{"id":"f3b266ae-15b0-4c75-bd3e-ae23f0c6b1d2","arxiv_id":"2509.04618","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"ITQDE can stably extract coarse spectral data with polynomially many shots and overlaps, provided the energy gaps of interest lie within a bandwidth set by the filter width tau and the shot budget K.","lead":"A new analysis of Imaginary-Time Quantum Dynamical Emulation (ITQDE) shows that coarse-grained spectral information, such as low-lying gaps and partition functions, can be estimated with polynomial resources, while exact ground-state resolution remains exponentially hard. It provides a quadrature formulation, stability criteria, and a smoothing step that create a practical 'free snack' regime for near-term quantum hardware.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sec. IV's quadrature overlap-count scaling is internally inconsistent: Eq. (19) simplifies to m̄ constant in s while the preceding balancing gives m̄ ∝ s, and the claimed m̄ ∼ O(n) for exponential τ does not follow.","rationale":"The reader's weakest-assumption analysis identifies exactly the same soft spot: the quadrature overlap-count scaling in Sec. IV. My reading of Eqs. (18)–(19) and the surrounding text confirms the internal inconsistency. The balancing argument forces m̄ ∝ s, while Eq. (19) algebraically reduces to m̄ independent of s; the subsequent inference that τ exponential in n implies m̄ ∼ O(n) is also invalid, since linear-in-τ is exponential in n when τ is exponential. This matters because the paper's 'free snack' narrative relies on the claim that quadrature makes the number of overlaps, and hence circuit count, polynomial even while τ grows exponentially for full spectral resolution. However, the central polynomial-resource claim for coarse, smoothed spectral information may survive in the regime where τeff is chosen polynomially (e.g. for gapped systems with constant gap), which is why the appropriate verdict remains conditional rather than rejection. The concern is real but localized to Sec. IV; Secs. V–VI's stability and sampling analysis, including the exponential shot-cost relation Eq. (37), appear internally consistent once the overlap-count issue is set aside.","tokens_in":21098,"tokens_out":8104,"duration_ms":79638,"concrete_test":"Independently re-derive the truncation requirement: take Eq. (18) with m = κs, set the target truncation error to a fixed ε, and solve for m̄ as a function of s. Then, for the 2D-Fermi-Hubbard model used in Fig. 2, compute the minimal m̄ needed to reach a fixed integrated error ϵ = 10^{-6} over the spectrum for s = τ‖H‖^2 spanning two orders of magnitude (e.g. by varying hopping J and rescaling τ). If the required m̄ grows approximately linearly in s, Eq. (19)'s constant-in-s form is falsified, and the text's m̄ ∼ O(n) claim for exponential τ has no support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The resource-reduction claim rests on the quadrature truncation scaling in Sec. IV. Eq. (18) sets (s/m)^m ∼ e^{-d m̄}, and choosing m = κs gives e^{-γ s} ∼ e^{-d m̄}, so balancing exponents yields m̄/s ∼ γ/d, i.e. m̄ ∝ s. However, Eq. (19) states m̄ ∝ s log(ε^{1/s}); since log(ε^{1/s}) = (ln ε)/s, the right-hand side is actually constant in s. The two statements cannot both hold: either m̄ must grow linearly with s (and hence exponentially with n when τ is exponential), or m̄ is fixed by the required precision and the linear-in-s balancing argument is wrong. The text then asserts that 'the effective number of overlaps required to approximate the staircase grows only linearly with τ, meaning m̄ ∼ O(n)'; this is also a non sequitur, because linear growth in τ with τ ∼ e^n is exponential in n. The numerical collapse in Fig. 2, plotted against m̄/s, does not discriminate between m̄ ∝ s and a precision-limited constant m̄, so it does not resolve the contradiction. Because the claimed surprising parsimony of ITQDE and the 'polynomial circuits for fixed precision' statement at the end of Sec. V depend on this step, the internal inconsistency is load-bearing for the free-snack resource claim over wide spectral windows.