{"id":"38656b3d-6da7-4360-87fc-178d37a4b4c5","arxiv_id":"2506.19778","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Noncontextual Pauli Hamiltonians have support up to 2^(n+1), block-diagonal spectra, and eigenstates with stabilizer rank linear in the number of qubits.","lead":"This paper gives proofs for several properties of noncontextual Pauli Hamiltonians, a class of quantum Hamiltonians that admit a classical description. It shows that such Hamiltonians can have up to 2^(n+1) terms and that every eigenstate can be written as a small superposition of stabilizer states.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proof of Theorem III.3 relies on a false LCU identity (Eq. 5); the stabilizer-rank claim is not established as written, though a projector-based repair appears available.","rationale":"The reader's CONDITIONAL verdict is reasonable. I identify the same central weak point in Theorem III.3's proof: Definition II.10, Eq. (5), is a false identity, and the claimed bound on stabilizer terms does not follow. The reader's named weakest assumption (the R=G∪A characterization, Corollary II.4) is a cited prior result and is likely sound; it is not where the proof breaks. I therefore partially agree. The false LCU identity is load-bearing because Theorem III.3 is the paper's main new contribution; without a valid proof of the O(n) stabilizer-rank statement, the abstract's central claim is unsupported as written. However, the theorem appears to be true and easily repairable via the projector (I±P*)/2, which is itself a linear combination of at most |A|+1 Paulis; applied to a stabilizer state it gives an eigenvector of the block. Thus the correct outcome is a conditional acceptance requiring revision of the proof, not rejection. This does not change the reader's verdict.","tokens_in":19107,"tokens_out":30352,"duration_ms":331561,"concrete_test":"Check Eq. (5) directly for the one-qubit anticommuting set A={X,Y,Z}: evaluate exp(-iθ Σ_{k≠w} iβ_k P_k P_w) in the Pauli basis and confirm it has more than three nonzero coefficients, so the identity is false. Then attempt to prove Theorem III.3 without Eq. (5) by applying P_±=(I±P*)/2 with P*=Σβ_i P_i to a computational basis state in the ν-block; if this yields a nonzero eigenvector with stabilizer rank at most |A|+1 for every block, the central claim is correct and only the proof needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Definition II.10, Eq. (5), claims R_LCU = exp(-iθ Σ iβ_k P_k P_w) = Σ_l d_l P_l. This identity is false in general: the exponential of a linear combination of Pauli operators is not a finite linear combination of |A| Pauli operators unless the exponent is a single Pauli or a very special algebra. For A={X,Y,Z} on one qubit, the exponent reduces to a linear combination of Paulis and the exponential has more than three Pauli-basis coefficients. Consequently, the proof of Theorem III.3 — that R†_LCU|ψ±(ν)⟩ has at most |A| stabilizer terms — does not follow from anything established in Section II.C. Since Theorem III.3 is the central claim of the paper, the submitted proof is invalid at its key step. This is a correctness gap in the argument as written, not merely a stylistic issue. The claim itself may be repairable: for a block H_nc(ν)=δI+||c||P* with P*=Σβ_i P_i, the operator P_±=(I±P*)/2 applied to any computational basis state in the stabilizer subspace gives an eigenvector expressed as at most |A|+1 stabilizer states, so a correct proof likely exists. But that repair is not in the manuscript, and Eq. (5) as written cannot be defended.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies noncontextual Pauli Hamiltonians, defined as Hamiltonians generated under the Jordan product from a set R = G ∪ A, where G is a commuting set of Pauli symmetries and A is a set of pairwise anticommuting Pauli operators. The main results are: (i) a support bound |H_nc| ≤ 2^{n+1}; (ii) a block-diagonal structure with respect to the symmetries G; (iii) a two-eigenvalue spectral theorem for each block, which is a constant plus a normalized linear combination of pairwise anticommuting Paulis; and (iv) the claim that every eigenvalue has an associated eigenvector of stabilizer rank O(n) (Theorem III.3). The paper also discusses degeneracies arising from the anticommuting set A. The central proof of Theorem III.3 relies on an LCU rotation identity in Definition