{"id":"a398c9ed-b1dd-46c1-bfc6-27dc78357b1a","arxiv_id":"2607.18596","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"VGP, not stoquasticity, is the invariant boundary: the VGP-local Hamiltonian problem is StoqMA-complete, and recognizing VGP is PSPACE-complete.","lead":"This paper replaces stoquasticity with vanishing geometric phase (VGP) as the correct boundary for quantum sign-problem and complexity classes, proving that VGP Hamiltonians still have a StoqMA-complete local Hamiltonian problem. It also shows deciding VGP is PSPACE-complete in general, giving a sharp computational separation between recognizing and exploiting the property.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Collapse VGPMA=StoqMA hinges on Lemma 3's unverified StoqMA measurement of nonlocal diagonal F; if that lemma fails, Theorem 4 and Corollary 6 do not follow.","rationale":"The reader's weakest assumption identifies the same load-bearing point I find: Lemma 3 is the only place where the proof of Theorem 4 goes beyond the known StoqMA containment for stoquastic local Hamiltonians. The rest of the argument—VGP implies a cospectral stoquastic proxy via Proposition 1, Lemma 2 supplies the decomposition with efficiently computable diagonal terms—is coherent assuming Lemma 3. My stress-test did not find an independent fatal flaw; the main additional observation is the paper's internal note that Lemma 3 is 'not necessary' for Theorem 4 even though the proof of Theorem 4 explicitly invokes it. This ambiguity, together with the non-self-contained nature of the Appendix B proof, justifies keeping the conditional verdict rather than accepting outright. The proposed concrete test—re-deriving the measurement primitive and simulating a small nonlocal instance—would settle whether the concern actually lands. If the test passes, the conditional could be upgraded; if it fails, the central collapse is unsupported.","tokens_in":40744,"tokens_out":17898,"duration_ms":221664,"concrete_test":"Independently re-derive Lemma 3 from Ref. [7]'s StoqMA verifier for X⊗Π. First, write the verifier explicitly for a general diagonal projector Π given only by a reversible circuit C_Π, and confirm that the acceptance probability is exactly affine in <psi|X⊗Π|psi> with no locality assumption and only one final Hadamard measurement. Then instantiate the binary-expansion construction for a concrete nonlocal F on two qubits, e.g. F = diag(1/2, 1, 0, 1/2) with M=1, and compute the acceptance probability on two distinct witness states by direct state-vector simulation. Check that the result matches α + β<psi|X⊗F|psi> ± δ with the claimed constants and error bound (b = ceil(log2(M/δ))+O(1)). If the affine formula fails for nonlocal F, or if the construction requires mid-circuit measurements, Lemma 3 is invalid and the collapse is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The collapse VGPMA=StoqMA (Corollary 6) rests on Theorem 4, and the proof of Theorem 4 departs from Ioannou et al. exactly at Lemma 3. Lemma 3 asserts that any nonnegative diagonal operator F with a poly-size reversible entry oracle can be measured in StoqMA, so that the acceptance probability is affine in <psi|X⊗F|psi>/M ± delta. This is load-bearing because the stoquastic proxy S(H) in Lemma 2 is generally nonlocal: the original Stoq-LH containment (Theorem 5) applies to local stoquastic Hamiltonians, and the proof extends it to S(H) only by replacing the local H_j with nonlocal but efficiently evaluable diagonal operators supplied by Lemma 3. The Appendix B proof of Lemma 3 is not self-contained: it invokes a 'measurement primitive' from Ref. [7] without re-deriving the affine acceptance formula, and the binary-expansion step 'samples ℓ ... using fresh ancillas exactly as ... Ref. [7]' does not explicitly establish that the resulting circuit—only X/CNOT/Toffoli with a single final Hadamard measurement—implements the claimed average. If the primitive secretly requires locality of the projector, or if the normalization 1/(1-2^{-b}) or the truncation error is mishandled, then Lemma 3 fails for nonlocal F, and the proof of Theorem 4 collapses. The paper itself notes that Lemma 3 is 'not necessary' for Theorem 4 while later invoking it in the proof of Theorem 4, making the dependency unclear; if a weaker lemma is intended, it should be stated and proven.