{"id":"c1844fe9-3416-454e-ba1c-6e13be0902eb","arxiv_id":"2608.10013","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"A model-based RF-budgeted compiler workflow determines validated multitone frame partitions for frequency-multiplexed superconducting-qubit control, with a 12-qubit BV microwave layer closing in three four-tone frames at 240 ns.","lead":"This paper introduces a simulation workflow that decides which qubit-control tones can safely share a single microwave pulse frame, combining qubit calibration records, a modeled RFSoC electronics chain, and qutrit-level quantum simulations. It reports model-based capacities such as a 12-qubit Bernstein-Vazirani layer closing in three four-tone frames at 240 ns, useful for frequency planning in scaled-up superconducting processors.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Frame-level validation only checks qubits addressed in that frame; off-resonant drive of qubits addressed in later frames is never validated at layer level, so the 240-ns three-frame BV closure may overstate final per-qubit accuracy.","rationale":"I read the paper as a model-based workflow claim rather than a hardware claim, and the authors are careful to label results as simulation diagnostics. The workflow is internally structured, with exact coloring and exact-cover audits, and the main contribution is a reproducible admission procedure. The load-bearing gap is not the use of priors or the absence of T1/T2, which the paper declares; it is that the admissibility predicate is applied per frame and per addressed qubit, whereas the central quantity—a validated layer partition—requires per-qubit error after the full sequence. This is checkable with the authors' own QuTiP machinery: propagate each of the 12 qutrits through the three frames, including all tones of each frame in the Hamiltonian of every qutrit, and compare final per-qubit errors with Table II. If the final errors exceed the budgets, the 'three validated frames' result is not a layer-level closure. This does not invalidate the workflow as a screening tool, but it should be stated as frame-level closure, or the layer-level check should be added. The reader's locality concern is partially right; my concern sharpens it to a specific omission that is inexpensive to fix.","tokens_in":24539,"tokens_out":8728,"duration_ms":91479,"concrete_test":"Simulate the 12-qubit BV layer at T=240 ns by applying the three four-tone frames sequentially to all 12 qutrits, evolving each qutrit under Eq. 20 with all tones of the current frame (including tones intended for other qubits), and compute each qubit's final rotation-angle error, axis-phase error, and maximum transient leakage after the full layer; compare these with the Table II thresholds and with the per-frame worst-case values. If any qubit exceeds the thresholds, the 'three validated frames' claim is not a layer-level closure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The workflow's closure tests are frame-local: Section V-A defines frame-level diagnostics as maxima over addressed qubits, and Section VI-C validates two- to four-qutrit risk patches rather than the full layer. For a partition into K frames, a qubit is addressed in one frame but exposed to the aggregate tones of the other K-1 frames through the same shared line. The pairwise scheduler bounds two-tone false addressing and leakage, but a frame's coherent sum of up to four tones can produce an aggregate spectator drive larger than any pairwise check; this spectator drive is never evaluated because the qubit is not addressed in that frame. Likewise, the qubit's own gate validation starts from ideal input states, not from the state accumulated after earlier frames. A 'validated partition' therefore certifies each frame in isolation, not the full microwave layer. The headline 12-qubit BV three-frame, 240-ns capacity is exactly such a partition, so the claimed layer closure is not established by the reported diagnostics.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a compilation workflow for frequency-multiplexed superconducting-qubit control. For a requested layer of single-qubit rotations, the workflow builds a QID database (transition frequencies, pulse parameters, drive calibration), constructs a QID-level conflict graph using pairwise RF and leakage checks, colors the graph with an exact DSATUR-style search, forms candidate multitone RF-frame descriptors, propagates each aggregate waveform through a behavioral RFSoC source-chain model (Table I), decodes the result into transition-frame drive coefficients, and validates each frame with local QuTiP qutrit-patch simulations against the Table II closure thresholds. The studies range from single-qutrit pulse closure to pairwise coexistence, RF-frame capacity on four synthetic QID maps, and Qiskit-derived BV and QAOA microwave layers. Under the nominal RF budget, the paper reports that a 12-qubit BV -Y90 layer closes in three validated four-tone frames at 240 ns, and QAOA mixer partitions vary with rotation