{"id":"96ef01df-7c00-4a31-abe0-2b4b31dfef16","arxiv_id":"2607.20616","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A device model extension that captures neutral-atom-specific features (coordinates, interaction radii, shuttling, zoned sites) enables specialized NA compilers to integrate with universal quantum ecosystems, demonstrated via QDMI.","lead":"Quantum ecosystems like Qiskit use device models built for superconducting chips, which assume qubits never move. This paper redesigns those models so neutral-atom hardware with movable, rearranged atoms and zoned operations can plug in, claiming up to a 100,000x routing-fidelity gain on a 16-qubit circuit.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 100,000x fidelity claim is undermined by applying the shuttling-derived idle-coherence formula (Eq. 1) to the SWAP-based SC scenario without physical justification.","rationale":"The reader's weakest_assumption already identifies the same load-bearing concern: the shuttling-derived coherence model is applied to SWAP routing. I agree. I do not see a separate, more damaging flaw. The device-model extension itself is reasonable, the implementation is public, and the qualitative examples (static coupling map cannot represent dynamic rearrangement; zoned devices are excluded) are convincing. The problem is localized to the quantitative benchmark: Eq. 1's t_idle is physically meaningful only for shuttling, and the paper gives no definition for the SC case. Fixing this could change the 100,000x headline by orders of magnitude. Because the paper's central conceptual contribution stands and the quantitative claim needs revision, the reader's CONDITIONAL verdict remains appropriate. No adjustment.","tokens_in":11589,"tokens_out":12125,"duration_ms":97881,"concrete_test":"Recompute the qft-16 row of Table I with the SC-model coherence term evaluated from actual SWAP time (e.g., t_idle = (number of inserted SWAPs) × 3 × 0.36 µs × n_atoms, or from the scheduled circuit duration) instead of n_atoms·10.9 µs per routing step. If the SC fidelity rises by orders of magnitude, the claimed ×10^5 improvement collapses and should be restated with a corrected model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. 1 (Sec. III-A) defines t_idle = n_atoms·10.9 µs, the time to shuttle one atom two sites at 0.55 µm/µs, and models all atoms as idle during that move. In Sec. V-C the paper states that 'for both settings' the routing-overhead fidelity is computed as the product of operation fidelities and this coherence term. For the SC-model column, routing is done by Qiskit via SWAP insertion. A SWAP is not an idle interval: it is implemented by three CZ gates, each 0.36 µs (≈1.08 µs total), and those CZ errors are already included in the operation-fidelity product. Charging the SC scenario an additional coherence penalty of n_atoms·10.9 µs per routing step—or any t_idle derived from shuttling speed—has no physical basis and double-counts decoherence. The qft 16-qubit row (SC 7.39×10^-7 vs NA 7.83×10^-2, improvement ×1.06×10^5) is the headline result; its order of magnitude depends directly on this asymmetric coherence penalty. Without a definition of how t_idle is accumulated in the SWAP case, the 100,000x factor is not supported. The qualitative claim—that static device models omit NA capabilities—is unaffected.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that universal quantum ecosystems (Qiskit, Cirq, PennyLane, QDMI) use device models shaped by superconducting hardware—static qubit positions and fixed coupling maps—and that these models cannot abstract the computational capabilities of neutral-atom devices, specifically dynamic rearrangement/shuttling and zoned operations. The authors propose targeted extensions to a device model (site coordinates, interaction/blocking radii, lattice description, shuttling operations, zone sites, unit handling), implement the model in QDMI, and evaluate it by comparing Qiskit's SWAP-based routing on a static model with the QMAP neutral-atom compiler using the extended model. They report routing-overhead fidelity improvements by a factor of up to 1.06e5 on a 16-qubit, 600-gate QFT circuit.","tokens_in":11876,"tokens_out":16956,"duration_ms":125481,"significance":"The qualitative contribution is solid and timely: the paper identifies a real abstraction gap, illustrates it with concrete examples (Steane-code state preparation and zoned devices), and proposes a generic, implementable device-model extension. The open-source QDMI implementation