse, Appendix A tries to derive Eq. (9) as an upper bound on the classical SFI ι = (∂ log p / ∂θ)². The Cauchy-Schwarz step chooses A=√ρ√Π, B=√ρ L√Π, which gives Tr(B†B)=Tr(Π L ρ L), not Tr(Π L² ρ). And the stated inequality |Tr(ΠLρ)|² ≤ Tr(Πρ)Tr(ΠL²ρ) is simply false; a counterexample is ρ=|+⟩⟨+|, Π=|0⟩⟨0|, L=σx, where the left side is 0.1875 and the right side is 0.0625. So the bound does not hold, and the CQFI is not established as a saturated version of the SFI.\n\nThe simulations in Sec. IV use f_Q,γ = ⟨ψ_γ|L²|ψ_γ⟩, which is the state-conditioned form. That is a valid and interesting quantity, and the numerical checks of the decomposition and speed limits are believable. But the paper's narrative that this object arises from a quantum generalization of the classical SFI is unsupported.\n\nBottom line: the authors have a decent core idea. The fix is to define the CQFI cleanly as the state-conditioned quantity, rework the motivation, and remove or repair the invalid derivation. This is not a desk-reject; it deserves peer review, and a good referee will require major revision. I'd bring it to the reading group as a cautionary example about definitional consistency.","headline":"The paper has a genuinely interesting idea—trajectory-level QFI with a negative interference term—but the definition is equivocal and the derivation from the classical SFI is invalid, so the central claim doesn't hold as written.","tokens_in":16551,"tokens_out":6815,"would_cite":false,"duration_ms":58754,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper introduces a conditional quantum Fisher information that turns quantum metrology from an ensemble average into a per-trajectory random variable, with a negative interference term that witnesses non-classicality.","keywords":["conditional quantum Fisher information","trajectory-level quantum speed limit","stochastic information geometry","symmetric logarithmic derivative","quantum jump unraveling","interference cross-term","single-shot metrology","quantum thermodynamics"],"falsifier":"Evaluate the ratio form and the state-conditioned form for the same POVM element on a state, e.g., ρ = |+⟩⟨+| and Π = |0⟩⟨0|; if the two numbers differ, the average-to-QFI property and speed limits demonstrated numerically belong to the state-conditioned object rather than the ratio-defined CQFI. Re-running the driven-qubit simulation with the ratio definition and checking whether ⟨f⟩ still equals F_Q would settle whether the reported trajectory-level results survive under the paper's primary definition.","tokens_in":15567,"feed_emoji":"⚛️","tokens_out":7451,"duration_ms":68389,"temperature":0.7,"pith_summary":"The paper introduces the conditional quantum Fisher information (CQFI) as a trajectory-level counterpart to the standard QFI: each measurement record gets its own information value, and averaging over records returns the usual ensemble QFI. The CQFI splits into an incoherent (population) term, a coherent (basis-rotation) term, and an interference cross-term that disappears in the ensemble average but can be negative on individual runs. The paper uses the CQFI to build a stochastic information geometry—stochastic length, action, and a speed limit—that bounds evolution along single quantum trajectories, not just on average. It validates the construction with quantum-jump trajectories of a driven thermal qubit and shows the trajectory-level bounds respect the expected thermodynamic hierarchy. If the claims hold, quantum metrology and stochastic thermodynamics gain access to single-shot sensitivity and fluctuations hidden by ensemble averaging.","feed_headline":"Single trajectories get their own quantum speed limit","feed_subtitle":"The conditional QFI is a random variable that recovers the standard QFI on average and bounds each run's speed.","key_machinery":"The central object is the conditional quantum Fisher information (CQFI), defined as Tr(Π_α L²ρ)/Tr(Π_αρ) in terms of the symmetric logarithmic derivative L of the state; in the trajectory simulations it is evaluated through the state-conditioned form ⟨α|L²|α⟩. Its role is to be a stochastic metric: it assigns a sensitivity value to a single measurement outcome α and to the instantaneous pure state of a trajectory. The argument is carried by the spectral decomposition of L into diagonal and off-diagonal parts, which splits the CQFI into incoherent, coherent, and interference contributions, and by the Cauchy-Schwarz inequality that converts the squared second moment into bounds on observable r","core_discovery":"The