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyses Imaginary-Time Quantum Dynamical Emulation (ITQDE), a recently proposed method for estimating spectral densities, partition functions, and energy gaps from overlaps of oppositely propagated unitary evolutions. The author recapitulates the ITQDE correspondence, argues that exact spectral resolution must inherit an exponential cost, and then develops a Gauss-Hermite quadrature approximation with truncation to reduce the number of required overlaps. Secs. V and VI provide stability criteria and sampling-error bounds, and introduce a classical smoothing step that effectively rescales the filter width. The central claim is a 'free snack': coarse-grained spectral information can be obtained with resources polynomial in system size, while the QMA-hardness of exact ground-state energy estimation is preserved through an exponential shot-budget scaling involving the ratio of largest to smallest gaps in a spectral window.","tokens_in":21475,"tokens_out":3697,"duration_ms":38775,"significance":"If the scaling arguments were correct, the paper would provide a useful resource analysis of a NISQ-friendly spectral estimator and identify a concrete intermediate regime between trivial simulation and intractable exact resolution. The paper has genuine strengths: Fig. 1 validates the ITQDE staircase against exact eigenenergies on small Fermi-Hubbard models; the stability analysis in Sec. V and the sampling analysis in Sec. VI and App. B are transparent and supported by numerical experiments exhibiting the predicted K^{-1/2} law; and the distinction between deterministic truncation error and statistical sampling error is well drawn. However, the quadrature-overlap scaling in Sec. IV, which is load-bearing for the resource-reduction claim, is internally inconsistent; until that is corrected, the central 'polynomial overlaps' and 'free snack' statements are not established.","major_comments":[{"comment":"The derivation of the truncation scaling is internally inconsistent. Matching (s/m)^m to e^{-d m_bar} and setting m = kappa s gives e^{-gamma s} ~ e^{-d m_bar}, hence m_bar ~ (gamma/d) s, i.e. linear growth in s. Equation (19), however, states m_bar proportional to s log(epsilon^{1/s}); since log(epsilon^{1/s}) = (ln epsilon)/s, the right-hand side is constant in s. These two statements cannot both hold. Because s = tau ||H||^2 and full spectral resolution requires tau ~ 1/Delta_min^2, which is exponential in n for typical dense spectra, the distinction between constant and linear scaling is exponentially significant. The numerical collapse in Fig. 2, plotted against m_bar/s, does not discriminate between these two possibilities.","section":"Sec. IV, Eqs. (18)-(19)"},{"comment":"The claim that 'the effective number of overlaps required to approximate the staircase grows only linearly with tau, meaning m_bar ~ O(n)' is a non sequitur. Even if m_bar grew linearly with tau, tau must itself scale exponentially with n when the smallest gaps shrink exponentially with n, so linear-in-tau growth implies exponential-in-n growth. The statement appears to conflate linear growth in tau with linear growth in n. This affects the 'polynomial circuits for fixed precision' statement in Sec. V and the resource-reduction part of the free-snack claim, so it must be corrected explicitly.","section":"Sec. IV, final paragraph"},{"comment":"The shot-budget bound K >= exp(c (Delta_max/Delta_min)^2) is a plausible and clearly stated mechanism for recovering exponential hardness, and the smoothing construction tau_eff = tau/(1 + tau (delta_lambda)^2) is well motivated. However, the resource count for the smoothed estimator is incomplete: after smoothing to tau_eff ~ alpha/Delta_j^2, the number of retained overlaps m_bar must be re-evaluated with the corrected scaling from Sec. IV. If m_bar is constant in s for fixed precision, the smoothed regime is indeed polynomial; if m_bar is linear in s, the overlap count must also be stated for the smoothed parameters. The paper should present the total resource count (overlaps and shots) after smoothing, not only the shot budget.","section":"Sec. VI, Eq. (37) and smoothing discussion"}],"minor_comments":[{"comment":"There is a typo 'TO achieve' at the end of the introductory section; it should read 'To achieve'.","section":"Sec. I, Introduction"},{"comment":"The caption text 'min tau = 0.5 max tau = 6' is formatted as plain text; using a legend or a more standard notation such as 'tau_min = 0.5, tau_max = 6' would improve readability.","section":"Fig. 1 caption"},{"comment":"Near the end of the appendix the text reads 'their relative errors. in In particular'; the stray 'in' should be removed.","section":"App. B, sampling error"},{"comment":"Ref. [39] repeats the same textbook as Ref. [25] (Nielsen and Chuang); the duplicate should be consolidated.","section":"References"},{"comment":"The use of tau both as the filter width and as the quadrature node time tau_k = sqrt(tau) x_k is a slight abuse that is flagged in the text, but the distinction would be clearer if the node times were denoted by a different symbol, such as t_k.","section":"Sec. IV, notation"}],"recommendation":"major_revision","confidential_remarks":"The paper leans very heavily on the author's own prior work (Ref. [1]) for the core ITQDE correspondence; although Fig. 1 provides some validation on small models, the manuscript should state explicitly which parts are re-derived versus assumed. The internal inconsistency in Sec. IV is the main obstacle; if the corrected scaling turns out to make the overlap count exponential for full resolution, the smoothed-regime resource analysis in Sec. VI would still be the natural route to salvage the central claim, provided it is written out completely. I would encourage the editor to send the revised version back to a referee who checks the corrected scaling carefully."