{"id":"64bb0a41-63ca-4d87-b6b4-901347c868ca","arxiv_id":"2501.02929","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Adaptive circuits with unary encoding prepare sparse states, Slater determinant sums, and Bethe wavefunctions in logarithmic or constant depth, trading circuit depth for extra width.","lead":"This paper designs quantum circuits that prepare specific states for quantum simulation with less depth by using mid-circuit measurements and classical feedforward, at the cost of more qubits. The methods target sparse quantum states, sums of Slater determinants, and Bethe wavefunctions, with claimed constant or logarithmic depth.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 7's proof gives O(M^3 log M) width for its first summand, which for M≈L/2 is Θ(L^3 log L) and contradicts the claimed O~(L^2 log L) width of the Bethe preparation in Theorem 4.","rationale":"I read the paper as a constructive claim: measurement/feedforward plus unary encoding gives constant-depth or log-depth preparation of sparse states, SOS, and Bethe states with stated widths. The central novelty is the depth-width tradeoff, so the width figures are load-bearing, not cosmetic. I checked the internal arithmetic of the Bethe construction. The weakest point is not a vague suspicion about uncomputation; it is the explicit width accounting in Appendix D. Lemma 5's width scales as O(ηζ logζ); with η=M and ζ=M^2 for the first summand of Eq. (16), this is O(M^3 log M), and Appendix D says so. Lemma 7 and Theorem 4 nevertheless claim O~(L^2 log L) under M≤L/2. Since M can be L/2, O(M^3 log M) is Θ(L^3 log L), which violates the claimed width. This is a falsifiable, load-bearing gap in the strongest claim. I do not recommend rejection: the depth result may survive with a corrected width O(M^3 log M + L^2 log L), and the Uncompress/Compress lemmas appear coherent. The reader's concern about uncomputation/postselection is related but less sharp; my independent check is narrower and settles against the current statement. I therefore keep the CONDITIONAL verdict, recommending the authors fix the width bound or restrict the allowed M regime.","tokens_in":31882,"tokens_out":15630,"duration_ms":230625,"concrete_test":"Analytically check Appendix D by substituting η=M and ζ=M^2 into Lemma 5's width bound and comparing with Theorem 4's width at M=L/2. If the first summand costs Θ(L^3 log L), the claimed O~(L^2 log L) is false. Then determine whether the intended restriction is M=O(L^{2/3}) or whether an alternative construction avoids Lemma 5; either Theorem 4's condition must be narrowed or its width claim revised to include O(M^3 log M).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim is the constant-depth Bethe preparation of Theorem 4 with width O~(L^2 log L). The proof is internally inconsistent. Lemma 5 has width O~(ηζ logζ) for state (9). Lemma 7 wants state (16), whose first summand uses η=M and ζ=M^2, so Lemma 5 alone costs O(M^3 log M). Appendix D explicitly states: 'the first part of the desired quantum state can be prepared using a constant-depth quantum circuit of width O(M^3 log M)'. For M≤L/2, the worst case M≈L/2 gives M^3 log M = Θ(L^3 log L), which asymptotically dominates the claimed O~(L^2 log L). The later Cleaning step, also quoted as O~(L^2 log L), cannot erase this first-summand cost. Thus the width bound in the central Bethe theorem is not derived; the construction as written supports only O(M^3 log M + L^2 log L). The reader's uncomputation/postselection concern is real, but this width accounting is a concrete algebraic inconsistency that does not depend on interpreting a cited probability.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes constant-depth quantum circuits with mid-circuit measurements and feedforward for preparing several quantum states relevant to quantum simulation: sparse quantum states (Theorem 1), single (anti)symmetric states (Theorem 2), sums of Slater determinants in first quantization (Theorem 3), and Bethe wavefunctions (Theorem 4). The main technical tool is a constant-depth transformation between binary and unary encodings (Uncompress and Compress, Lemmas 2 and 3), combined with known constant-depth primitives such as the quantum fan-out gate, Equal, GreaterThan, and Dicke-state preparation. For Bethe wavefunctions, the authors combine a prepared Dicke state with permutation registers and controlled pairwise commuting phase gates to encode the phases A_σ(θ) and exp(i k·x), claiming constant depth, width O~(L^2 log L), and success probability O(1/M!).","tokens_in":32119,"tokens_out":6970,"duration_ms":62985,"significance":"If the constructions are correct, the