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that vanishing geometric phase (VGP), rather than stoquasticity, is the correct diagonal-unitary-invariant boundary for sign-problem-free Monte Carlo simulation and for the complexity of ground-state energy problems. It defines the class VGPMA as problems reducible to the local Hamiltonian problem under a VGP promise, and claims that VGP-LH is StoqMA-complete (Theorem 4), yielding VGPMA = StoqMA (Corollary 6). The proof extends the StoqMA containment of Ioannou et al. to the nonlocal stoquastic proxy S(H) via a generalized measurement lemma (Lemma 3). The paper also constructs VGP Hamiltonians that are hard to stoquastize (Theorem 2), proves PSPACE-completeness of VGP recognition for local and geometrically local Hamiltonians (Theorem 8), and identifies tractable cases for recovering stoquastizing diagonals (Theorem 9, Corollary 9).","tokens_in":41167,"tokens_out":39931,"duration_ms":435639,"significance":"If the proof of Lemma 3 can be made fully rigorous, the collapse VGPMA = StoqMA is a significant result: it shows that the StoqMA-completeness of the local Hamiltonian problem is preserved under diagonal unitary transformations, so stoquasticity is not the operative feature. The separation examples (Example 4, Theorem 2) and the efficient stoquastization of PMR-local VGP Hamiltonians are valuable contributions. The PSPACE-completeness of VGP recognition is a striking contrast to the tractability of using the VGP promise. However, the current manuscript leaves a load-bearing lemma insufficiently proved and contains an internal contradiction about its role, so the headline claims are not yet established to the standard required for publication.","major_comments":[{"comment":"Lemma 3 is load-bearing for Theorem 4 and hence for Corollary 6, but its proof is not self-contained. Appendix B invokes an unstated 'measurement primitive of Ref. [7]' and asserts that locality is irrelevant without proving the affine acceptance formula for nonlocal projectors. The binary-expansion step ('sample ℓ ... using fresh ancillas exactly as Ref. [7]') must be specified as a coherent circuit over {X,CNOT,Toffoli} with no intermediate measurements, and the constants α, β must be shown to preserve the 1/poly gap. Moreover, the sentence after Lemma 3 says it is 'not necessary' for Theorem 4, while the proof of Theorem 4 explicitly invokes it; if a weaker lemma suffices, it should be stated and proved.","section":"Section V.A / Lemma 3 / Appendix B"},{"comment":"The claim that the Ioannou et al. decomposition can be constructed for S(H) using only the operators eH_{S,x} is asserted in property (c), not proved. Since S(H) can have exponential Pauli rank (Example 4), this is nontrivial. The proof must either provide the actual construction or state precisely which properties of the Ref. [7] construction are used, and why they survive when the diagonal terms are nonlocal but efficiently evaluable.","section":"Section V.A / Lemma 2"},{"comment":"The PSPACE-hardness reduction is presented at a high level and contains at least one questionable step: the footnote allowing an Or vertex with self-loops is not a legal NCL gadget, and the parity argument depends on it. The equivalence between the existence of an odd-length cycle and FreeNCLRev acceptance needs a rigorous proof, especially the claim that illegal configurations cannot contribute non-VGP cycles. This is a headline result and needs a complete, checkable proof.","section":"Section VI.B / Appendix D (Theorem 8)"}],"minor_comments":[{"comment":"The statement that Lemma 3 is 'not necessary' for Theorem 4 directly contradicts its use in the proof of Theorem 4. Please reconcile this or remove the disclaimer.","section":"Section V.A"},{"comment":"The proof asserts that each PMR term D_iP_i may be decomposed as a sum of whole local terms h_β. In general a PMR term is a sum of parts of local terms. The conclusion for geometrically local Hamiltonians is still correct when the interaction graph has bounded degree, but the proof should be rewritten.","section":"Section IV.A / Proposition 3"},{"comment":"The StoqMA verifier definition should clarify whether intermediate measurements are allowed. If not, the 'sampling' in Lemma 3 must be implemented coherently; this is related to Major Comment 1 and should be explicit.","section":"Definition 26"},{"comment":"Several missing spaces (e.g., 'isStoqMA-complete') and the dense formatting of Table I make the paper hard to read. Please fix typographical issues.","section":"Abstract / Table I"}],"recommendation":"major_revision","confidential_remarks":"The paper is ambitious and the main intuition is plausible, but the proof of Lemma 3 must be completed before the central claim can be accepted. The NCL self-loop issue in Theorem 8 also needs to be resolved. I would not reject on the current evidence; the issues appear fixable within the manuscript's scope. If the authors can supply a full proof of Lemma 3 and tighten the PSPACE-hardness reduction, this would be a solid contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Things to know: this paper argues that the stoquastic/non-stoquastic divide should be replaced by a diagonal-unitary-invariant notion, VGP, and proves that the local Hamiltonian problem under the VGP promise is StoqMA-complete, collapsing VGPMA to StoqMA. If the proof holds, that is a major reframing of where the sign problem boundary sits. The paper also has a PSPACE-completeness result for recognizing VGP in local Hamiltonians and some nice examples separating VGP from stoquasticity.