{"id":"505995a2-7c13-4995-8b00-b489ec560532","arxiv_id":"2411.08119","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"In slow Trotterized annealing on MaxCut, genuine multipartite entanglement rises to a barrier near the minimum gap and falls again, and the chance of finding the optimum is bounded by one minus the final entanglement.","lead":"The authors measure genuine multipartite entanglement, using a distance-to-product-states metric, in Trotterized quantum annealing and QAOA on MaxCut instances. They find a multipartite entanglement barrier linked to the energy gap and prove that final success probability is capped by one minus the final entanglement, with simple overlaps bounding entanglement throughout.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The analytical bound Eq. (14) is sound, but the claimed generic entanglement barrier depends on the hand-set symmetry-breaking f=0.05; without an f-scan, the numerical central claim is not established.","rationale":"I read the paper's central analytical claim as Eq. (14) and its corollary relating final success probability to final GGM. That part is mathematically sound: for any pure state, the GGM is the minimum distance to a bipartite product state, and each cost-Hamiltonian eigenstate is a product state, so 1−|ci|² is a valid upper bound. The paper's stronger and more interesting claim is the numerical multipartite entanglement barrier and its correlation with success. This is where the argument is least secure. The barrier is explicitly linked in App. C to a near-GHZ superposition of the two Z2-related optimal solutions |e0⟩ and |e1⟩. Such a superposition is a direct consequence of the near-degeneracy of those states, which the authors create by adding f Z0 with f = 0.05. They motivate this value by saying it lifts the degeneracy while preserving the eigenlevel structure, but they do not demonstrate that the barrier is robust to the choice of f or to the form of the symmetry-breaking term. A reader cannot tell whether the observed peak at G2 ≈ 1/2 is a generic feature of annealing through a symmetry-related avoided crossing or a finite-bias artifact. The paper does include some independent support: exact diagonalization over 100 instances, a public dataset, and a clear explanation of the pairing mechanism. It also honestly notes that some instances do not show a barrier. But without an f-scan, the generality claim remains conditional. I therefore keep the CONDITIONAL verdict. The proposed test is computationally inexpensive at N = 8 and would directly settle whether the barrier is robust. I agree with the reader that f = 0.05 is the weakest assumption; my concern is the same one, phrased as a robustness question rather than an internal contradiction.","tokens_in":623,"tokens_out":6384,"duration_ms":183116,"concrete_test":"Re-run the exact-diagonalization/TQA analysis of Fig. 5(c) for the same 100 eight-vertex instances with a range of symmetry-breaking offsets f ∈ {0.005, 0.01, 0.02, 0.05, 0.1, 0.2} and also with a random local field h_i Z_i instead of f Z0. For each f, compute the instance-averaged G2(s/p), the peak height and position, the fraction of instances with a barrier (G2_max > G2_final + 10^−4), and the final |c0|². If the barrier peak and the |e0⟩/|e1⟩ pairing persist for f down to 0.005, or vanish only for f values that quantitatively change the gap structure, the claim is robust. If the averaged peak disappears or shifts significantly for f ≤ 0.02, the barrier is an artifact of the hand-set f = 0.05.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central inequality G2(|ψ⟩) ≤ min{1−|c0|², 1−|d+|², 1/2} is a direct consequence of the definition of GGM and of the fact that the eigenstates |ei⟩ of HC in Eq. (8) are computational-basis product states. I see no flaw in that part. The paper's headline numerical phenomenon, the multipartite entanglement barrier, is what the central claim actually rests on. That barrier is shown in App. C and Fig. 12 to arise from a near-equal superposition of the Z2-paired basis states |e0⟩ and |e1⟩, producing a near-GHZ state with GGM ≈ 1/2. The pairing is created by the ad hoc symmetry-breaking term f Z0 with f = 0.05 fixed for all instances. The paper does not test whether the barrier survives for smaller f, for f → 0, or for a different symmetry-breaking term, such as a random local field. Since the perturbation is chosen precisely to keep the paired states close in energy while making the ground state unique, the observed barrier could be an artifact of that particular perturbation scale rather than a generic feature of quantum optimization. Moreover, the success-probability bound |c0|² ≤ 1 − G2_final is trivial for any product-state solution; its significance relies on the final GGM being non-negligible in typical runs. If the barrier is an artifact, the claimed resource-theoretic role of genuine multipartite entanglement in optimization is not supported. This is a numerical-robustness issue, not an internal inconsistency, but it is the load-bearing assumption.","agreement_with_reader":"agree"},"referee_report":null,"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: one clean analytical result and one interesting numerical observation. The analytical result is Eq. (14): G2 ≤ min{1−|c0|², 1−|d+|², 1/2}. It follows directly from the GGM definition and the fact that the cost-Hamiltonian eigenstates are product states; I see no flaw there. The numerical observation, a multipartite entanglement barrier in Trotterized quantum annealing, is plausible but not fully nailed down, because it rests on a single value of the symmetry-breaking offset f = 0.05.\n\nWhat's new: they compute the generalized geometric measure across TQA and QAOA, and derive the overlap-based upper bounds. These bounds are simple, experimentally accessible, and not in the prior literature they cite. The paper also does something useful in showing how GGM differs from fixed-bipartition von Neumann entropy, including a disconnected-graph example where the entropy is large but GGM vanishes. The data is on Zenodo, which helps.\n\nThe soft spots: the barrier mechanism, shown in App. C, is a near-GHZ superposition of the Z2-paired states |e0⟩ and |e1⟩. That pairing is created by the ad hoc field fZ0 with f fixed to 0.05. The paper never scans f or tries a different symmetry-breaking term, so we don't know whether the barrier survives as f → 0 or for a random local field. That's a real gap, and it's the load-bearing part of the numerical claim. I'd call it an addressable weakness rather than a fatal one. Also minor: the QAOA study uses a single instance; the conditional-probability curves use arbitrary ε thresholds; and there's no simulation code, though the dataset is public.\n\nWho this is for: people benchmarking near-term optimization hardware, and anyone interested in whether multipartite entanglement is a resource or a byproduct in quantum optimization. It deserves a serious referee: the analytical bound is solid, and the barrier observation is worth testing even if it isn't proven generic. If I were the editor, I'd send it to review and ask the authors to add an f-scan and at least one more graph family. I'd probably cite the bound myself; I'm more cautious about the barrier claim.","headline":"The analytical GGM bound is solid and the barrier observation is worth reporting, but the paper should scan the symmetry-breaking offset before claiming the barrier is generic.","tokens_in":20668,"tokens_out":4452,"would_cite":true,"duration_ms":39511,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-12T21:57:37.194759+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}