paper would demonstrate a significant depth reduction for state preparation tasks that previously required depth O(log d) or O(L), at the cost of increased width. The unary-encoding transformation between integer sets is clearly presented and makes use of established constant-depth feedforward primitives in a coherent way, which is a genuine strength. The claimed constant-depth Bethe wavefunction preparation, together with explicit resource counts, is a falsifiable and potentially useful contribution to early fault-tolerant quantum simulation. However, the significance is conditional on resolving several load-bearing proof gaps, most notably a width accounting inconsistency in the Bethe theorem that currently invalidates the claimed O~(L^2 log L) bound in the worst case.","major_comments":[{"comment":"The width bound in Theorem 4 is not derived from the provided construction. In Appendix D, the proof of Lemma 7 states that 'the first part of the desired quantum state can be prepared using a constant-depth quantum circuit of width O(M^3 log M)', which follows from Lemma 5 with η = M and ζ = M^2. For the allowed range M ≤ L/2, taking M ≈ L/2 gives Θ(L^3 log L), which asymptotically dominates the claimed O~(L^2 log L) width of Theorem 4. The later Cleaning step, also quoted as O~(L^2 log L), cannot reduce the cost of the first summand. The theorem's stated width is therefore not supported by the proof; as written, the construction supports only O(M^3 log M + L^2 log L).","section":"Appendix D / Lemma 7 / Theorem 4"},{"comment":"Both SOS constructions end with an asserted uncomputation: the first version concludes 'Finally, applying the inverse operations of Compress and Uncompress ensures that only the information of r0 remains', and the second version ends with 'we obtain the desired state after uncomputing all redundant states'. The uncomputation of the coefficient register and of the intermediate index registers is not shown. If these inverse operations cannot be implemented in constant depth (or within the stated width), the O(d) and O(log d) depth claims of Theorem 3 fail. The proof should either provide an explicit circuit for the final uncomputation or invoke a lemma that guarantees the required disentangling within the claimed resources.","section":"Appendix C (both versions of Theorem 3)"},{"comment":"The success probability in Theorem 4 is stated as O(1/M!), but the proof does not derive this for the authors' own circuit. Step 3 asserts that after applying the inverse phase-encoding gates, the probability of measuring |00...0⟩ is 'known to be O(1/M!) [40]'. Reference [40] derives this for a different Bethe-state preparation scheme, not for the circuit constructed here, which includes additional registers and phase gates (Lemmas 8–10 and the Cleaning step). Since the success probability is a central part of the theorem's claim, the authors should either compute the postselection amplitude for their specific circuit or prove that the cited bound carries over to their construction.","section":"Appendix H / Theorem 4"},{"comment":"The proof of Lemma 8 contains a circular reference: it states that the phase-attaching gates can be parallelized 'Using Theorem 4', but Lemma 8 is itself used in the proof of Theorem 4. This is not a valid proof step. The intended meaning is presumably that the parallelization follows from the quantum fan-out technique of Refs. [17,18], but the proof of Lemma 8 must be self-contained or explicitly reference the appropriate prior lemma rather than the theorem being proved.","section":"Appendix E / Lemma 8"},{"comment":"The first sentence of Theorem 4 claims that the Bethe wavefunction for M ≤ L/2 can be prepared in constant depth with width O~(L^2 log L). The subsequent sentence, however, splits the result into two cases: exact preparation restricted to M = O(√L), or approximate preparation with infidelity O(1/2^{LL}) and no restriction on M. The theorem statement should reflect this dichotomy explicitly, since as written it suggests exact constant-depth preparation for all M ≤ L/2, which is not established.","section":"Theorem 4 statement"}],"minor_comments":[{"comment":"The resource count transitions from O(d log^2 n log log^2 n) to O(d n log n); the equality is not justified, and the two expressions are not asymptotically equal (the first is smaller for large n). The bound is true in the sense of being an upper bound, but the presentation is misleading.","section":"Theorem 1 proof"},{"comment":"In the definition of the Bethe wavefunction, the sum over σ is written as S_η, but the number of particles is denoted M; it should be S_M. The same inconsistency appears in the