{"id":"18b3b9a4-c85a-4309-9747-a77dcebbbb82","arxiv_id":"2607.28479","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"GRAPE-optimized global drives compress Hadamard sequences from ~3000 ns to ~320 ns and raise fidelity under strong relaxation from ~0.64 to ~0.96 in a 15-qubit globally driven ladder.","lead":"Pulse shaping of a shared global drive can cut gate times by about 10× and largely restore one-qubit fidelities under relaxation in a ladder-style superconducting processor. That matters because every physical qubit feels the noise, so you cannot hide the logical register behind local shielding.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"Favorable |gg⟩ inputs and global fidelity may overstate the restored-fidelity claim under pulse compression.","rationale":"The reader correctly flags the phenomenological local GKSL model (Sec. II, App. C) as a modeling limitation; Appendix C shows invariance of bare σ−/σz under RF+RWA, but that is not a secular dressed derivation, so the concern is real. It is not, however, the single most load-bearing issue for the strongest claim as written. Inside the authors’ own model the numerics are coherent, data are deposited, and the time-compression mechanism is physically transparent. The sharper threat to the claim “restoring high gate fidelities” is the narrow logical input and the global-F metric: the paper repeatedly uses the most favorable |gg⟩ state, notes that a full logical average would be needed, and never reports a noisy process fidelity or superposition sweep for the optimized pulse. That gap directly limits how far Table I/Fig. 6 can be read as evidence that global optimal control restores computational gates when dissipation hits every qubit. Keeping CONDITIONAL is right; the condition should emphasize broader logical-state/process averaging (and clearer waveform artifacts) at least as strongly as dressed noise. No change to REJECT is warranted: the simulation result inside the stated model stands.","tokens_in":22522,"tokens_out":740,"duration_ms":15498,"concrete_test":"Re-evaluate the GRAPE Hadamard of Fig. 5/6 under the same γ− sweep and 2% disorder, but average the logical process fidelity (or at least F over the six Pauli eigenstates of the target qubit, including |+⟩/|−⟩ and |e⟩) rather than only |gg⟩ global overlap. If F_process(0.08) falls below ~0.9 or the gain vs. the rectangular pulse drops below ~3×, the headline mitigation claim needs qualification.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim (Table I, Fig. 6) is that GRAPE compression restores high Hadamard fidelity under relaxation acting on all physical qubits. That result is shown almost exclusively for the favorable payload |ψ_in⟩=|g1 g2⟩ (Sec. III A–C), which the authors themselves call “the most robust payload” because it carries neither logical excitations for σ− nor static coherences for σz. The only population sweep (Fig. 4) is noiseless and at fixed relative phase φ=0, so it does not probe dephasing of logical superpositions or relaxation of |e⟩ under the short optimized drive. Global fidelity F (Eq. 9) further folds in the entire Néel/ferro background; a high F can therefore coexist with degraded logical-subspace process fidelity once the ICC carries superpositions. If the order-of-magnitude gain shrinks or vanishes for a Haar-averaged or process-fidelity metric on logical inputs, the abstract claim that shaping “restores high gate fidelities” is overstated for computation.","agreement_with_reader":"partial"},"referee_report":{"model":"grok-4.5","summary":"The manuscript studies decoherence in a globally driven superconducting ladder architecture that encodes logical qubits in domain-wall (well-formed) states. Using an RF+RWA effective Hamiltonian and a local GKSL model with amplitude-damping and dephasing jumps, the authors simulate ICC translation, Hadamard, and CZ operations via MPS quantum trajectories. They find that relaxation is typically more damaging than dephasing at equal bare rates, that CZ is relatively robust because it is short, and that GRAPE-optimized global pulses compress the Hadamard from ~3000 ns to ~320 ns, raising fidelity under relaxation from F≈0.64 to F≈0.96 at γ−=0.08 (including 2% frequency disorder). The central message is that decoherence acting on the entire physical lattice—including spectator qubits that sustain Néel/ferro order—can be mitigated purely by temporal shaping