{"id":"cdc85701-ec70-4674-b3c4-b262417314fd","arxiv_id":"2607.27094","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.5,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"A 65-ns native CCZ gate at 99.39(5)% fidelity is demonstrated on three fluxonium qubits with a microwave-driven transmon coupler, matching what would require ~99.94% CZ gates in a standard decomposition.","lead":"Researchers built a three-qubit superconducting chip that runs a native CCZ gate (Toffoli-equivalent) in 65 ns at 99.39% fidelity using one microwave pulse on a shared coupler. This could cut the multi-gate overhead that currently slows algorithms on superconducting processors while keeping control simple and parasitic interactions low.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The manuscript is a complete, self-contained experimental report. Device Hamiltonian parameters, state-dependent coupler frequencies, calibration primitives, XEB protocol, coherence-limit derivation, and SI1000 decomposition simulation are all present and internally consistent. The reader correctly identified the only material residual uncertainty (XEB local-noise assumption). That assumption is already stress-tested in the text by Table III and by agreement with the analytic reference model; it does not rise to a load-bearing flaw that would move the verdict off ACCEPT. No stronger concern (hidden coherent error, unfair decomposition accounting, or scalability contradiction) survives a good-faith reading of the full text and supplements. Therefore the verdict and confidence stay as the reader stated.","tokens_in":16570,"tokens_out":496,"duration_ms":72973,"concrete_test":"Re-fit the interleaved XEB decay of the 65-ns gate after injecting a controlled non-local noise term (e.g., a residual ZZ or correlated dephasing channel at the level of the measured static ZZ ~10 kHz) into the reference model of Appendix I; if the extracted FCCZ shifts by more than the reported 0.05% uncertainty, the local-noise isolation is insufficient.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central experimental claim (65-ns native CCZ at 99.39(5)% XEB fidelity, single-pulse calibration, competitive with a nearest-neighbour decomposition that would need ~99.94% CZ gates) is supported by the device parameters, calibration sequences, XEB curves, coherence-limit estimate, and decomposition simulation in the manuscript and appendices. The softest point remains exactly the one the reader flagged: isolation of the three-qubit error via interleaved XEB with single-qubit references under a local-noise assumption (Appendix I, Eqs. I1–I2, Table III). The authors supply the natural checks (simultaneous vs individual single-qubit fidelities agree; reference decay matches the analytic local-noise model), and no internal inconsistency or missing control appears strong enough to overturn the headline number or the architectural argument. Coupler participation during the 2π pulse is a secondary interpretive caveat for the precise slope of the blue coherence band (Eq. H1 omits the transmon), but it does not undermine the measured fidelity or the decomposition comparison.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript reports the experimental realization of a native three-qubit CCZ gate (locally equivalent to Toffoli) on a fluxonium–transmon–fluxonium processor unit. A single microwave 2π pulse applied to a shared transmon coupler selectively addresses the coupler transition conditioned on |111⟩, producing a 65-ns gate with interleaved-XEB fidelity 99.39(5)%. Calibration uses two independent measurements (Rabi-like amplitude and Ramsey-type conditional phase). The authors argue that matching this fidelity via a nearest-neighbour eight-CZ decomposition would require CZ fidelities of ~99.94% (SI1000 circuit-level noise model), that performance is coherence-limited, and that the unit extends to 2D layouts with suppressed parasitics while supporting multiplexed phase control on |000⟩ and |111⟩.","tokens_in":16776,"tokens_out":1353,"duration_ms":49287,"significance":"If the result holds, it is a concrete step toward treating multi-qubit gates as hardware primitives rather than expensive decompositions. The combination of (i) single-pulse control with a short, transparent calibration, (ii) coherence-limited ~99.4% fidelity at a duration comparable to two-qubit gates, and (iii) an explicit scaling argument that preserves the parasitic-suppression