is a strength. However, the quantitative evaluation does not currently isolate the effect of the device model from compiler specialization and hardware assumptions, and the fidelity calculation is not fully specified. If the evaluation is made rigorous, the proposed model could be a useful step toward standardizing neutral-atom information in quantum software ecosystems.","major_comments":[{"comment":"The comparison conflates device-model expressiveness with compiler specialization and hardware choice. The SC-model column uses Qiskit's generic SWAP router on a monolithic neutral-atom abstraction, while the NA column uses QMAP, a specialized shuttling-aware compiler, on a zoned device. The reported improvements (up to 1.06e5) therefore include differences in compiler capability and target architecture, not only the effect of the proposed model. To attribute the gain to the device model, the evaluation should control for the compiler (e.g., by ablating the newly added model properties while keeping the rest of the stack fixed) or explicitly identify and discuss these confounds. As written, the central quantitative claim is not supported by this comparison.","section":"Sec. V-C, Table I"},{"comment":"The routing-overhead fidelity calculation is underspecified in the SC setting. Eq. (1) defines t_idle = n_atoms·10.9 µs for one shuttling move, and the text states that this coherence term is used 'for both settings'. No definition is given for how t_idle is accumulated when routing is performed by SWAP insertion. A SWAP is not an idle shuttling interval; its CZ errors are already included in the operation-fidelity product. Charging the SC scenario a shuttling-derived idle-time penalty has no stated physical basis, and the Table I numbers are not reproducible from the text. The coherence term may not be the dominant contributor to the 100,000x factor, but the calculation must be specified for the quantitative claim to be grounded.","section":"Sec. V-C, Eq. (1)"}],"minor_comments":[{"comment":"The table formatting merges the gate count with the SC fidelity value; for example, 'qft 16 6027.39×10−7' appears to combine gates=602 with SC fidelity=7.39×10−7. Please separate the columns clearly.","section":"Table I"},{"comment":"The text says single-qubit gates have 99.97% fidelity in 52 µs; this is likely a unit typo (52 ns?) and should be corrected.","section":"Sec. V-C"},{"comment":"For reproducibility, specify the exact Qiskit version, routing method, and optimization level used for the SC baseline, and the QMAP version/options used for the NA baseline.","section":"Sec. V-C"},{"comment":"Clarify whether 'routing overhead fidelity' includes only inserted routing overhead or the total compiled-circuit fidelity; the current wording is ambiguous.","section":"Sec. V-C"},{"comment":"In the ghz 128 row, the NA model is worse by a factor of 3.72. This is an interesting exception but is not explained beyond a brief mention of long shuttling moves; please add a short discussion.","section":"Sec. V-C, Table I"},{"comment":"The 'idling fidelity' property is introduced but its semantics are not formally defined. A precise definition or an example would help.","section":"Sec. IV-E"}],"recommendation":"major_revision","confidential_remarks":"The qualitative contribution is sound and likely of interest to the quantum software/architecture community. The evaluation, however, relies on the authors' own compiler and interface stack and compares against a generic baseline in a way that makes the headline factor difficult to interpret. I do not question the motivation or the need for the proposed model, but the quantitative claim should be either strengthened with a controlled comparison or toned down. The paper is within the journal's scope; the self-citation pattern is relevant and not inappropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth a serious look: it makes a concrete, implementable proposal to extend generic device models with NA-specific features—site coordinates, interaction/blocking radii, lattice/zone markers, shuttling properties, and zone extents—implemented in QDMI with public code. That is a genuine gap: Qiskit, Cirq, PennyLane, and CUDA-Q assume static coupling maps and cannot natively represent shuttling or zoned operations. The qualitative argument is correct and well-illustrated with the Steane-code example.