of the global drive.","tokens_in":22785,"tokens_out":1381,"duration_ms":38989,"significance":"If the numerical claims hold under broader logical inputs and a more carefully justified open-system model, the work is a useful and timely contribution to globally controlled superconducting architectures. It connects an existing ladder encoding to concrete decoherence metrics and shows that optimal control can simultaneously address static disorder and dynamical noise without local addressing—an attractive hardware-level message. Strengths include an explicit control decomposition (Appendix B), a documented trajectory/MPS pipeline with stated cutoffs (N_traj=2000, max bond 200, λ=10−12), public data, and a clear Table I / Fig. 6 comparison of standard vs optimized Hadamard performance. The result is incremental relative to the authors’ prior architecture and disorder papers, but the decoherence-mitigation numbers are new and actionable for near-term experiments.","major_comments":[{"comment":"Sec. III A–C and Table I(b)/Fig. 6: the restored-fidelity claim under GRAPE is demonstrated almost exclusively for the favorable payload |ψ_in⟩=|g1 g2⟩, which the text itself calls “the most robust payload” (no logical excitations for σ−, no static logical coherences for σz). Fig. 4 only checks the optimized pulse in the noiseless limit over populations at fixed φ=0. The abstract and conclusion state that shaping “restores high gate fidelities” for computation, but that is not yet shown for logical superpositions or |e⟩ under dissipation. Please add at least (i) optimized Hadamard fidelities vs γ− for a small set of logical inputs spanning superpositions (or a Haar/process average on the ICC), and/or (ii) a logical-subspace process fidelity alongside global F (Eq. 9). If the gain shrinks substantially, the abstract wording should be narrowed accordingly.","section":"Sec. III–IV, Table I, Fig. 6"},{"comment":"Sec. II and Appendix C: the open-system model uses on-site σ− and σz jumps that remain form-invariant under RF+RWA. The paper correctly flags this as a phenomenological choice and notes that a secular derivation in the strongly interacting (η=20) blockade regime would produce dressed, frequency-dependent jumps. Because the mitigation mechanism is “shorten exposure while preserving blockade,” a qualitative change in jump structure (e.g., correlated or blockade-conditioned loss) could alter the reported gains. A short robustness check—e.g., a comparison with a simple dressed or correlated dissipator on a reduced ladder, or a clear statement of the regime of validity and how it would fail—would make the central claim much more secure.","section":"Sec. II, Appendix C"},{"comment":"Sec. IV C: optimization maximizes a noiseless cost over a training set, then fidelity under noise is evaluated post hoc. That is a standard and legitimate workflow, but the manuscript should state the training-set size/composition and confirm that the optimized pulse was not implicitly tuned on the same |gg⟩ dissipative trajectories used in Fig. 6. A brief hold-out (unseen logical inputs under noise) would remove any residual concern that the 0.96 figure is overfit to the reported configuration.","section":"Sec. IV B–C"}],"minor_comments":[{"comment":"Eq. (9) vs Eq. (11): the text mixes “root” fidelity F with the trajectory average of the squared overlap. State once, prominently, which quantity is plotted in every figure (including whether error bars are std. err. of the mean over trajectories).","section":"Sec. III A"},{"comment":"Fig. 2–3 and Fig. 6: axis labels give γ in µs−1 while the calibration Ωχ=10 µs−1 is in the text; a secondary axis or explicit γ/Ω would help comparison with device T1 values.","section":"Figs. 2, 3, 6"},{"comment":"The CZ gate is omitted from the optimization study because it is already short; a one-sentence estimate of residual headroom (or a statement that GRAPE was tried and gave negligible further compression) would round out Sec. IV.","section":"Sec. IV"},{"comment":"Typos/notation: “N ´eel” spacing is inconsistent; “RW A” vs “RWA”; arXiv IDs in the reference list that look like future/placeholder numbers should be checked for correctness before publication.","section":"Throughout / References"},{"comment":"Appendix D: max(m)=200 and λ=10−12 are stated; a one-line bond-dimension convergence check for the longest (standard Hadamard) noisy runs would strengthen reproducibility.","section":"Appendix D"}],"recommendation":"major_revision","confidential_remarks":"Fit for a solid specialized quant-ph / quantum-control venue is good; for a very high-impact general journal the evidence base is still narrow (N=2, favorable inputs, phenomenological dissipator, single gate optimized). The self-citation cluster defining the architecture is appropriate but dense; the decoherence-mitigation result is the novel piece and should remain the focus after revision. I do not see integrity issues—data DOI is provided and methods are reconstructible."