paper's central discovery is a family of identities turning the quantum Fisher information into a trajectory-level random variable. Defining the CQFI by f_Q,α = Tr(Π_α L²ρ)/Tr(Π_αρ), with L the symmetric logarithmic derivative and Π_α a POVM element, the authors show it averages to F_Q = Tr(ρL²) over measurement outcomes. For projective measurements they expand it as f_Q,α = f^IC_Q,α + f^C_Q,α + f^X_Q,α, where the incoherent part comes from population derivatives, the coherent part from eigenbasis rotations, and the cross-term from their interference; the cross-term can be negative along single trajectories while vanishing on average. From the induced metric they define single-trajectory","pith_inferences":["If the per-trajectory speed limit is stable under feedback, the CQFI could be used for real-time metrology—e.g., stopping or switching measurement basis when a trajectory's sensitivity spikes; the paper discusses this as an application but does not simulate a closed-loop protocol.","The negative cross-term could be tested as an empirical non-classicality witness by correlating its occurrence with other single-shot indicators such as contextuality or negativity in phase-space distributions.","The paper's two definitions of the CQFI—the ratio form and the state-conditioned form—are asserted to be equivalent; a direct check for arbitrary mixed states is a natural extension, since the reported numerics use the state-conditioned form.","A multiparameter version of the CQFI could yield a conditional Fisher matrix and trajectory-level versions of the quantum Cramér–Rao bound, extending the framework beyond single-parameter time estimation."],"forward_implications":["Averaging the CQFI over measurement outcomes recovers the standard QFI, so single-shot information remains consistent with the ensemble metrological bound.","The trajectory-level speed limit ∫|ȯ_γ|/Δ_γ O ≤ 2ℓ holds for each realization, potentially giving tighter bounds than ensemble QSLs for rare, information-rich trajectories.","The interference cross-term can be negative along a single trajectory and vanishes on average, offering a local, trajectory-level witness of destructive interference between classical and quantum information channels.","The stochastic length and action satisfy ℓ² ≤ j for every trajectory, extending thermodynamic length and action to individual quantum realizations.","In the Gaussian force-sensing example, the CQFI coincides with the QFI and is independent of the measurement outcome, showing a regime where trajectory-level and ensemble sensitivities agree."],"fun_headline_variants":["Each quantum run gets its own speed limit","Conditional Fisher info yields per-trajectory speed limits","CQFI: a trajectory-level quantum speed limit","Per-run quantum speed limit from conditional Fisher info","Negative cross-term per trajectory reveals quantum interference"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The argument assumes that the ratio definition of the CQFI and the state-conditioned form ⟨α|L²|α⟩ give the same value for every mixed state; the trajectory-level calculations and numerical checks use the state-conditioned form, so that equivalence is load-bearing.","fun_headline_variants_meta":{"raw":{"variants":["Each quantum run gets its own speed limit","Conditional Fisher info yields per-trajectory speed limits","CQFI: a trajectory-level quantum speed limit","Per-run quantum speed limit from conditional Fisher info","Negative cross-term per trajectory reveals quantum interference"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001236,"raw_usage":{"total_tokens":4892,"prompt_tokens":707,"completion_tokens":4185,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":451,"completion_tokens_details":{"reasoning_tokens":4113}},"tokens_in":451,"tokens_out":4185,"duration_ms":28205,"temperature":1.0,"reasoning_tokens":4113,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T09:47:15.417854+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the ratio form and the state-conditioned form for the same POVM element on a state, e.g., ρ = |+⟩⟨+| and Π = |0⟩⟨0|; if the two numbers differ, the average-to-QFI property and speed limits demonstrated numerically belong to the state-conditioned object rather than the ratio-defined CQFI. Re-running the driven-qubit simulation with the ratio definition and checking whether ⟨f⟩ still equals F_Q would settle whether the reported trajectory-level results survive under the paper's primary definition.","supporting_citations":[],"review_version":1}