surrounding discussion.","section":"Eq. (13)"},{"comment":"Equation (4) omits the summation on the right-hand side; it should read Σ_i α_i |0⟩_{log N} |e_i⟩_η.","section":"Lemma 3"},{"comment":"The 'Cleaning operation' from Ref. [17] is invoked without definition; its action and width scaling should be specified so that the reader can verify the claimed O~(L^2 log L) contribution.","section":"Appendix D"},{"comment":"There are several typographical issues, including 'Greatherthan' in Table II and Appendix A, 'Preparaing' in the heading of Section IV.B, 'spririt' in Appendix H, and inconsistent capitalization of 'Uncompress'/'uncompress'.","section":"Throughout"},{"comment":"The labels in Figure 1b are partially garbled (e.g., '𝑟𝑖log𝑁𝑟 0𝜂 0log𝑁𝑝'), making the figure difficult to interpret.","section":"Figure 1"}],"recommendation":"major_revision","confidential_remarks":"The paper presents an appealing depth/width trade-off for state preparation, but the main theorem's width claim is undermined by an explicit contradiction in Appendix D. This is not a mere presentation issue: the construction as written gives O(M^3 log M + L^2 log L) width, which for M~L/2 is Θ(L^3 log L), contradicting the abstract and Theorem 4. The other gaps (uncomputation in Theorem 3, postselection probability in Theorem 4) also need to be addressed before the results can be considered reliable. I recommend the editor ask for a thorough revision with corrected resource bounds and complete proofs, and a resulting re-evaluation of the claimed contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper has a genuinely useful core—the unary-encoding bridge plus constant-depth fan-out tools gives new depth-width tradeoffs for sparse states and Slater determinant sums. The sparse-state result (Thm 1) looks solid, and the two SOS circuits in Thm 3 are plausible and worth taking seriously. But the Bethe wavefunction claim (Thm 4) has a width-accounting problem that the authors' own appendix exposes.\n\nWhat's new: the parallelized constant-depth unary transformation (Lemmas 2,3) is a real contribution, even though the bridge itself was in Ref [47]. Combining it with the controlled-pairwise-commuting-gate trick from Ref [18] to get O(log d) sparse prep and O(log d) or O(d) SOS prep is a decent step forward. The paper is clearly written and properly credits the primitives it imports. No fitted parameters, no hand-waving about resources of external tools—the constructions are explicit enough to check.\n\nSoft spots: the Bethe width. Lemma 7's proof (Appendix D) states the first summand costs O(M^3 log M). For M=O(√L) that's fine, but the theorem claims O~(L^2 log L) without the constraint (using the approximate Dicke route of Ref [13]). For M≈L/2, M^3 log M = Θ(L^3 log L), which dominates the stated width. So the claimed width bound in Theorem 4 is not derived; the construction as written gives O(M^3 log M + L^2 log L). That needs a fix, either by restricting M or by changing the bound. Second, the uncomputation steps in Appendix C and the postselection probability in Appendix H are asserted, not proven. Those are load-bearing for the depth claims. Minor typos and notation slips don't affect the math, but the paper would benefit from a careful pass.\n\nThe reader's CONDITIONAL verdict is on the mark. The core ideas are sound and new enough to deserve referee time, but the Bethe width inconsistency must be addressed before this can be accepted.\n\nRecommendation: send it to peer review, but with a clear request for the authors to supply a corrected width analysis for Theorem 4 and explicit circuit decompositions for the uncomputation steps. My guess is the sparse and SOS results survive; the Bethe result may need to be scaled back.","headline":"Clever depth-width tradeoffs for state preparation, but the headline Bethe width claim doesn't survive a close check of its own appendix.","tokens_in":32640,"tokens_out":3775,"would_cite":true,"duration_ms":33446,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68"],"pacs":["03.67.Lx"],"model":"deepseek-v4-flash","headline":"The paper claims that adaptive circuits with unary encoding prepare sparse states, sums of Slater determinants, and Bethe wavefunctions in logarithmic or constant depth, at the price of more qubits.","keywords":["quantum state preparation","circuit depth reduction","measurements and feedforward","unary encoding","Bethe wavefunction","Slater determinants","sparse quantum state","quantum simulation"],"falsifier":"Compile the Compress subroutine for small sizes, e.g. $\\eta = 2$ and $\\log N = 2$, apply it to a superposition with known amplitudes, and check whether the binary register returns exactly to $|00\\ldots0\\rangle$ with the original amplitudes intact; any residual phase or amplitude error falsifies Lemma 3. For the Bethe claim, run Theorem 4's circuit for $L=4, M=2$ with the exact Dicke option, measure the frequency with which the postselection register reads $|00\\ldots0\\rangle$, and compare the output state's fidelity to $|\\psi(\\theta,\\vec{k})\\rangle$ against $1 - \\mathcal{O}(1/2^L)$; a success probability far from $1/M!