{"id":"6f442ed8-7427-499f-8b57-acc75c4902be","arxiv_id":"2504.12893","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Single-layer unitary cluster Jastrow circuits can emulate quadratic-Ising IQP circuits, so sampling from some quantum chemistry ansaetze is worst-case classically hard unless the polynomial hierarchy collapses.","lead":"This paper proves that sampling from the output of certain quantum chemistry circuits is classically hard unless the polynomial hierarchy collapses. It works by showing these chemistry circuits can simulate IQP circuits, which are already known to be hard for classical computers.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The hardness transfer rests on the unstated premise that the Hadamard gadget in Ref. [31] uses only 1- and 2-local Z-diagonal gates; the paper asserts this in one paragraph but does not demonstrate it.","rationale":"The reader's weakest_assumption identifies exactly the same point: the restricted quadratic-form IQP family must inherit the post-IQP = post-BQP hardness of Ref. [31]. This is the most load-bearing premise for the paper's headline claim, because if it fails, the inclusion 1-UCJ ⊇ IQP does not yield worst-case classical hardness. I reviewed the rest of the proof in detail: the mapping from IQP to 1-UCJ' is explicit, the parity invariant for V gates is correct, the Rz and controlled-phase constructions act properly on the encoded subspace, the reference-state encoding is consistent, and the spin-down sector can be set to identity without affecting the hardness transfer. The only concern is the compressed treatment of the restricted-IQP hardness. It is a missing justification rather than an identified error: the standard Hadamard gadget for IQP is known to be implementable with T and CZ gates, which are exactly of the quadratic-Z form, so the concern is likely to resolve. For this reason, I do not recommend changing the reader's ACCEPT verdict, but I would encourage the authors to make the gadget check explicit in a revision.","tokens_in":15198,"tokens_out":46122,"duration_ms":486078,"concrete_test":"Reconstruct the Hadamard-gadget construction from Ref. [31]'s proof of post-IQP = post-BQP and enumerate every diagonal gate used inside the gadget, ignoring only the outer H layers. If every such gate is of the form exp(iθZ) or exp(iθ(I−Z)⊗(I−Z)) for real θ (with the usual T and CZ cases included), then the quadratic-D IQP family inherits post-IQP = post-BQP, and the hardness reduction to 1-UCJ is complete. If any gate is a 3-qubit or higher Z-diagonal gate, then Theorem III.1 would need additional argument or would fail.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central theorem Theorem III.1 (1-UCJ_JW contains IQP) is intended to inherit worst-case hardness from Bremner et al.'s IQP result via the restricted family of circuits with C_diag = e^{iD}, D = Σ w_αβ Z_α Z_β + Σ v_α Z_α (Eq. 1). The load-bearing step is Section II.B's claim that 'the complexity argument is not affected' because e^{iD} can implement e^{iθZ} and e^{iθ(I−Z)⊗(I−Z)} gates, which 'in turn can implement any diagonal gate in the universal gate set adopted in the proof.' This is only sufficient if the Hadamard-gadget argument used to prove post-IQP = post-BQP in Ref. [31] never needs a diagonal gate with a higher-order Z term, such as a CCZ gate exp(iθ ZZZ) or a product of such terms. The paper does not reproduce the gadget or list the diagonal gates it uses. If the gadget requires any 3-qubit or higher diagonal interaction, the quadratic-D family would not contain the hard IQP circuits, and Theorem III.1 would not imply PH-collapse. The remainder of the reduction — the V-gate construction, the particle-number invariant w(b)=α, the Rz and CP constructions in Appendix A, and the marginalization over the spin-down sector — is explicit and appears sound; this missing justification is the single unverified link.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proves worst-case classical hardness of sampling from the output distribution of single-layer unitary cluster Jastrow (1-UCJ) circuits, a chemistry-motivated ansatz. The main theorem (Theorem III.1) states that the class of Jordan-Wigner-transformed 1-UCJ circuits contains IQP circuits whose diagonal part is e^{iD}, where D is quadratic in Pauli Z operators. The proof explicitly constructs each IQP Hadamard gate from a restricted Givens rotation V_alpha, and each diagonal gate from products of D_pq operators, with a clear identification of IQP qubits with particle-number-conserving two-orbital subspaces. Combined with the Bremner-Jozsa-Shepherd result that IQP circuits are hard to weakly simulate under the assumption that the polynomial hierarchy does not collapse, the paper concludes that efficient classical sampling of 1-UCJ circuits with multiplicative error 1 <= c < sqrt(2) would collapse the polynomial hierarchy to the third