angle and pulse duration. All results are explicitly decoherence-free, model-based simulation diagnostics.","tokens_in":24727,"tokens_out":6697,"duration_ms":69675,"significance":"If the workflow is accepted as a design-time diagnostic, it is a useful and fairly complete framework: it combines a circuit-informed RF-chain model, leakage-aware qutrit dynamics, explicit closure thresholds, and a scheduler with model-in-the-loop no-good constraints. The paper gives credit for reproducibility: deterministic seeds, a stated QuTiP integrator configuration, convergence checks, exact-cover audits for selected partitions, and a 100-seed robustness audit for one headline case. The distinction between screening diagnostics and final closure criteria is carefully drawn, and the authors are transparent that the RF-chain parameters are priors rather than measured hardware specifications. The main quantitative claims, however, are conditional on the frame-local validation scheme, and the abstract's wording that a BV layer 'closes in three validated four-tone frames' overstates what the reported diagnostics establish at the layer level.","major_comments":[{"comment":"The workflow's validation is frame-local, and the advertised layer-level 'closure' is not established. Section V-A defines frame-level diagnostics as maxima over addressed qubits, and the qutrit Hamiltonian in Eq. (20) is evaluated only for addressed qutrits in that frame. A qubit whose gate is scheduled in a later frame is never propagated under the aggregate waveform of an earlier frame; only the pairwise two-tone screens of Sec. IV-A bound its off-resonant drive. A four-tone frame can produce a coherent spectator drive on a later-frame qubit that exceeds the per-pair false-addressing threshold even when every pair passes, and this aggregate spectator drive is never checked because the qubit is not addressed in that frame. The exact-cover audit in Sec. VI-D does not remedy this, because it selects from the same frame-local validated library. Additionally, each gate validation starts from an ideal |0> state (Sec. V-B), not from the state accumulated after earlier frames in the partition. Consequently, the claim that a 12-qubit BV -Y90 layer 'closes in three validated four-tone frames at 240 ns' (Abstract; Fig. 11(a)) is a claim about frame-by-frame validation, not about the full microwave layer. I recommend adding a layer-level spectator check (for each qubit, simulate or bound its dynamics under the aggregate of all frames in the partition in which it is not addressed) or explicitly re-scoping the 'validated partition' language to frame-local validation.","section":"Secs. V-A, V-C, VI-C; Fig. 11(a)"},{"comment":"The headline capacities (e.g., the three-frame 240-ns BV partition and the frame counts in Figs. 8 and 13) are computed only at the nominal values of the RF priors. Table I states that only the δu_eff sweep is exercised and that the stress brackets for ENOB, jitter, S21, compression, and clipping are 'configured but not exercised.' Because the admission decisions depend on these priors through the source-chain model, the reported numbers are a single point in prior space. Since the abstract emphasizes a concrete 240-ns, three-frame result, the paper should either exercise the configured stress brackets for the headline workloads and report how the frame counts change, or explicitly state that no sensitivity of the headline numbers to the unexercised priors is claimed. As written, a reader could reasonably take the 240-ns result as a robust capacity estimate rather than as a nominal-prior-only simulation outcome.","section":"Table I and Secs. VI-C, VI-D"},{"comment":"The notation and the implementation description are in tension. Eq. (11) defines the aggregate waveform from the complex envelopes d_i(t) of Eq. (10), which include the DRAG-like quadrature term, but the text in Sec. III-B states that in the implemented workflow only the in-phase envelope g_i(t) is propagated through the source chain and that the quadrature is applied later at the decoded qutrit-drive stage. This means the headroom, clipping, and crest-factor checks do not see the full drive waveform, which is a modeling choice that could affect the reported capacities. Please align the equations with the implemented path (e.g., define a separate in-phase aggregate s_F^I(t) for the RF-chain propagation) and justify why the quadrature component is exempt from source-chain impairment.","section":"Sec. III-B and Eqs. (10)-(11)"}],"minor_comments":[{"comment":"References [7] and [8] list DOIs '10.1103/xw1s-dt16' and '10.1103/fmqf-w6ht', which do not appear to be valid CrossRef DOIs; please verify or replace them. The manuscript header also contains the placeholder 'Digital Object Identifier 10.1109/TQE.2026.Doi Number'.","section":"References"},{"comment":"The sentence 'All 100 robustness seeds close, corresponding to a Wilson 95% confidence interval [0.963,1.000]' does not state what the seeds randomize (jitter realizations, noise seeds, or