\n\nWhat's genuinely new: the StoqMA-completeness of VGP-LH and the collapse VGPMA=StoqMA are not in the prior literature. The VGP concept is from Hen 2021, and the containment technique is from Ioannou et al., but the extension is nontrivial. The PMR-locality notion is new and useful, and the efficient stoquastization results for PMR-local VGP Hamiltonians look solid. The PSPACE-hardness reduction through NCL is intricate and I did not find an obvious bug.\n\nThe soft spot: the proof of Theorem 4 rests on Lemma 3, which extends the StoqMA measurement primitive to nonlocal diagonal operators via a reversible entry oracle. That lemma is load-bearing. The appendix proof is a sketch: it defers to a 'measurement primitive' from Ref [7] without re-deriving the affine acceptance formula, and the binary-expansion sampling step needs more detail on how the circuit implements the claimed average with the proper normalization. The paper's own remark that Lemma 3 'is not necessary' for Theorem 4 is confusing, since the proof of Theorem 4 explicitly uses it. If a weaker lemma suffices, it should be stated and proven. This is not a fatal flaw on its face—the lemma looks plausible—but it needs independent verification.\n\nA smaller issue: the definition of VGPMA via many-one reducibility to a promise problem is a bit unusual, though not wrong. And some example calculations, e.g. Example 4, are terse.\n\nOverall: the central argument is coherent, the paper is honest about its limitations, and the main claim could well be true. The right response is a serious referee, with instructions to dig into Lemma 3 and the NCL reduction, not a desk reject.","headline":"VGP replaces stoquasticity as the real boundary for StoqMA — plausible, significant, but the main containment depends on Lemma 3, which needs a closer look.","tokens_in":41658,"tokens_out":5669,"would_cite":true,"duration_ms":60199,"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":"The local Hamiltonian problem stays StoqMA-complete under a vanishing-geometric-phase promise, so diagonal unitary transformations cannot enlarge StoqMA.","keywords":["vanishing geometric phase","stoquastic Hamiltonians","sign problem","StoqMA","local Hamiltonian problem","adiabatic quantum computation","quantum Monte Carlo","PSPACE-completeness"],"falsifier":"Construct a VGP local Hamiltonian whose stoquastic proxy S(H) has entries that no polynomial-size reversible circuit can approximate, or exhibit a nonnegative diagonal operator F for which an entry-oracle measurement fails to produce acceptance probability affine in ⟨ψ|X⊗F|ψ⟩ at inverse-polynomial precision; either would invalidate the measurement lemma and break the paper's central containment.","tokens_in":40634,"feed_emoji":"⚛️","tokens_out":5798,"duration_ms":66403,"temperature":0.7,"pith_summary":"This paper argues that the real boundary for sign-problem-free quantum systems and for quantum complexity is not the stoquastic/non-stoquastic dichotomy, but a basis-independent geometric condition: vanishing geometric phase (VGP), the requirement that every closed walk in a Hamiltonian's transition graph accumulates phase 1. The central result is that the local Hamiltonian ground-state energy problem remains StoqMA-complete under the promise that the input has VGP, so the class VGPMA collapses to StoqMA. The paper also constructs VGP Hamiltonians that are provably hard to stoquastize yet admit efficient recognition of VGP, shows that recognizing VGP is PSPACE-complete for general local Hamiltonians and coNP-hard already for 2-local ones, and argues that non-VGP, not merely non-stoquastic, instantaneous Hamiltonians are necessary for the standard route to universal adiabatic quantum advantage. If right, the 'stoquastic vs non-stoquastic' framing should be replaced by 'VGP vs non-VGP' in discussions of the sign problem, Hamiltonian complexity, and quantum annealing.","feed_headline":"Stoquasticity is not the boundary; geometric phase is","feed_subtitle":"A proof that VGP Hamiltonians remain StoqMA-complete resets the sign-problem and adiabatic-advantage debate.","key_machinery":"The central object is the holonomy of a closed walk in the Hamiltonian's transition graph G_H: the product of normalized off-diagonal entries along the walk, multiplied by (-1) to the walk length. VGP is the condition that every such holonomy equals 1, equivalently that a stoquastizing diagonal unitary exists. The stoquastic proxy map S(H) carries the