level. A side result shows post-1-UCJ_JW = post-BQP.","tokens_in":15524,"tokens_out":47022,"duration_ms":446231,"significance":"If correct, this is a meaningful step toward quantum advantage in practical quantum chemistry, as it targets a circuit family used in VQE-type and QSCI algorithms rather than an artificial sampling problem. The reduction is explicit and checkable, and the paper honestly frames the result as worst-case hardness. The technical core—the emulation of IQP by 1-UCJ—is sound; I specifically verified that the Hadamard gadget in Ref. [31] requires only a controlled-Z gate and postselection (together with the boundary Hadamard gates), so the restriction to quadratic D does not weaken the hardness argument. This addresses the main potential concern about the hardness transfer.","major_comments":[],"minor_comments":[{"comment":"The justification that the complexity argument is unaffected by restricting IQP's diagonal part to the quadratic form of Eq. (1) is compressed; since this step is load-bearing for the hardness transfer, I recommend adding a brief explanation that the Hadamard gadget in Ref. [31] requires only a controlled-Z gate and postselected measurement (together with the boundary Hadamard gates), all of which are available in the e^{iD} family.","section":"II.B"},{"comment":"The statement '1-UCJ_JW ⊇ IQP' might be misread as referring to the unrestricted IQP class; because the paper adopts the quadratic-D restriction in Eq. (1), please define the notation explicitly (e.g., 'IQP with quadratic diagonal') to avoid ambiguity.","section":"Theorem III.1"},{"comment":"In Eq. (A7), the subscript 'n−α−0' should read 'n−α−1'; this is a typographical error in an otherwise clear derivation.","section":"Appendix A, Eq. (A7)"},{"comment":"The abstract refers to a 'single-layer UCJ circuit on 4n spin orbitals', while the proof constructs a 2n-qubit circuit after ignoring the spin-down sector; please clarify that the spin-down sector is a passive product state that is traced out, so the effective hard circuit has 2n qubits.","section":"Abstract and Section III.B"}],"recommendation":"minor_revision","confidential_remarks":"The paper is well-written and the main reduction is sound. The only issue is the terseness of the IQP hardness transfer in Section II.B, which can be addressed with a short clarification. I recommend acceptance after minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is the first hardness result for the UCJ ansatz, and the reduction from quadratic-Ising IQP to 1-UCJ is explicit and checks out. Second, the one soft spot the stress test flags—the compressed claim that the quadratic family implements the BJS universal diagonal gates—is not a real flaw. Bremner-Jozsa-Shepherd's Hadamard gadget uses only controlled-Z and single-qubit Z rotations, both 2- and 1-local, so the restricted IQP class in Eq. (1) contains the hard instances. The paper should have said that explicitly instead of in a single sentence, but the argument is sound.\n\nWhat the paper does well: the V-gate construction correctly implements Ry(-π/2) via the parity cancellation w(b)=α, the Dpq construction for Rz and CP is explicit and correct in Appendix A, and the authors state their limitations honestly (worst-case, no noise, no average-case). They also disclose the concurrent UCCSD work [43] and correctly note that their result doesn't follow from it.\n\nSoft spots, in proportion. The hardness is worst-case only; it doesn't tell you that the typical UCJ circuits used in QSCI or VQE are hard. That's a known limitation and they say so. The exposition of the restricted-IQP premise is too short—one paragraph—and a careful reader has to go back to the BJS gate set to verify. That's a minor presentation fix, not a correctness issue. Also, the construction uses 4n spin orbitals for n IQP qubits, and the parameters for the V gates are highly specific; the hard instances are contrived, which is fine for complexity theory but worth remembering when interpreting 'potential quantum advantage.'\n\nCitation pattern looks fine; the reduction relies on an external result with no author overlap, and the concurrent work is cited.\n\nBottom line: this deserves a serious referee. It's a clean, checkable reduction that answers a natural question. I'd recommend acceptance after the authors expand the paragraph on the BJS gate set and Hadamard gadget, and maybe clarify the spin-down tracing. If you're on the program committee, give it to a competent referee; it's not a desk reject.","headline":"A clean worst-case hardness reduction from quadratic-Ising IQP to single-layer UCJ; the stress-test's missing-gadget concern does not survive a check of the BJS gate set.","tokens_in":16050,"tokens_out":26001,"would_cite":true,"duration_ms":269301,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-16T12:23:22.913657+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}