II.10/Eq. (5).","tokens_in":19303,"tokens_out":19828,"duration_ms":197924,"significance":"If the stabilizer-rank theorem can be correctly established, the paper identifies a genuinely new family of Hamiltonians whose eigenstates admit efficient classical descriptions, going beyond commuting and stabilizer Hamiltonians. The block-diagonalization theorem (Theorem III.1) and the two-eigenvalue spectral structure (Theorem III.2) are elementary, clearly stated, and the proofs in Appendix B are essentially correct. The support bound in Corollary III.3 is a useful quantitative result, although the comparison with 'diagonal Hamiltonians' is misstated. The paper does not present numerical or code artifacts; its value is theoretical. The advertised main contribution is Theorem III.3, and as written that theorem is not established because its proof uses a false identity.","major_comments":[{"comment":"The displayed identity is mathematically false. The exponential of a linear combination of Pauli operators is not, in general, a finite linear combination of |A| Pauli operators. For example, with A={X,Y,Z} on one qubit, exp(-iθ(aX+bY+cZ)) = cos(θ)I - i sin(θ)(aX+bY+cZ), which contains the identity in addition to the three Pauli operators and is not of the form Σ_{l=1}^{|A|} d_l P_l. The standard LCU construction from [63] is a linear combination of unitaries, not an exponential of the summed Pauli products. Since the proof of Theorem III.3 in Section III.E explicitly uses Eq. (5) to bound the number of stabilizer terms by |A|, the central stabilizer-rank claim is not established as written.","section":"II.C, Definition II.10, Eq. (5)"},{"comment":"The eigenvector is written as |ψ_nc±(ν)⟩ = R†(ν)|ψ±(ν)⟩, but the convention adopted in Eq. (6) is R† Ô R = P. With H_nc(ν)=δ(ν)I+||c(ν)||Ô and Ô = R P_A R†, the eigenvector is R(ν)|ψ±(ν)⟩, not R†(ν)|ψ±(ν)⟩, unless R is Hermitian and self-inverse. The proof of Theorem III.3 should either apply the rotation on the correct side or explicitly prove that the LCU unitary is Hermitian and satisfies R²=I; otherwise the displayed vector is not generally an eigenvector.","section":"III.E, Eq. (20)"},{"comment":"The comparison underlying the claimed support result is misstated. The maximum size of a commuting set of Pauli operators on n qubits is 2^n, not 2n, and the support bound proved in Corollary III.3 is 2^{n+1}, not 2n+1. The abstract and introduction compare noncontextual Hamiltonians with 'diagonal Hamiltonians' using the numbers 2n+1 versus 2n, which is internally inconsistent with Corollary III.3. The correct comparison should be restated as 2^{n+1} versus 2^n.","section":"Introduction and II.C"}],"minor_comments":[{"comment":"The expansion of R_S as a product of rotations equal to a linear combination of 2|A|-1 Pauli operators also appears incorrect in general: a product of noncommuting single-Pauli rotations does not reduce to that few Pauli terms. This construction should be aligned with the precise statement in [63].","section":"II.C, Definition II.9, Eq. (4)"},{"comment":"The eigenvalue formula is written as E±(ν)=δ(ν)I±(... ), but an eigenvalue is a scalar, not an operator; the identity operator should be removed.","section":"III.D, Eq. (19)"},{"comment":"The paper aims to be self-contained, but Corollary II.4, which supplies the G∪A characterization used throughout Section III, is stated without proof and only attributed to 'Appendix B of [4]'. It should either be proved or explicitly labeled as an imported theorem.","section":"II.F, Corollary II.4"},{"comment":"The statement that every eigenstate of H_nc must be an eigenstate of each G_i is too strong when H_nc has degeneracies; the correct statement is that there exists a common eigenbasis. This imprecision should be corrected.","section":"II.F, near Definition II.14"},{"comment":"There are several typographical issues, including 'noncontexutal' on page 3, 'The proof we provide are straightforward' on page 2, and missing parentheses in Eq. (21). These should be fixed in a revision.","section":"Global (typos)"},{"comment":"The caption says 'Different 4-qubit noncontextual Pauli Hamiltonians by size of A', but it is not clear from the caption or the text whether the plots show all structures for each size of A or representative examples; this should be clarified.","section":"Figure 1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript has a repairable but central gap: the LCU identity in Eq. (5) is false, and Theorem III.3 depends on it. The projector-based construction (P±=(I±P*)/2) mentioned in review appears capable of repairing the theorem, but that repair is not present in the manuscript. I recommend major revision rather than rejection, since the correct fix appears to be local and within the scope of the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the take: this paper gives the first rigorous proofs for several structural facts about noncontextual Pauli Hamiltonians, but the proof of its headline eigenstate result relies on an LCU identity that is false as written, so the manuscript needs a real repair before that theorem is established.