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the ITQDE quadrature paper. Short version: the stability and sampling analysis is genuinely good, but the headline resource claim over wide spectral windows rests on a Sec. IV scaling argument that is internally inconsistent, and the m̄∼O(n) sentence does not follow from the preceding equations. I would still send it to a referee, because the core ideas are worth engaging, but Sec. IV needs to be corrected before the polynomial-resource claims are taken at face value.\n\nWhat is actually new and right: the quadrature formulation of ITQDE, the stability threshold Δ_ε and tail-ratio argument, the local gap-bandwidth sampling formula K ≳ exp(c κ_loc^2), and the τ→τ_eff smoothing map. These are not in Ref. [1], and the numerical checks on small Fermi-Hubbard models are consistent with the stability and sampling predictions. The paper does not claim to break QMA-hardness; it tries to locate the obstruction in the shot budget. That is a useful and honest framing.\n\nThe soft spot is Sec. IV. Eq. (18) balances exponents to get m̄/s ∼ γ/d, i.e. m̄ ∝ s. Eq. (19) then gives m̄ ∝ s log(ε^{1/s}), which is algebraically constant in s. Both cannot hold. On top of that, the text says the overlap count grows \"only linearly with τ, meaning m̄ ∼ O(n)\". Even if linear-in-τ were correct, τ ∼ exp(n) when you chase exponentially small gaps, so linear in τ is exponential in n, not O(n). The Fig. 2 collapse against m̄/s is consistent with m̄ ∝ s and does not discriminate between the two versions. This is not a cosmetic typo: the fixed-precision polynomial-circuit statement in Sec. V and the wide-window version of the free snack depend on it. The coarse-grained regime with τ_eff matched to a fixed gap may survive a corrected scaling, but as written the argument is broken.\n\nA smaller issue: the polynomial-shot claim in Sec. VI uses the trace-averaged variant with stochastic trace estimation, and the cost of preparing the random states and evolving out to times √τ x_k is not fully counted. If those times are exponential, circuit depth is exponential no matter how few overlaps you keep.\n\nThe reliance on the author's own Ref. [1] is a bit heavy, but Fig. 1 provides a direct numerical check, so I would not count that against the paper.\n\nWho it is for: people working on NISQ spectral estimation and imaginary-time methods. It deserves a serious referee: the stability analysis and the κ_loc formula are citable, and the flaw is fixable. Recommendation: engage, but keep the free-snack language tied to the corrected math.","headline":"Useful ITQDE analysis with a broken Sec. IV scaling argument; worth refereeing once the m̄ inconsistency is fixed.","tokens_in":21958,"tokens_out":6217,"would_cite":true,"duration_ms":59915,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Ac"],"model":"deepseek-v4-flash","headline":"A quadrature-enhanced imaginary-time method extracts coarse spectral information from a Hamiltonian at polynomial sampling cost, while exact ground-state estimation remains exponentially hard.","keywords":["ITQDE","imaginary-time evolution","spectral staircase","Gauss-Hermite quadrature","QMA-completeness","ground-state energy estimation","quantum sampling complexity","coarse-grained spectral information"],"falsifier":"Directly measure the number of quadrature terms $\\bar{m}$ needed to hold the integrated staircase error $\\epsilon$ fixed as $s = \\tau\\|H\\|^2$ is increased across many orders of magnitude, for example by rescaling a fixed Hamiltonian's hopping amplitude as in Fig. 2. If $\\bar{m}$ must grow linearly with $s$ to maintain constant $\\epsilon$, the claimed $O(n)$ overlap count is wrong; if $\\bar{m}$ stays bounded or grows logarithmically, the polynomial-resource claim survives. A second check: with $K$ fixed and smoothing width $\\delta\\lambda$ varied, verify the predicted precision $\\delta_j \\approx \\Delta_j/\\sqrt{\\ln K}$ and the largest-stable-gap formula $\\Delta_\\varepsilon \\approx \\sqrt{\\ln K/(2\\tau_{\\mathrm{eff}})}$ against exact diagonalization on small Fermi-Hubbard lattices.","tokens_in":20869,"feed_emoji":"🍿","tokens_out":8096,"duration_ms":67255,"temperature":0.7,"pith_summary":"This paper tries to map the boundary between cheap and expensive spectral estimation for a recently proposed quantum method. It argues that Imaginary-Time Quantum Dynamical Emulation (ITQDE), which reconstructs a Hamiltonian's