something else). Please specify the stochastic variation in the text or caption.","section":"Sec. VI-D, Fig. 14"},{"comment":"The phrase 'Muted RF frames are excluded from the transition-frame and QuTiP counts' uses 'muted' in an unusual way; 'rejected' or 'RF-failed' would be clearer.","section":"Fig. 7(d) and Sec. VI-B"},{"comment":"The sentence 'The QuTiP stage validates local two- to four-qutrit risk patches rather than a fully coupled 16-qutrit Hamiltonian' is an important scope limitation and should also appear in the abstract or conclusion, alongside the already-stated decoherence-free caveat.","section":"Sec. VI-C"}],"recommendation":"major_revision","confidential_remarks":"The frame-local validation gap identified in the major comments is the key issue: the abstract overstates the layer-level closure. The paper's own scope statements (model-based, decoherence-free, local patches) are honest but do not directly acknowledge the cross-frame spectator problem. In addition, the reference list contains DOIs that appear to be placeholder strings; the editor may wish to verify the integrity of references [7], [8], and any other newly appearing entries before publication. The paper's topic fits the journal, and the workflow contribution is potentially valuable; the revision should either close the layer-level gap computationally or explicitly narrow the claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real engineering contribution — a compiler workflow that decides which tones can share an RF frame under source-chain and leakage budgets — and it is honestly scoped as model-based. But the headline \"layer closure\" claim is stronger than the validation supports: each frame is checked only for the qubits it addresses, so aggregate spectator drive from later frames and state accumulation across frames are never evaluated. The 240-ns three-frame BV result is a per-frame validation count, not a validated full-layer trajectory.\n\nWhat's new and good: the QID record abstraction is a sensible compiler-side container for transition frequencies, pulse parameters, and drive calibration; the RFSoC behavioral model is detailed but honestly labeled as priors; the model-in-the-loop no-good constraints and recoloring inside an exact DSATUR search is a nice touch. The paper earns credit for reproducibility hygiene: versions, seeds, convergence checks, a 100-seed robustness audit, exact-cover audits for several partitions, and clear statements that all results are decoherence-free simulation diagnostics. The pairwise leakage-guard calibration (150/60/45/30 MHz) is useful in its own right.\n\nSoft spots, in order: (1) The frame-local validation gap. Section V-A defines diagnostics as maxima over addressed qubits; Section VI-C validates two-to-four-qutrit patches. A qubit addressed in frame 1 is exposed to the aggregate tones of frames 2..K on the same shared line, and its input state after frame 1 is not the ideal state used in its own gate validation. The stress-test concern read the paper correctly; this is a real limitation on the layer-level capacity claims. It is not fatal if the paper is read as a per-frame admission tool, but the abstract and conclusion sell it as layer closure. (2) The RF priors are unmeasured; the paper says so, but the headline numbers depend on them. (3) The reproducibility package is referenced but not linked; no repo or commit hash. Minor: the paper is dense and sometimes hard to navigate, but that's not a correctness issue.\n\nWho it's for: hardware architects and compiler people deciding frequency plans and pulse durations for shared-line control. It deserves a serious referee. The referee should push for either a full-layer validation (even at reduced scale) or a rewritten claim that says \"each frame passes its own closure budget\" rather than \"the layer closes.\" My recommendation: send it to review, with the expectation of a major revision on the validation claim.","headline":"The workflow is a genuine contribution and deserves refereeing, but the headline layer-closure numbers are frame-local validation, not full-layer trajectories.","tokens_in":25317,"tokens_out":3691,"would_cite":true,"duration_ms":35918,"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":"This paper claims that shared-RF multitone control of superconducting qubits can be made a compile-time admission problem, and that under its nominal RF budget a 12-qubit Bernstein–Vazirani microwave layer closes in three validated…","keywords":["superconducting qubits","frequency-multiplexed control","RF-budgeted compilation","qubit-control identity records","RFSoC source-chain modeling","leakage-aware scheduling","crosstalk-aware control","microwave control"],"falsifier":"Run the paper's three-frame, four-tone Bernstein–Vazirani partition on a real shared-line RFSoC transmitter driving 12 frequency-distinct transmons and measure each qubit's $|2\\rangle$ population and rotation angle after the frame; if any qubit's transient leakage exceeds $10^{-3}$ or the layer needs more