\n\nWhat's genuinely new: the support bound |H_nc| ≤ 2^{n+1} (Cor. III.3), the block-diagonal form with two eigenvalues per block (Thm III.1 and III.2), the degeneracy counting (Cor. III.5), and the explicit construction of eigenstates as R†|ψ±⟩. The proofs of III.1 and III.2 are elementary and correct. The counting argument for the support is clean. The paper is also honest that it is formalizing ideas from prior CS-VQE papers; several results are 'straightforward' but were unproven. That is a legitimate contribution.\n\nThe soft spot is Definition II.10 / Eq. (5). As written, R_LCU is defined as exp(-iθ Σ iβ_k P_k P_w) and then asserted equal to a linear combination of |A| Pauli operators. That identity is not true in general: the exponential of a linear combination of anticommuting Paulis has infinitely many Pauli-basis components unless the algebra is special. The proof of Theorem III.3 explicitly uses the 'at most |A| terms' property of R_LCU to bound stabilizer rank, so the proof does not go through as written. This is a correctness gap at the key step, not a stylistic issue.\n\nThat said, the underlying claim is likely repairable. As the stress-test note observes, for each block you can apply (I ± P*)/2 to a stabilizer-state eigenvector of the symmetry generators and get a linear stabilizer-rank eigenvector directly, without the LCU rotation. So the theorem may survive in a revised version. The authors also overstate the novelty in the abstract: linear stabilizer-rank states are already known; the new point is that these specific Hamiltonian eigenstates have that property, which is useful but not a new class of simulatable states.\n\nShould it go to peer review? Yes. The structural results are solid and the central claim is plausible and important for CS-VQE. A serious referee should ask for a corrected Definition II.10 and a rewritten proof of Theorem III.3, and then the paper is in good shape. If the repair fails, the linear stabilizer-rank theorem should be removed or qualified.","headline":"Solid structural results for noncontextual Pauli Hamiltonians, but the proof of the central stabilizer-rank theorem depends on a false LCU identity and needs a real repair.","tokens_in":19930,"tokens_out":3001,"would_cite":false,"duration_ms":29902,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that every eigenvalue of a noncontextual Pauli Hamiltonian has an associated eigenvector expressible as a sum of at most $O(n)$ stabilizer states, opening a new class of efficiently simulatable states.","keywords":["noncontextual Pauli Hamiltonian","stabilizer rank","contextuality","Jordan product","unitary partitioning","efficient classical simulation","Pauli Hamiltonians","eigenspectrum"],"falsifier":"Search the small-qubit cases: if any set of Pauli operators satisfies the paper's noncontextuality condition (commutation becomes transitive after removing universally commuting operators) yet its Jordan-closed support has no independent generating set of the form $G \\cup A$, then Corollary II.4 and the theorems built on it fail. Alternatively, construct the $R_{LCU}$ eigenstates for random noncontextual Hamiltonians and test whether their stabilizer rank ever exceeds $2(n-|G|)+1$.","tokens_in":18819,"feed_emoji":"⚛️","tokens_out":12303,"duration_ms":108153,"temperature":0.7,"pith_summary":"The paper establishes that noncontextual Pauli Hamiltonians—quantum Hamiltonians built from Pauli operators that admit a consistent noncontextual value assignment—remain classically tractable even though they can contain more interaction terms than diagonal Hamiltonians. Its central result is that for every eigenvalue of such a Hamiltonian there is an associated eigenvector whose stabilizer rank is linear in the number of qubits, so the eigenstate can be written as a superposition