mechanisms of linear fluxonium–transmon architectures is stronger than prior superconducting native three-qubit demonstrations that either required multi-drive calibration or suffered large parasitics. The multiplexed |000⟩/|111⟩ control and the quantitative decomposition comparison (Fig. 3c, Appendix J) are useful, falsifiable contributions. Device parameters, spectroscopy, calibration sequences, XEB circuits, and supplementary derivations (Appendices A–J) are reported in enough detail for independent assessment.","major_comments":[{"comment":"Appendix H, Eq. (H1): the coherence-limit band in Fig. 3a is computed from fluxonium T1 and Tφ only. During the resonant 2π pulse the coupler is appreciably excited on the target computational subspace (T1,T ≈ 6 μs, Table I), so coupler relaxation and dephasing contribute to the process error. Because the abstract and main text claim “coherence-limited performance,” the estimate should either include a weighted coupler term (with the duty cycle set by the 2π trajectory and the 1/8 computational weight of |111⟩) or explicitly justify why that contribution is negligible relative to the blue band. A revised curve (or a short numerical estimate) is needed to keep the claim load-bearing.","section":"Appendix H, Eq. (H1); Fig. 3a"},{"comment":"Appendix I, Eqs. (I1)–(I2) and Table III: the reported 99.39(5)% rests on interleaved XEB with single-qubit references under a local-noise assumption (pref = 1 − 16/21 Σ ei). The simultaneous-vs-individual single-qubit comparison and the analytic reference-decay check are the right controls and are partially supplied, but the main text should state more clearly (i) the circuit depths used, (ii) that the extracted pCCZ is stable when the reference model is replaced by the measured simultaneous single-qubit decay, and (iii) any residual coherent error budget (off-resonant coupler transitions, residual ZZ) that is not captured by the depolarizing fit. Without that, the isolation of the three-qubit error—and therefore the 99.94% equivalent-CZ claim—remains the softest point of the central result.","section":"Appendix I; Fig. 3b; HIGH-FIDELITY NATIVE CCZ GATE"}],"minor_comments":[{"comment":"Fig. 1b and Table I: the state-dependent coupler detunings are central to spectral selectivity; adding the fitted χi (or the non-additive shifts) next to the level diagram would help the reader connect Eq. (1) to the measured 78/130 MHz isolations without flipping to the SI.","section":"Fig. 1b; Table I"},{"comment":"Fig. 3c: the gray “equivalent CCZ from CZ decomposition” curve should state in the caption that single-qubit errors are p/10, idling errors are included, and T gates are ideal (Appendix J), so the 99.94% figure is not misread as a pure two-qubit threshold.","section":"Fig. 3c; Appendix J"},{"comment":"Main text (MULTIPLEXED CONTROL): the 80-ns multiplexed vs sequential comparison is limited by AWG power; a sentence on whether the fidelity gap is fully explained by duration (vs residual cross-talk between the two drive tones) would tighten the claim that multiplexed control adds no coherent error.","section":"MULTIPLEXED CONTROL; Fig. 4"},{"comment":"Typos / notation: “HIGH-FIDELITY NA TIVE” and “SUPPLEMENT AR Y INFORMA TION” have stray spaces; “CCZ” vs “CCZ” (overline) should be introduced once in a display equation so the two operations remain visually distinct in print.","section":"Section headings; throughout"},{"comment":"Appendix A–B: the two-mode fluxonium harmonic mode is declared “far detuned and omitted”; quoting its frequency (or a lower bound) would reassure readers that it does not participate in the driven dynamics.","section":"Appendix A; Appendix B"}],"recommendation":"minor_revision","confidential_remarks":"The work is a solid experimental advance and is appropriate in scope for a high-impact quantum-information venue. The two major points are clarifications of load-bearing error accounting, not conceptual flaws; I expect them to be addressable in revision without new devices. I did not find evidence of circular fitting or undisclosed free parameters beyond ordinary Hamiltonian extraction and pulse calibration. Novelty relative to the authors’ own prior coupler-activated CZ/CCZ theory papers is adequate because this is the first high-fidelity experimental native CCZ in this architecture with the scaling and multiplexed-control arguments."