than three frames to meet the Table II thresholds, the isolated-patch validation is too optimistic. A cheaper variant is a fully coupled multi-qutrit simulation of one candidate frame with direct coupling and energy relaxation added.","tokens_in":24251,"feed_emoji":"📡","tokens_out":8589,"duration_ms":84672,"temperature":0.7,"pith_summary":"Superconducting-qubit control that shares one microwave source across many qubits cannot be planned by assigning carrier frequencies alone; the shared RF budget can corrupt multi-qubit rotations through headroom, quantization, jitter, spurs, compression, crosstalk, and leakage. This paper introduces a compiler workflow that treats the question as an admission problem: for a requested layer of single-qubit gates, it partitions qubits into RF frames only if the full aggregate waveform survives a source-chain model and local qutrit-dynamics validation, and it repartitions when validation fails. The load-bearing quantitative claim is that under the paper's nominal RF budget a 12-qubit Bernstein–Vazirani Hadamard layer, implemented as physical $-Y_{90}$ rotations, closes in three validated four-tone frames at 240 ns, while QAOA mixer layers need between three and six frames depending on rotation angle. The contribution is a reproducible method for deciding how many tones can share one frame and for attributing which constraint, when violated, forces a split. All results are model-based, decoherence-free simulation diagnostics rather than measured hardware fidelities.","feed_headline":"Three shared RF frames close a 12-qubit layer in 240 ns","feed_subtitle":"A compiler decides when tones can safely share one control line by checking RF budget, leakage, and crosstalk before execution.","key_machinery":"The central object is the qubit-control identity (QID) record, a compiler-side entry per qubit holding computational and first-leakage transition frequencies, pulse parameters, and drive-scale calibration, with the RF-chain profile and effective crosstalk matrix supplied as separate inputs. Around it runs an admission loop: pairwise conflict rules build a graph, an exact graph-coloring search proposes a minimum-frame partition, each candidate frame's aggregate waveform is propagated through a behavioral RF system-on-chip (RFSoC) source-chain model, decoded into transition-frame projections, and validated by local qutrit-patch dynamics with a Hamiltonian that includes crosstalk couplings and leakage transitions. The mechanism that carries the argument is the loop itself, where a failed validation adds a no-good constraint and triggers recoloring, so the workflow's output is a validated frame partition rather than a heuristic spacing rule.","core_discovery":"On the paper's own terms, the discovery is that frequency-multiplexed microwave control is a compile-time admission problem rather than a carrier-assignment problem. A candidate multitone frame is admissible only when the aggregate waveform passes every configured RF-budget check and every local qutrit closure test: intended-drive mismatch below $5\\times10^{-3}$, false addressing and leakage-drive projection below $10^{-3}$, rotation-angle and axis-phase error below $1^\\circ$, and leakage-aware fidelity loss, computational-subspace survival loss, and maximum transient leakage below $10^{-3}$. When a candidate frame fails, the failure is returned as a no-good constraint and the layer is recolored, so the output is a validated partition and an attributed failure mechanism rather than an unrestricted simultaneous-control claim. Under the nominal RF-chain priors, the paper reports that a 12-qubit $-Y_{90}$ layer from a Bernstein–Vazirani circuit closes in three four-tone frames at 240 ns, and that a 12-qubit $X_{90}$ layer closes in five frames at 120 ns and three at 240 ns for uniform, jittered, and heavy-tail frequency maps but requires nine and five frames for a clustered map. These numbers are conditional on the configured model budgets and are explicitly not hardware measurements.","pith_inferences":["Editorial extension: if the isolated-patch validation is optimistic, the three-frame Bernstein–Vazirani partition is the number most likely to degrade on hardware, because it packs four tones per frame with aggregate peaks near half full scale; measuring spectator-qubit $|2\\rangle$ population during that exact frame would be the sharpest test.","Editorial extension: the same admission loop could be inverted into a frequency-allocation co-design tool, using validated frame count as a cost function when choosing transmon frequency maps; the paper demonstrates the sensitivity but does not propose the optimization.","Editorial extension: adding energy relaxation and dephasing to the qutrit patches would probably reduce the admissible frame count at longer pulse durations, so the reported capacities should be read as decoherence-free upper bounds on what a dissipative device would tolerate."],"forward_implications":["Longer pulses admit more tones per frame but do not necessarily minimize