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The new piece is the open-system characterization plus the mitigation demo. They take their existing globally driven ladder (N=2, 15 qubits), run Lindblad trajectories with MPS/TDVP under local amplitude damping and dephasing, map how ICC shift, Hadamard, and CZ degrade, then show a GRAPE pulse that compresses the Hadamard from ~3000 ns to ~320 ns and keeps F≈0.96 at γ−=0.08 (with 2% frequency disorder) where the rectangular sequence falls to ~0.64. Table I and Fig. 6 match the text. Data are deposited; appendices give the RF+RWA Lindblad invariance, pulse decompositions, and cutoffs (2000 trajectories, bond 200, λ=10−12). That pipeline is clean and reproducible for this class of work.\n\nWhat they do well: they are explicit that dissipation hits every physical qubit, including the Néel/ferro spectators that enforce the blockade, so you cannot just protect a logical subsystem. The CZ being naturally short and more robust is a useful structural observation. Self-citations define the architecture; the decoherence numbers themselves are new forward simulations, not circular.\n\nSoft spots, in proportion. The load-bearing restored-fidelity plot is almost entirely for |ψ_in⟩=|g1 g2⟩, which they correctly call the most robust payload. The only population sweep (Fig. 4) is noiseless and at fixed phase, so it does not stress logical superpositions or |e⟩ under the short drive. Global fidelity also folds in the whole background; a high F can mask logical-subspace error. The dissipator is phenomenological local jumps, not a dressed secular form in the blockade regime—they flag this. None of that sinks the numerical claim inside the model; it does mean the abstract’s “restores high gate fidelities” is a bit ahead of the logical-process evidence.\n\nThis is for people working on global-control superconducting (or Rydberg) architectures who need a concrete, hardware-light lever against T1. Methods people will care about the MPS-GRAPE stack. I would send it to referees: the result is real within its assumptions, the limitations are mostly stated, and the fix (broader logical averaging, process fidelity, clearer waveforms) is ordinary revision, not a rewrite. Worth engaging if you touch this platform.","headline":"Solid open-system numerics on their ladder: GRAPE shortens the global Hadamard ~10× and restores high F under local T1, but the restored-fidelity claim is still mostly for the easy |gg⟩ payload.","tokens_in":23464,"tokens_out":591,"would_cite":true,"duration_ms":13566,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Lx","03.65.Yz","85.25.-j","42.50.Lc"],"model":"grok-4.5","headline":"Shaping the global drive compresses gate times by ~10× and restores high fidelities under decoherence in a ladder processor.","keywords":["global optimal control","decoherence mitigation","superconducting qubits","ladder architecture","GRAPE","tensor networks","quantum trajectories","pseudo-Rydberg blockade"],"falsifier":"Implement the GRAPE-optimized Hadamard on a physical ladder device (or a high-fidelity master-equation simulation with secular dressed jump operators) at γ−≈0.08 µs−1 and check whether measured fidelity remains near 0.96 or collapses toward the standard-pulse value.","tokens_in":23359,"feed_emoji":"⚡","tokens_out":890,"duration_ms":18850,"temperature":0.7,"pith_summary":"In a globally driven superconducting ladder, every physical qubit feels the same noise, including the background qubits that hold the ordered phases around the logical information. The paper shows that amplitude damping and dephasing both degrade information flow and gate fidelity, with relaxation the more damaging channel because it breaks the blockade that keeps the computation inside the well-formed subspace. The authors then show that reshaping the shared control pulses