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The headline result is real: a 65-ns native CCZ on three fluxoniums through a microwave-driven transmon coupler at 99.39(5)% XEB fidelity, single control pulse, two-step calibration, and performance that tracks the coherence limit. Same fidelity via nearest-neighbor Toffoli decomposition would need CZ gates around 99.94%. That comparison is the right one to make, and they run it cleanly with SI1000.\n\nWhat is new is the combination, not the ingredients. Microwave-activated fluxonium–transmon gates and other three-qubit demos already exist (Liu, Kim, Warren, Roy, their own prior theory). Here they get coherence-limited native CCZ with simple control, multiplexed independent phase on |000> and |111>, and a concrete argument that the three-qubit unit cell extends the linear fluxonium parasitic-suppression story into 2D without extra coupler complexity. Device table, spectroscopy, calibration sequences, XEB circuits, and appendices A–J are complete enough to audit.\n\nSoft spots are modest. The weakest link is exactly what the reader flagged: interleaved XEB with single-qubit references under a local-noise assumption. They do the natural checks (simultaneous vs individual single-qubit fidelities agree; reference decay matches the analytic model), so I would not discount the 99.39% figure, but it is still model-dependent isolation of the three-qubit error. The blue coherence band also omits coupler participation during the 2π pulse; that slightly softens the “coherence-limited” slope claim without touching the measured fidelity or the decomposition comparison. Novelty is incremental relative to the broader multi-qubit-gate literature; significance is high inside superconducting hardware.\n\nMath, data, and citations look solid. No circular fitting of the headline number. Free parameters are the usual device energies, pulse amp/freq, and the p sweep in the decomposition sim.\n\nThis is for people building or benchmarking superconducting processors, especially fluxonium and native multi-qubit gate work. Bring it to reading group if that is your crowd. I would cite it when discussing native CCZ or fluxonium coupler architectures. Send it to peer review; it deserves referee time.","headline":"Solid experimental native CCZ at 99.39% with single-pulse control; the number and the scaling story both hold up.","tokens_in":17544,"tokens_out":549,"would_cite":true,"duration_ms":16102,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A single microwave pulse drives a native 65-ns CCZ gate at 99.39% fidelity on three fluxonium qubits, matching what eight near-perfect CZ gates would need.","keywords":["native multi-qubit gates","CCZ gate","Toffoli","fluxonium","transmon coupler","microwave-driven gates","superconducting qubits","cross-entropy benchmarking"],"falsifier":"An independent full process tomography or a Clifford-based three-qubit randomized benchmark on the same device that returns a process fidelity substantially below the reported XEB number, or a multi-unit two-dimensional chip in which parasitic ZZ or coupler crosstalk collapses the spectral selectivity needed for the single-pulse CCZ.","tokens_in":17458,"feed_emoji":"⚛️","tokens_out":1109,"duration_ms":35877,"temperature":0.7,"pith_summary":"Quantum algorithms lean on multi-qubit operations such as the Toffoli gate, but superconducting hardware usually builds them from many two-qubit gates, stacking error and time. This paper shows a three-qubit processor unit in which a controlled-controlled-phase (CCZ) gate—locally equivalent to Toffoli—is native: one microwave pulse on a shared transmon coupler selectively flips the coupler when all three fluxonium qubits are in a chosen computational state, returning a conditional phase of π. The gate lasts 65 ns, reaches 99.39(5)% fidelity, and is limited by the qubits’ own coherence rather than by multi-drive calibration overhead. Reaching the same fidelity by decomposing into nearest-neighbour CZ gates would demand roughly 99.94% two-qubit fidelity. The same coupling layout is argued to tile into two-dimensional arrays while keeping unwanted qubit–qubit interactions small, so native multi-qubit gates need not sacrifice the simplicity and scalability of ordinary two-qubit architectures.","feed_headline":"Native CCZ gate hits 99.39% fidelity in 65 ns","feed_subtitle":"One microwave pulse on fluxonium qubits matches what eight near-perfect CZ gates would need","key_machinery":"State-dependent