validated layer time: at 240 ns the 12-qubit $X_{90}$ layer needs three frames, while at 120 ns it needs five frames but reaches the highest time-normalized aggregation of 20 qubits/GHz/$\\mu$s.","Frequency-map clustering is costly: for the 12-qubit $X_{90}$ layer the clustered map requires 9 frames at 120 ns and 5 at 240 ns, versus 5 and 3 for uniform, jittered, and heavy-tail maps, so validated frame count should be a co-objective in frequency planning.","Duration-dependent directed leakage guards, calibrated to 150, 60, 45, and 30 MHz for pulse durations of 80, 120, 160, and 240 ns, are scheduling preconditions but do not replace aggregate-frame validation because multitone leakage can stack coherently.","Algorithm-derived layers can be covered by compact partitions: the 12-qubit Bernstein–Vazirani $-Y_{90}$ layer closes in three four-tone frames at 240 ns, and QAOA mixer layers close in 3, 5, 6, and 5 frames for physical rotation angles of 90°, 120°, 150°, and 180°.","A candidate frame must pass aggregate RF headroom and source-chain checks before qutrit evaluation; per-tone phase scheduling could recover headroom, but any phase-optimized frame must be revalidated through the same model because phase changes coherent false-addressing and leakage contributions."],"supporting_citations":[{"why":"Supplies the leakage-elimination pulse design that underlies the DRAG-like quadrature used in the QID pulse parameters.","marker":"[15]"},{"why":"Establishes that simultaneous gates in frequency-crowded multilevel systems require leakage-aware treatment, motivating the frame-level leakage checks.","marker":"[19]"},{"why":"Supplies the qutrit-dynamics solver used for the local validation patches.","marker":"[26]"},{"why":"Supplies the circuit-construction layer from which the Bernstein–Vazirani and QAOA microwave workloads are extracted.","marker":"[30]"},{"why":"Defines the Bernstein–Vazirani algorithm whose Hadamard layers become the fixed $-Y_{90}$ workload.","marker":"[33]"},{"why":"Defines the QAOA whose mixer layers become the variable-angle $R_x$ workload.","marker":"[34]"},{"why":"Supplies the exact graph-coloring search used to propose minimum-frame partitions from the pairwise conflict graph.","marker":"[35]"},{"why":"Supplies the RF-DAC and clocking specifications that anchor the behavioral RFSoC source-chain model priors.","marker":"[40]"}],"fun_headline_variants":["RF budget, not tone count, decides which qubit frames share a line","Compiler admits frames only after RF, leakage, and crosstalk checks","Three four-tone frames close a 12-qubit layer in 240 ns (simulated)","Qubit control sharing becomes an admission problem at compile time","RF-budgeted frame compilation: multi-tone sharing passes only if closure"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The most fragile premise is that validating each qubit as an isolated three-level patch, with no direct qubit-to-qubit coupling and no energy relaxation, captures what a real shared control line does to all qubits simultaneously.","fun_headline_variants_meta":{"raw":{"variants":["RF budget, not tone count, decides which qubit frames share a line","Compiler admits frames only after RF, leakage, and crosstalk checks","Three four-tone frames close a 12-qubit layer in 240 ns (simulated)","Qubit control sharing becomes an admission problem at compile time","RF-budgeted frame compilation: multi-tone sharing passes only if closure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00042,"raw_usage":{"total_tokens":2271,"prompt_tokens":1165,"completion_tokens":1106,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":781,"completion_tokens_details":{"reasoning_tokens":1005}},"tokens_in":781,"tokens_out":1106,"duration_ms":11057,"temperature":1.0,"reasoning_tokens":1005,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:29:45.033483+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the paper's three-frame, four-tone Bernstein–Vazirani partition on a real shared-line RFSoC transmitter driving 12 frequency-distinct transmons and measure each qubit's $|2\\rangle$ population and rotation angle after the frame; if any qubit's transient leakage exceeds $10^{-3}$ or the layer needs more than three frames to meet the Table II thresholds, the isolated-patch validation is too optimistic. A cheaper variant is a fully coupled multi-qutrit simulation of one candidate frame with direct coupling and energy relaxation added.","supporting_citations":[{"cited_title":"Simultaneous gates in frequency-crowded multilevel systems using fast, robust, analytic control shapes,","cited_arxiv_id":null,"evidence_quote":"Establishes that simultaneous gates in frequency-crowded multilevel systems require leakage-aware treatment, motivating the frame-level leakage checks."},{"cited_title":"Zynq UltraScale+ RFSoC Data Sheet: DC and AC Switching Characteristics,","cited_arxiv_id":null,"evidence_quote":"Supplies the RF-DAC and clocking specifications that anchor the behavioral RFSoC source-chain model priors."}],"review_version":1}