$ or a fidelity below that bound would falsify Theorem 4.","tokens_in":31697,"feed_emoji":"⚛️","tokens_out":7940,"duration_ms":69892,"temperature":0.7,"pith_summary":"The paper argues that mid-circuit measurement and feedforward, which allow constant-depth implementations of quantum fan-out and related logic gates, can be converted into a depth-width trade-off for preparing quantum states that commonly seed quantum simulations. By using unary encoding as a reversible bridge between two superpositions over ordered integer tuples, it constructs an $\\mathcal{O}(\\log d)$-depth, $\\mathcal{O}(d n \\log n)$-width circuit for arbitrary $d$-sparse $n$-qubit states, and extends the same machinery to sums of Slater determinants in first quantization at depth $\\mathcal{O}(d)$ or $\\mathcal{O}(\\log d)$. Its central result is Theorem 4: an $L$-site $M$-particle Bethe wavefunction with $M \\le L/2$ can be prepared with a constant-depth adaptive circuit of width $\\widetilde{\\mathcal{O}}(L^2 \\log L)$ and success probability $\\mathcal{O}(1/M!)$. If correct, initial-state preparation for these state families would no longer be the depth bottleneck in quantum simulation, at the cost of substantially wider circuits and probabilistic success.","feed_headline":"Measurement feedback makes Bethe state preparation constant-depth","feed_subtitle":"Unary encoding as a bridge cuts depth for sparse states, Slater sums, and Bethe states—at the price of extra qubits.","key_machinery":"The load-bearing objects are (i) the Clifford-ladder circuit, which realizes the quantum fan-out gate in constant depth through Bell pairs, mid-circuit measurements, and feedforward; (ii) the Uncompress and Compress operations, which reversibly map a binary integer to its unary (one-hot) representation and back in constant depth, so that a superposition over integer tuples can be re-routed through an addressable unary register; and (iii) the parallelization of a product of controlled pairwise-commuting gates by simultaneously diagonalizing them and using a unary-encoded control register. These routines are what let the paper trade width for depth: each previously sequential step is replaced by parallel fan-out-mediated operations acting on many replicated registers. The width bounds in Theorems 1 through 4 are dominated by these replications.","core_discovery":"On the paper's own terms, the discovery is that two constant-depth primitives, the binary-unary-binary conversion and the parallelization of controlled pairwise-commuting phase gates, are enough to restructure the whole state-preparation pipeline for structured states. The first primitive reduces the problem of transforming a superposition over one ordered integer set into another to constant depth with width $\\widetilde{\\mathcal{O}}(\\eta \\log N)$, yielding the sparse-state and Slater-determinant results. The second primitive attaches the scattering phases $A_\\sigma \\exp(i \\sum_\\ell k_{\\sigma(\\ell)} x_\\ell)$ of the Bethe wavefunction to a Dicke-state backbone in one parallel layer, after the inverse permutation has been encoded in unary; combining these with a probabilistic postselection gives the constant-depth Bethe preparation. The authors state these results as Theorems 1, 3, and 4, and give width bounds for each.","pith_inferences":["The same binary-unary-binary bridge should apply to other index-symmetric state families in first quantization, such as permanents or fermionic antisymmetrized geminals, whenever the amplitudes factor over ordered tuples; the paper notes permanents as a byproduct but does not develop the consequences.","A direct computation of the Bethe postselection probability from the authors' own inverse-phase circuit, rather than a citation, would sharpen the runtime estimate; the paper's constant-depth claim is otherwise insensitive to the multiplicative success overhead.","The width bounds suggest a practical crossover: on devices where idle qubits are cheap and coherence is short, these adaptive circuits should outperform alternatives, while on qubit-starved hardware the $\\widetilde{\\mathcal{O}}(L^2 \\log L)$ width for Bethe states will dominate the resource budget.","Extending the