coupler frequency: strong fluxonium–transmon capacitive coupling shifts the transmon 0–1 frequency by tens to hundreds of MHz according to the joint computational state of the three qubits, so a single near-resonant 2π microwave pulse selectively completes a Rabi cycle only for the target basis state and returns a controlled-controlled phase.","core_discovery":"The authors experimentally realize a native CCZ operation, locally equivalent to a Toffoli gate, lasting 65 ns with fidelity 99.39(5)% in a three-fluxonium unit coupled by one microwave-driven transmon. A single 2π pulse on the coupler transition conditioned on the |111⟩ (or |000⟩) state imparts the conditional phase; calibration uses only amplitude and frequency sweeps and yields coherence-limited performance. The same fidelity via conventional eight-CZ decomposition on linear connectivity would require CZ gates of about 99.94% fidelity. Multiplexed drives further allow independent simultaneous phase control on |000⟩ and |111⟩, and the unit is claimed to extend to two-dimensional layouts wi","pith_inferences":["If the coherence-limited single-pulse picture holds at scale, Toffoli-heavy workloads (search, factoring, chemistry, measurement-free error correction) could cut entangling depth by roughly the eight-CZ factor without waiting for order-of-magnitude better two-qubit gates.","The same spectral-selectivity mechanism naturally supports continuous families of three-qubit phase gates, not only discrete CCZ, which the multiplexed π/2 demonstration already hints at.","Whether this approach outruns native multi-qubit gates on ions or atoms will turn on whether 2D fluxonium layouts keep the reported isolation when many couplers and islands share a chip."],"forward_implications":["Native Toffoli-class gates become competitive with decompositions once two-qubit fidelities sit near 99.94%, without needing still higher two-qubit performance.","Single-pulse, two-parameter calibration removes the multi-drive coherent-error tax that has limited other superconducting three-qubit schemes.","The three-qubit coupler element lets each fluxonium island touch at most one transmon even in 2D, preserving the low-parasitic mechanism of linear fluxonium–transmon arrays.","Multiplexed coupler tones support simultaneous independent phase gates on distinct computational states and, by analogy, parallel native three-qubit gates on adjacent couplers.","Further gains in fluxonium T1 and T2 should translate nearly one-to-one into higher native CCZ fidelity, since the gate already sits at the coherence limit."],"fun_headline_variants":["65-ns native CCZ hits 99.39% on fluxoniums with one pulse","Single microwave pulse yields 99.39% CCZ in 65 ns","Native CCZ at 99.39% fidelity matches eight 99.94% CZs","Fluxonium CCZ gate: 65 ns, 99.39%, coherence-limited","Microwave-driven coupler enables 99.39% native CCZ"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The quoted 99.39% fidelity correctly isolates the three-qubit gate error, which rests on interleaved cross-entropy benchmarking whose single-qubit reference decay assumes local noise dominates and that residual coherent or non-local errors are negligible.","fun_headline_variants_meta":{"raw":{"variants":["65-ns native CCZ hits 99.39% on fluxoniums with one pulse","Single microwave pulse yields 99.39% CCZ in 65 ns","Native CCZ at 99.39% fidelity matches eight 99.94% CZs","Fluxonium CCZ gate: 65 ns, 99.39%, coherence-limited","Microwave-driven coupler enables 99.39% native CCZ"]},"model":"grok-4.5","effort":"low","cost_usd":0.004639,"raw_usage":{"total_tokens":1371,"prompt_tokens":788,"num_sources_used":0,"completion_tokens":93,"cost_in_usd_ticks":46388000,"prompt_tokens_details":{"text_tokens":788,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":490,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":788,"tokens_out":93,"duration_ms":8203,"temperature":1.0,"reasoning_tokens":490,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T11:17:00.267245+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"An independent full process tomography or a Clifford-based three-qubit randomized benchmark on the same device that returns a process fidelity substantially below the reported XEB number, or a multi-unit two-dimensional chip in which parasitic ZZ or coupler crosstalk collapses the spectral selectivity needed for the single-pulse CCZ.","supporting_citations":[],"review_version":1}