of at most about $2n$ stabilizer states. The authors also prove that a noncontextual Hamiltonian can contain up to $2^{n+1}$ distinct Pauli operators, twice the ceiling for a fully commuting Hamiltonian, and that each block of the Hamiltonian carries only two eigenvalues. If the proofs are correct, noncontextual eigenstates form a new class of classically simulatable states usable as inputs to quantum algorithms.","feed_headline":"Every noncontextual Hamiltonian has an O(n)-term eigenstate","feed_subtitle":"Each eigenvalue gets a state made of at most ~2n stabilizer terms, so it is classically describable.","key_machinery":"The load-bearing object is the generating set $R = G \\cup A$ of a noncontextual Pauli Hamiltonian: $G$ is a set of independent, mutually commuting Pauli operators serving as $Z_2$ symmetries, and $A$ is a set of pairwise anticommuting Pauli operators. Under the Jordan product—the symmetrized product $\\{A,B\\}/2$, which vanishes on anticommuting pairs—every noncontextual Hamiltonian is generated from such a set. The main technical engine is Theorem III.2, stating that any real linear combination of pairwise anticommuting Pauli operators is unitarily equivalent to a single Pauli operator and therefore has exactly two eigenvalues equal to plus and minus the Euclidean norm of its coefficients. The explicit unitary partitioning rotation $R_{\\mathrm{LCU}}$ realizes that equivalence as a linear combination of at most $|A|$ Pauli terms, and this expansion is what carries the stabilizer-rank bound.","core_discovery":"On the paper's own terms, the central discovery is Theorem III.3: for a general $n$-qubit noncontextual Pauli Hamiltonian, every eigenvalue has an associated eigenvector with stabilizer rank linear in $n$. The argument writes the Hamiltonian as generated by a Jordan-independent set $R = G \\cup A$, where $G$ is a set of commuting Pauli symmetry operators and $A$ is a set of pairwise anticommuting Pauli operators. The symmetries in $G$ block-diagonalize the Hamiltonian, and inside each block the remaining operator is a constant plus a normalized real linear combination of anticommuting Pauli operators, which the unitary partitioning rotation $R_{\\mathrm{LCU}}$ maps to a single Pauli operator. Applying that rotation to a stabilizer state produces a superposition of at most $|A| \\leq 2(n-|G|)+1$ stabilizer states, which gives the linear bound. The same structure yields the two-eigenvalue-per-block spectrum and the support bound $|H_{nc}| \\leq 2^{n+1}$.","pith_inferences":["A direct numerical extension would build the $R_{\\mathrm{LCU}}$ expansion for random noncontextual Hamiltonians and check that the stabilizer rank of the constructed eigenstates never exceeds $2(n-|G|)+1$, and whether the bound is ever tight.","If the converse question the authors raise holds—that every state of linear stabilizer rank has a noncontextual parent Hamiltonian—then noncontextual Hamiltonians would characterize a general class of classically simulable Hamiltonians; if it fails, an even broader simulable class exists.","The NP-completeness of the ground-state search means the efficient eigenstate description is best read as a verification and initialization tool: it makes a chosen sector classically tractable, but it does not by itself remove the hardness of choosing the right sector."],"forward_implications":["Noncontextual Pauli Hamiltonians can include up to $2^{n+1}$ distinct Pauli terms, twice the maximum for a fully commuting (diagonal) Hamiltonian, so they cover more physical interactions while staying classically describable.","Each symmetry block of a noncontextual Hamiltonian has exactly two eigenvalues $E_{\\pm}(\\vec{u}) = \\delta(\\vec{u}) \\pm \\|\\vec{c}(\\vec{u})\\|$, so the spectrum can be evaluated classically sector by sector rather than by diagonalization.","For every eigenvalue, an eigenvector can be constructed as a superposition of at most $2(n-|G|)+1$ stabilizer states; when $A$ is empty the Hamiltonian is a stabilizer Hamiltonian and its eigenvectors are single stabilizer states.","Eigenvalue degeneracies come in factors of $2^{\\lceil(|A|-1)/2\\rceil-1}$ from the anticommuting part, with additional degeneracies possible from symmetries beyond the Pauli $Z_2$ symmetries in $G$.","Finding the noncontextual ground state remains NP-complete because there are $2^{|G|}$ symmetry sectors, but once a sector is chosen the corresponding eigenstate