\n\nWhere I'd push back is the evaluation. The benchmark compares Qiskit's generic SWAP-based router (on a monolithic coupling-map model) to QMAP, a purpose-built NA compiler (using the new model). That conflates the device model's expressiveness with compiler specialization. The 100,000x fidelity improvement is real for this pairing, but it does not isolate the model's contribution. A cleaner test would run the same compiler with and without the extended information, or extend Qiskit's routing to accept shuttling and compare directly. That said, showing that the model enables a specialized compiler to work through a standard interface is the paper's point, so this is a framing weakness, not a fatal one.\n\nThe stress-test concern about Eq. 1—applying the shuttling-derived idle-coherence formula to the SWAP-based scenario—is overstated. The coherence term exp(-t_idle/T_eff) with T_eff = 1.5 s and idle times in the tens of microseconds is extremely close to 1; even charging 10.9 µs per SWAP instead of 1.08 µs changes the product by well under 1%. The 100,000x factor comes from the number of CZ gates (SWAP insertions vs. shuttling moves), not from the coherence penalty. The formula's application is sloppy and should be clarified, but it is not load-bearing.\n\nThe paper is honest about its scope, uses literature hardware constants, and the central qualitative claim is externally grounded. Self-citation is heavy but the cited prior work is relevant. I'd send this to peer review: the core idea and implementation are solid, and the evaluation needs tightening rather than rethinking.","headline":"A well-motivated QDMI extension for neutral-atom features that fills a real gap, but the headline fidelity gain is confounded by comparing different compilers, not just different device models.","tokens_in":12378,"tokens_out":5075,"would_cite":true,"duration_ms":39674,"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":"Universal quantum ecosystems' device models, shaped by superconducting hardware, cannot represent neutral atoms' dynamic rearrangement and zoned operations; a generalized device model closes the gap and improves routing fidelity by up to 10","keywords":["neutral atom quantum computing","device model","qubit routing","atom shuttling","zoned operations","quantum compiler","hardware abstraction","fidelity improvement"],"falsifier":"Recompute the fidelity comparison using measured per-operation error rates—CZ gates, single-qubit gates, trap transfers, and shuttling—rather than the single idle-time coherence formula, and check whether the 16-qubit QFT circuit still shows a ~100,000x ratio. A quicker check: run the same circuit on a real neutral-atom device under both compilation regimes and compare measured output fidelities.","tokens_in":11461,"feed_emoji":"⚛️","tokens_out":7203,"duration_ms":60234,"temperature":0.7,"pith_summary":"The paper tries to establish that the device models used by universal quantum ecosystems—the standardized descriptions of a device's qubits, gates, and connectivity that compilers read—are too narrow for neutral-atom hardware. Because those models were shaped by superconducting chips, they assume fixed qubit positions and static coupling maps, and so can neither express dynamic atom shuttling nor represent zoned devices built around storage, entanglement, and measurement regions. As a result, neutral-atom compilers cannot get the hardware information they need from standard ecosystems, a 'technology lock' that the paper argues blocks integration. The paper's proposed remedy is a more generic device model that adds site coordinates, interaction and blocking radii, shuttling properties, zone metadata, and unit conventions, so that specialized neutral-atom compilers can be plugged in unchanged. Evaluations show that routing-overhead fidelity improves by a factor of up to 100,000 on a 16-qubit, 600-gate circuit when the full neutral-atom capability set is exposed rather than flattened to a superconducting-style coupling map.","feed_headline":"Device model unlocks 100,000x fidelity gain for neutral atoms","feed_subtitle":"Standard quantum software can now expose neutral-atom shuttling and zoned gates instead of forcing SWAP-heavy routing.","key_machinery":"The central object is the device model itself: the machine-readable description of a quantum device's computational capabilities that a compiler consumes. The paper's contribution is a generalized model that adds, per site, spatial coordinates plus interaction and blocking radii (the two radii that determine which atoms can perform a Rydberg gate and which atoms must stay away); per operation, shuttling speed, minimum atom separation, and a zoned