argument: it is cospectral to any VGP Hamiltonian and provides a stoquastic representative without exhibiting the diagonal. The paper's key technical step is a measurement lemma that lets a StoqMA verifier measure nonlocal nonnegative diagonal terms given only a polynomial-size reversible circuit that approximates their entries, which is what e","core_discovery":"On its own terms, the paper establishes that VGP is the diagonal-unitary-invariant notion of sign-problem-freeness that inherits StoqMA's computational boundary. A Hamiltonian has VGP exactly when a diagonal unitary D makes it cospectral to a stoquastic Hamiltonian, namely the stoquastic proxy S(H) that replaces each off-diagonal entry by its negative absolute value. The proof that VGP-LH is in StoqMA never constructs D and never verifies VGP; it uses the VGP promise only to guarantee that S(H) is cospectral to H, then feeds efficiently queryable entries of S(H) into a StoqMA verifier through a generalized measurement lemma that tolerates nonlocal diagonal terms. The resulting collapse VGPMA","pith_inferences":["If the collapse VGPMA = StoqMA is correct, any future separation between StoqMA and QMA must be witnessed by Hamiltonians that are provably non-VGP rather than merely non-stoquastic.","A testable extension is to benchmark quantum annealing devices with VGP-but-not-stoquastic drivers, such as bipartite exchange or XY drivers, against their explicitly stoquastic twins; the paper predicts no sign-problem-derived performance difference.","The PSPACE-completeness of VGP recognition suggests a practical screening discipline: use the paper's efficient sufficient conditions for VGP before attempting expensive stoquastization or non-VGP certification.","The generalized measurement lemma, if it survives scrutiny, opens the door to other nonlocal 'sign-curing' maps beyond the absolute-value proxy that admit efficient entry oracles and preserve StoqMA containment.",""],"forward_implications":["VGP-LH is StoqMA-complete, so hiding stoquasticity behind any diagonal unitary adds no verification power; equivalently VGPMA = StoqMA.","The frustration-free local Hamiltonian problem for PMR-local VGP Hamiltonians is in MA, matching the stoquastic case.","Recognizing VGP is PSPACE-complete for local, geometrically local, and PMR-local Hamiltonians, and coNP-hard for 2-local Hamiltonians, so exploiting the VGP promise is much easier than checking it.","For standard adiabatic schedules with sign-preserving drivers, VGP is independent of schedule and instance; non-VGP drivers are necessary for the usual universality route, while non-stoquastic but VGP drivers confer no sign-problem-based advantage.","Under a VGP promise, the partition function, thermodynamic quantities, and diagonal observables agree exactly with the stoquastic proxy; differences appear only for state-level tasks that require phase data.",""],"fun_headline_variants":["Geometric phase, not stoquasticity, sets quantum boundaries","VGP Hamiltonians inherit StoqMA-complete status","Sign-problem boundary is geometric phase, not stoquasticity","New proof: geometric phase defines StoqMA boundary","Vanishing geometric phase: the true computational divide"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that a small reversible circuit that approximates entries of the (possibly nonlocal) stoquastic proxy is sufficient to implement the StoqMA verifier's measurement with the required accuracy; if this generalized measurement premise fails, the proof that VGP-LH is in StoqMA fails, and with it the collapse VGPMA = StoqMA.","fun_headline_variants_meta":{"raw":{"variants":["Geometric phase, not stoquasticity, sets quantum boundaries","VGP Hamiltonians inherit StoqMA-complete status","Sign-problem boundary is geometric phase, not stoquasticity","New proof: geometric phase defines StoqMA boundary","Vanishing geometric phase: the true computational divide"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000177,"raw_usage":{"total_tokens":1155,"prompt_tokens":795,"completion_tokens":360,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":539,"completion_tokens_details":{"reasoning_tokens":278}},"tokens_in":539,"tokens_out":360,"duration_ms":4517,"temperature":1.0,"reasoning_tokens":278,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T14:56:00.466453+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct a VGP local Hamiltonian whose stoquastic proxy S(H) has entries that no polynomial-size reversible circuit can approximate, or exhibit a nonnegative diagonal operator F for which an entry-oracle measurement fails to produce acceptance probability affine in ⟨ψ|X⊗F|ψ⟩ at inverse-polynomial precision; either would invalidate the measurement lemma and break the paper's central containment.","supporting_citations":[],"review_version":1}