with gradient-ascent optimal control shortens the multi-step Hadamard sequence from roughly 3000 ns to roughly 320 ns. That compression alone keeps the fidelity above 0.96 across the full relaxation range they study, even with 2% frequency disorder present. The result matters because the architecture cannot protect an isolated logical subsystem; the only available lever is the temporal shape of the global drive.","feed_headline":"Global pulses cut gate time 10×, keep fidelity near 0.96","feed_subtitle":"In a ladder where every qubit feels the noise, reshaping the shared drive outruns relaxation.","key_machinery":"MPS-based GRAPE pulse optimization on the full 15-qubit ladder: control amplitudes are bounded and discretized at ~1 ns, gradients are obtained by forward/backward TDVP propagation, and the cost is a noiseless fidelity averaged over training states, yielding a ~320 ns Hadamard that still implements the target rotation.","core_discovery":"Global optimal control, by compressing gate sequences by nearly an order of magnitude in time, restores high one-qubit gate fidelities under decoherence in a globally driven ladder architecture. For the Hadamard under pure relaxation the standard rectangular protocol falls to F≈0.64 at γ−=0.08 µs−1 while the optimized pulse stays near F≈0.96, and the gain is obtained purely by reshaping the shared drives rather than by shielding any subsystem.","pith_inferences":["If the compression factor continues to scale with system size, global-drive architectures may remain competitive with locally addressed processors even when T1 is only tens of microseconds.","The same temporal-shaping idea should transfer to other blockade-based global platforms (Rydberg arrays, spin chains) whose logical subspace is protected by an ordered background.","Non-Markovian or thermal corrections that reintroduce dressed jumps would be the natural next stress test of the mitigation claim."],"forward_implications":["Gate duration, not bare channel rates alone, is the primary lever controlling decoherence accumulation in this architecture.","The same optimized pulses can simultaneously absorb a few-percent static frequency disorder and dynamical relaxation.","CZ remains comparatively robust because its native duration is already ~π/Ω, leaving little room for further compression.","Fault-tolerant pulse design for globally driven ladders can start from the same MPS-GRAPE engine.","Relaxation on spectator qubits that sustain Néel and ferromagnetic order is the dominant error pathway that pulse shortening must outrun."],"fun_headline_variants":["Global drive shaping cuts gates 10×, holds F≈0.96 under noise","Optimized shared pulses outrun decoherence on full ladder","Tensor-network study: global control restores Hadamard to F≈0.96","Reshaping global drives compresses sequences, beats relaxation","Order-of-magnitude faster gates via global optimal control"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The noise is modeled by simple on-site relaxation and dephasing jumps that stay unchanged after moving to the rotating frame; if the true jumps are dressed by the strong interactions, the reported fidelity gains from shorter pulses need not hold.","fun_headline_variants_meta":{"raw":{"variants":["Global drive shaping cuts gates 10×, holds F≈0.96 under noise","Optimized shared pulses outrun decoherence on full ladder","Tensor-network study: global control restores Hadamard to F≈0.96","Reshaping global drives compresses sequences, beats relaxation","Order-of-magnitude faster gates via global optimal control"]},"model":"grok-4.5","effort":"low","cost_usd":0.003902,"raw_usage":{"total_tokens":1160,"prompt_tokens":710,"num_sources_used":0,"completion_tokens":74,"cost_in_usd_ticks":39024000,"prompt_tokens_details":{"text_tokens":710,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":376,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":710,"tokens_out":74,"duration_ms":7421,"temperature":1.0,"reasoning_tokens":376,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T06:04:42.832952+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Implement the GRAPE-optimized Hadamard on a physical ladder device (or a high-fidelity master-equation simulation with secular dressed jump operators) at γ−≈0.08 µs−1 and check whether measured fidelity remains near 0.96 or collapses toward the standard-pulse value.","supporting_citations":[],"review_version":1}