spectrum from overlaps of forward and backward unitary evolutions, can be implemented with a Gauss-Hermite quadrature approximation so that only a modest set of weighted overlap measurements is needed. The paper's central claim is that exact eigenvalue resolution remains exponentially hard, but coarse-grained spectral information is available at polynomial cost after a classical smoothing step. That would matter because it puts an intermediate regime, including spectral densities, low-lying gaps, and partition functions, within reach of near-term hardware, without contradicting the QMA-hardness of ground-state energy estimation, the quantum analogue of NP-hardness.","feed_headline":"Quantum 'free snack': coarse spectra at polynomial cost","feed_subtitle":"ITQDE with quadrature smoothing resolves gaps relative to the local energy scale while keeping shot counts polynomial.","key_machinery":"The central object is the spectral staircase $H_\\tau(\\lambda) = \\langle N_\\tau^{(\\lambda)}\\rangle/\\langle Z_\\tau^{(\\lambda)}\\rangle$, a Gaussian-filtered weighted mean of eigenvalues that steps through plateaux at the eigenenergies as the probe $\\lambda$ is scanned. The mechanism is the ITQDE correspondence: the non-unitary filter $e^{-\\tau(H-\\lambda)^2}$ is replaced by a sum of unitary overlaps $\\langle\\psi_{-\\tau_k}|O|\\psi_{\\tau_k}\\rangle$, and a Gauss-Hermite quadrature, replacing the Hubbard-Stratonovich integral by weighted nodes $x_k$ with $\\tau_k=\\sqrt{\\tau}x_k$, makes the approximation exponentially accurate in the number of retained terms $\\bar{m}$. Truncation stability is controlled by the tail ratio $r(\\lambda)=Z_\\varepsilon/Z_\\tau(\\lambda)$, which sets a largest resolvable gap; sampling stability comes from estimating numerator and denominator from the same batch of random states so their covariance cancels the $1/D$ orthogonality catastrophe; and the smoothing identity $\\tau_{\\mathrm{eff}} = \\tau/(1+\\tau\\delta\\lambda^2)$ converts resolution into a tunable bias-variance trade-off.","core_discovery":"On its own terms, the paper's discovery is an explicit resource accounting for ITQDE that identifies the shot budget, not circuit depth or number of overlaps, as the locus of exponential cost. If a spectral window contains both a smallest gap $\\Delta_{\\min}$ and a largest gap $\\Delta_{\\max}$, resolving both simultaneously requires $K \\gtrsim \\exp(c(\\Delta_{\\max}/\\Delta_{\\min})^2)$ shots with $c=O(1)$; this is the mechanism by which QMA-completeness is preserved. If one instead convolves the staircase with a classical window of width $\\delta\\lambda$, the filter width governing the estimate changes from $\\tau$ to $\\tau_{\\mathrm{eff}} = \\tau/(1+\\tau\\delta\\lambda^2)$, and the precision of each estimated level becomes $\\delta_j \\approx \\Delta_j/\\sqrt{\\ln K}$, proportional to the local gap. With $K$ polynomial in system size, the method therefore delivers efficiently smoothed spectral information, the 'free snack', while leaving exact ground-state estimation exponentially expensive.","pith_inferences":["Taken literally, the derivation of Eq. (19) is internally inconsistent: $\\bar{m} \\propto s\\log(\\varepsilon^{1/s})$ reduces algebraically to $\\bar{m} \\propto \\ln\\varepsilon$, independent of $s$, while the surrounding text claims linear-in-$s$ growth. The $\\bar{m}\\sim O(n)$ overlap scaling is therefore not established by the algebra shown, and it should be checked numerically for large $s$ before th","The local bandwidth ratio $\\kappa_{\\mathrm{loc}} = \\Delta_{\\max}/\\Delta_{\\min}$ may be a more natural complexity parameter for spectral estimation than $n$ itself; the exponential $K \\gtrsim \\exp(c\\kappa_{\\mathrm{loc}}^2)$ echoes the scale-coexistence mechanism behind spectral-gap undecidability, suggesting a continuum between easy and undecidable spectral problems.","The shared-batch ratio trick behind the variance analysis is a general principle: any estimator formed as a ratio of two positively correlated trace estimates can cancel an exponential concentration of the individual traces, which may apply to other quantum algorithms that compute normalized quantities.","A testable extension would replace the Gaussian filter with other positive kernels, such as Lorentzians or polynomial filters; if the same overlap scaling and $\\tau_{\\mathrm{eff}}$ smoothing hold, the 'free snack' regime is a property of coarse spectral filtering generally, not of Gaussians specifically."],"forward_implications":["Near-term devices can use ITQDE with quadrature to estimate spectral densities, partition functions, and low-lying gaps at polynomial sampling cost, provided the target precision is expressed relative to the local gap rather than as absolute $1/\\mathrm{poly}(n)$ accuracy.","Over-resolved spectral windows