is efficiently constructible."],"supporting_citations":[{"why":"Supplies the definition of noncontextual Pauli Hamiltonians and the $G \\cup A$ generating-set characterization (Corollary II.4) on which the spectral theorems rest.","marker":"[4]"},{"why":"Provides Theorem II.2, the equivalence between noncontextuality and transitivity of commutation, from which the generating-set structure follows.","marker":"[3]"},{"why":"Constructs the unitary partitioning rotations $R_S$ and $R_{LCU}$ that map an anticommuting linear combination to a single Pauli operator.","marker":"[63]"},{"why":"Bounds the size of a pairwise anticommuting set on $m$ qubits by $2m+1$, giving the linear scaling in the stabilizer-rank theorem.","marker":"[64]"},{"why":"Introduces the contextual-subspace framework and the closure-under-inference definition used in the compatibility-graph characterization.","marker":"[2]"},{"why":"Details the $R_{LCU}$ construction and its term-count scaling, supporting the claim that the eigenstate expansion has at most $|A|$ stabilizer terms.","marker":"[6]"},{"why":"Establishes that states of bounded stabilizer rank admit efficient classical simulation, which is why the theorem matters.","marker":"[39]"},{"why":"Provides the Clifford tapering construction used to map symmetry generators to single-qubit $Z$ operators and block-diagonalize the Hamiltonian.","marker":"[61]"}],"fun_headline_variants":["Noncontextual Pauli eigenstates have linear stabilizer rank","Linear stabilizer rank for noncontextual Pauli eigenstates","Noncontextual Hamiltonians: more Paulis, still classically simulable","Proof: noncontextual eigenstates need only O(n) stabilizer terms","Noncontextual Pauli Hamiltonians admit efficient classical description"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that every noncontextual set of Pauli operators is generated under the Jordan product by a set $R = G \\cup A$ with $G$ universally commuting and $A$ pairwise anticommuting; the paper takes this characterization from earlier work without reproving it, and if it is incomplete the block structure and stabilizer-rank theorems cover only a subclass of noncontextual Hamiltonians.","fun_headline_variants_meta":{"raw":{"variants":["Noncontextual Pauli eigenstates have linear stabilizer rank","Linear stabilizer rank for noncontextual Pauli eigenstates","Noncontextual Hamiltonians: more Paulis, still classically simulable","Proof: noncontextual eigenstates need only O(n) stabilizer terms","Noncontextual Pauli Hamiltonians admit efficient classical description"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000607,"raw_usage":{"total_tokens":2821,"prompt_tokens":931,"completion_tokens":1890,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":547,"completion_tokens_details":{"reasoning_tokens":1799}},"tokens_in":547,"tokens_out":1890,"duration_ms":14166,"temperature":1.0,"reasoning_tokens":1799,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:25:47.343413+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search the small-qubit cases: if any set of Pauli operators satisfies the paper's noncontextuality condition (commutation becomes transitive after removing universally commuting operators) yet its Jordan-closed support has no independent generating set of the form $G \\cup A$, then Corollary II.4 and the theorems built on it fail. Alternatively, construct the $R_{LCU}$ eigenstates for random noncontextual Hamiltonians and test whether their stabilizer rank ever exceeds $2(n-|G|)+1$.","supporting_citations":[{"cited_title":"Classical Simulation of Noncontextual Pauli Hamiltonians","cited_arxiv_id":"2002.05693","evidence_quote":"Supplies the definition of noncontextual Pauli Hamiltonians and the $G \\cup A$ generating-set characterization (Corollary II.4) on which the spectral theorems rest."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides Theorem II.2, the equivalence between noncontextuality and transitivity of commutation, from which the generating-set structure follows."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Bounds the size of a pairwise anticommuting set on $m$ qubits by $2m+1$, giving the linear scaling in the stabilizer-rank theorem."},{"cited_title":"Ralli, T","cited_arxiv_id":null,"evidence_quote":"Details the $R_{LCU}$ construction and its term-count scaling, supporting the claim that the eigenstate expansion has at most $|A|$ stabilizer terms."}],"review_version":1}