flag; zone sites with spatial extents; lattice metadata so clients can reconstruct trap-site layouts; and integer scale factors for lengths and durations. The identity doing the quantitative work is the routing-overhead fidelity estimate exp(-t_idle","core_discovery":"The core claim is that the abstraction gap—not missing compiler algorithms—is what keeps neutral-atom devices and their specialized compilers out of universal quantum ecosystems. The paper demonstrates the failure mode with a Steane-code state-preparation circuit: an interaction graph that requires a connection between non-adjacent atoms forces SWAP insertion (at least 0.15 infidelity for one SWAP) when the device is described only by a static coupling map, whereas a single shuttling move (10.9 µs, infidelity ~6.5×10^-5 under the paper's coherence model) satisfies the same connection. It then proposes concrete device-model extensions—three-dimensional site coordinates, interaction and blocki","pith_inferences":["If this model becomes a common interface, benchmark comparisons between neutral-atom compilers and between hardware vendors could be standardized, making device exchange at computing centers a matter of swapping a driver rather than rewriting compilation passes.","The same abstractions—movable qubits, zone-like regions, radius-defined connectivity—could be used to integrate trapped-ion devices, whose shuttling and zoned operations the paper notes are also missing from current models.","The lattice-reconstruction protocol means a client can recover a device's logical zones entirely from queried properties, which could enable generic placement and routing algorithms that are unaware of the underlying technology."],"forward_implications":["Neutral-atom devices with shuttling and zoned operations can be represented in universal quantum ecosystems instead of being approximated by static coupling maps or excluded entirely.","Specialized neutral-atom compilers can retrieve the hardware information they need through the standardized model, so the ecosystem no longer forces them to bypass the abstraction.","For the benchmarked circuits, routing-overhead fidelity improves by factors from about 1.1x to 100,000x, with the largest gains on wide circuits where SWAP insertion would otherwise be catastrophic.","Zoned devices, whose entanglement zones make SWAP-based routing infeasible, become targetable by any compiler that understands the extended model.","The device-model concepts generalize to other ecosystems and other reconfigurable hardware, since the extensions are expressed as generic properties rather than neutral-atom-specific code."],"fun_headline_variants":["Neutral atom device model ends static-coupling lockout","Shuttling beats SWAPs: new model folds neutral atoms in","100,000x fidelity gain from a device model redesign","Quantum ecosystems finally see neutral atom capabilities","Redesigned device model fits neutral atoms' dynamic moves"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The quantitative headline rests on treating both routing strategies as losing fidelity mainly through coherence decay, with idle time equal to the number of atoms times 10.9 microseconds and effective coherence time 1.5 seconds; if SWAP-based routing actually loses fidelity mainly through gate errors, the claimed 100,000x improvement would shrink.","fun_headline_variants_meta":{"raw":{"variants":["Neutral atom device model ends static-coupling lockout","Shuttling beats SWAPs: new model folds neutral atoms in","100,000x fidelity gain from a device model redesign","Quantum ecosystems finally see neutral atom capabilities","Redesigned device model fits neutral atoms' dynamic moves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000175,"raw_usage":{"total_tokens":1138,"prompt_tokens":775,"completion_tokens":363,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":283}},"tokens_in":519,"tokens_out":363,"duration_ms":4515,"temperature":1.0,"reasoning_tokens":283,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T09:48:15.420805+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the fidelity comparison using measured per-operation error rates—CZ gates, single-qubit gates, trap transfers, and shuttling—rather than the single idle-time coherence formula, and check whether the 16-qubit QFT circuit still shows a ~100,000x ratio. A quicker check: run the same circuit on a real neutral-atom device under both compilation regimes and compare measured output fidelities.","supporting_citations":[],"review_version":1}