are self-diagnosing: the oscillatory truncation error predicted by the tail-ratio analysis appears precisely where $\\tau$ is too large for the gap, so a user can detect and coarsen those windows.","For gapped systems whose gap stays finite as $n$ grows, the achievable precision of each energy level decouples from system size, making the method particularly useful for phase identification and thermodynamic estimates.","The same quadrature and smoothing analysis transfers to other non-unitary dynamics, including non-Hermitian Hamiltonian evolution and Lindbladian dissipation, because the underlying emulation correspondence is not specific to Hermitian filters.","If the overlap-count scaling holds, the number of distinct circuits needed is only polynomial for fixed resolution, so the method is compatible with shallow NISQ-era circuits and modest classical post-processing."],"supporting_citations":[{"why":"Establishes the ITQDE correspondence that maps the non-unitary filter onto unitary overlap measurements, the object this paper analyzes.","marker":"[1]"},{"why":"Provides the Gauss-Hermite quadrature nodes, weights, and asymptotics used to derive the truncation and scaling laws.","marker":"[64]"},{"why":"Supplies the unitary 2-design and stochastic trace formalism underlying the shot-budget variance analysis.","marker":"[65]"},{"why":"States the QMA-completeness framework for the local Hamiltonian problem, the hardness barrier the paper must preserve.","marker":"[24]"},{"why":"Gives the formal complexity classification of the local Hamiltonian problem used for the exponential-gap scaling argument.","marker":"[10]"},{"why":"Describes the orthogonality catastrophe whose 1/D concentration the shared-batch ratio estimator cancels.","marker":"[66]"},{"why":"Supports the use of shared-batch covariance to reduce the variance of ratio estimates.","marker":"[67]"},{"why":"Connects scale coexistence in spectral gaps to undecidability, the analogy motivating the local-bandwidth scaling expression.","marker":"[68]"}],"fun_headline_variants":["Quantum free snack: cheap spectral estimates","Coarse spectra without the exponential price","Polynomial-cost spectral info via ITQDE","Free snack in quantum complexity: smoothed spectra","ITQDE yields spectral snacks at polynomial cost"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that Gauss-Hermite truncation leaves only $\\bar{m}\\sim O(n)$ retained overlap terms even when the filter width $\\tau$ must grow exponentially with $n$; the paper's derivation of this scaling appears internally inconsistent, since its Eq. (19) reduces to a constant rather than a linear-in-$s$ law, so if the true scaling is linear in $s$, the polynomial-overlap resource claim loses its foundation.","fun_headline_variants_meta":{"raw":{"variants":["Quantum free snack: cheap spectral estimates","Coarse spectra without the exponential price","Polynomial-cost spectral info via ITQDE","Free snack in quantum complexity: smoothed spectra","ITQDE yields spectral snacks at polynomial cost"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00017,"raw_usage":{"total_tokens":1247,"prompt_tokens":904,"completion_tokens":343,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":520,"completion_tokens_details":{"reasoning_tokens":277}},"tokens_in":520,"tokens_out":343,"duration_ms":2893,"temperature":1.0,"reasoning_tokens":277,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:30:26.303688+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly measure the number of quadrature terms $\\bar{m}$ needed to hold the integrated staircase error $\\epsilon$ fixed as $s = \\tau\\|H\\|^2$ is increased across many orders of magnitude, for example by rescaling a fixed Hamiltonian's hopping amplitude as in Fig. 2. If $\\bar{m}$ must grow linearly with $s$ to maintain constant $\\epsilon$, the claimed $O(n)$ overlap count is wrong; if $\\bar{m}$ stays bounded or grows logarithmically, the polynomial-resource claim survives. A second check: with $K$ fixed and smoothing width $\\delta\\lambda$ varied, verify the predicted precision $\\delta_j \\approx \\Delta_j/\\sqrt{\\ln K}$ and the largest-stable-gap formula $\\Delta_\\varepsilon \\approx \\sqrt{\\ln K/(2\\tau_{\\mathrm{eff}})}$ against exact diagonalization on small Fermi-Hubbard lattices.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Gauss-Hermite quadrature nodes, weights, and asymptotics used to derive the truncation and scaling laws."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the QMA-completeness framework for the local Hamiltonian problem, the hardness barrier the paper must preserve."},{"cited_title":"Kempe, A","cited_arxiv_id":null,"evidence_quote":"Gives the formal complexity classification of the local Hamiltonian problem used for the exponential-gap scaling argument."}],"review_version":2}