phase-attachment lemma to complex scattering matrices would require encoding more than pairwise ordering information per permutation; the paper mentions this as a future direction but leaves it open."],"forward_implications":["Any $d$-sparse $n$-qubit state can be initialized in depth $\\mathcal{O}(\\log d)$ with $\\mathcal{O}(d n \\log n)$ qubits, matching the best previous depth bound at a polynomial width cost.","Sums of Slater determinants in first quantization can be prepared in depth $\\mathcal{O}(\\log d)$ with width $\\widetilde{\\mathcal{O}}(\\eta^2 d^2 \\log N)$, or in depth $\\mathcal{O}(d)$ with width $\\widetilde{\\mathcal{O}}(\\eta^2(\\eta \\log \\eta + \\log N) + \\log d)$.","Bethe wavefunctions with up to half-filling ($M \\le L/2$) can be prepared in constant depth with width $\\widetilde{\\mathcal{O}}(L^2 \\log L)$ and success probability $\\mathcal{O}(1/M!)$, either exactly under $M = \\mathcal{O}(\\sqrt{L})$ or with infidelity $\\mathcal{O}(1/2^L)$ in the general case.","Because integer-tuple superpositions can be rewired in constant depth, the unary bridge is a reusable subroutine that can be attached to any amplitude-loading circuit that produces a superposition over ordered indices."],"supporting_citations":[{"why":"Provides the constant-depth implementations of fan-out, Equal, Greater-than, and Dicke-state preparation that all later routines build on.","marker":"[17]"},{"why":"Supplies the O(log d)-depth, O(d)-width amplitude encoding for a d-dimensional register that Theorem 1 converts into sparse-state preparation.","marker":"[31]"},{"why":"Introduces the unary-encoding bridge for sparse states that this paper rewires into constant depth.","marker":"[47]"},{"why":"The constant-depth arbitrary state preparation whose width scaling motivates the depth-width trade-off comparison.","marker":"[48]"},{"why":"Gives the bitonic-sort (anti)symmetrization that Theorem 3's sums of Slater determinants must beat on depth.","marker":"[49]"},{"why":"Provides the approximate constant-depth Dicke-state preparation used for the unconstrained M <= L/2 Bethe case.","marker":"[13]"},{"why":"Is the cited source for the O(1/M!) postselection success probability of the Bethe preparation.","marker":"[40]"},{"why":"Defines the generalized Bethe wavefunction that Theorem 4 prepares.","marker":"[33]"},{"why":"Is the previous O(L)-depth Bethe eigenstate preparation that Theorem 4 improves on.","marker":"[39]"}],"fun_headline_variants":["Constant-depth Bethe states via measurement feedback","Feedforward gets Bethe states to constant depth","Unary trick enables constant-depth state prep","Measurement feedback shrinks state-prep depth","Depth reduction via feedback for state prep"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central bet is that every auxiliary register can be uncomputed again in constant depth without disturbing the target superposition, and that the Bethe postselection succeeds with the cited $\\mathcal{O}(1/M!)$ probability, so the final state is left clean at the claimed depth.","fun_headline_variants_meta":{"raw":{"variants":["Constant-depth Bethe states via measurement feedback","Feedforward gets Bethe states to constant depth","Unary trick enables constant-depth state prep","Measurement feedback shrinks state-prep depth","Depth reduction via feedback for state prep"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000825,"raw_usage":{"total_tokens":3596,"prompt_tokens":922,"completion_tokens":2674,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":2609}},"tokens_in":538,"tokens_out":2674,"duration_ms":19613,"temperature":1.0,"reasoning_tokens":2609,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:00:23.498506+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compile the Compress subroutine for small sizes, e.g. $\\eta = 2$ and $\\log N = 2$, apply it to a superposition with known amplitudes, and check whether the binary register returns exactly to $|00\\ldots0\\rangle$ with the original amplitudes intact; any residual phase or amplitude error falsifies Lemma 3. For the Bethe claim, run Theorem 4's circuit for $L=4, M=2$ with the exact Dicke option, measure the frequency with which the postselection register reads $|00\\ldots0\\rangle$, and compare the output state's fidelity to $|\\psi(\\theta,\\vec{k})\\rangle$ against $1 - \\mathcal{O}(1/2^L)$; a success probability far from $1/M!$ or a fidelity below that bound would falsify Theorem 4.","supporting_citations":[{"cited_title":"Fomichev, K","cited_arxiv_id":null,"evidence_quote":"Gives the bitonic-sort (anti)symmetrization that Theorem 3's sums of Slater determinants must beat on depth."}],"review_version":1}