{"id":"b0ac2684-0995-4470-a2a7-9f32af41e8c4","arxiv_id":"2504.17509","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The 15-to-1 magic state distillation protocol can be made deterministic and measurement-free with a coherent feedback network, suppressing noise as O(p^2) instead of O(p^3).","lead":"Magic state distillation is usually done by measuring and discarding runs that fail. This paper shows a version of the 15-to-1 protocol that instead applies coherent error corrections, so no measurements or post-selection are needed, at the cost of weaker noise suppression per round.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Measurement-free claim rests on fault-tolerant non-Clifford multi-controlled gates or an undefined coherently-controlled-reset primitive; cost and measurement-free status are unverified.","rationale":"The reader's verdict is CONDITIONAL and identifies essentially the same uncertainty: the correction network is load-bearing and its measurement-free, fault-tolerant realization is not demonstrated. My pass sharpens this concern in two ways. First, the multi-qubit-controlled gates in Fig. 3 are not Clifford, so they cannot be treated as free in the standard MSD cost model, and their fault-tolerant implementation may require the very measurements or magic-state resources the protocol claims to avoid. Second, the Fig. 4 alternative relies on a 'coherently-controlled-reset' primitive that is neither defined nor referenced; a reset is non-unitary, and common implementations reintroduce measurement or ancilla discard. These issues do not invalidate the 15-to-1 result under the stated ideal-gate assumptions, and the numerical evidence for O(p^2) scaling is convincing. But they do substantiate the existing CONDITIONAL verdict: the headline no-measurement advantage should not be accepted until a concrete fault-tolerant construction of the CFN primitives is supplied with resource counts. I therefore recommend no change to the reader's verdict.","tokens_in":17814,"tokens_out":10022,"duration_ms":108166,"concrete_test":"Compile the explicit CFN of Fig. 3 and the reset-based CFN of Fig. 4 into the measurement-free Bacon-Shor gate set of Refs. [17,59] with a full resource count, expressing every multi-controlled gate in an explicit fault-tolerant basis. If the compilation contains any mid-circuit measurement, or if it consumes more injected non-Clifford resource states than the distillation round produces, then the title claim 'without measurements and post-selection' is not supported by the proposed implementation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is credible at the abstracted gate level: under the twirled noise model, the CFN gives O(p^2) single-round suppression and the stabilizer/statevector simulations reproduce the expected multi-round exponential scaling. The load-bearing gap is the 'without measurements' assertion itself. In the section 'Coherent feedback network', the paper says the CFN can be applied to logical qubits using fault-tolerant multi-qubit-controlled gates, citing Refs. [17,59], and Fig. 4 instead uses a 'coherently-controlled-reset' (R) primitive. These are the only bridges from a unitary circuit diagram to a fault-tolerant, measurement-free implementation. The multi-controlled Z gates used in Fig. 3 are non-Clifford for two or more controls; standard fault-tolerant implementations of such gates consume magic states or use lattice-surgery measurements, which would reintroduce exactly the operations the protocol promises to avoid, or would make the CFN as expensive as the measurement baseline. The reset primitive in Fig. 4 is not defined; a physical reset is non-unitary, and if it is implemented by measurement or by swap-and-discard of an ancilla, the protocol is not measurement-free. The paper does not provide a cost analysis of these primitives, nor does it show that Refs. [17,59] cover the specific control structures required. Its own wording ('seems convenient', 'may be desirable') is hedged, so the practical no-measurement advantage is conditional on an unverified hardware primitive.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript presents a deterministic, measurement-free version of the 15-to-1 magic state distillation protocol. The standard syndrome measurement and post-selection are replaced by a unitary decoding circuit E15^† of the [[15,1,3]] code followed by a coherent feedback network (CFN) that applies syndrome-dependent corrections to the output magic state. The author derives a single-round output error rate p_L = 105p^2 + O(p^3) for twirled Z noise, reports statevector and stabilizer simulations confirming this scaling, and shows that repeated rounds give exponential suppression for input error rates below about 1%. The claimed trade-off is a reduction from O(p^3) to O(p^2) per round in exchange for eliminating measurements and post-selection, with the advantage that distillation becomes deterministic and synchronous with logical clock cycles.","tokens_in":18083,"tokens_out":9119,"duration_ms":96775,"significance":"If the implementation gap is closed, this is a valuable contribution to magic state distillation: it shows that the syndrome-measurement step of a standard 15-to-1 protocol can be replaced by coherent feedback, making distillation deterministic and potentially compatible with architectures where mid-circuit measurements are slow or disruptive. The paper's strengths include explicit checkable circuits (Fig. 3), a combinatorial count of 105 uncorrectable weight-2 Z errors, and numerical verification by independent statevector and stabilizer methods whose agreement supports the simplified stabilizer model. The analytic prediction is not fitted to data, so the comparison between theory and simulation is not circular. The main caveat is that the practical 'without measurements' advantage is conditional on an unverified fault-tolerant implementation of the CFN gates or of the reset primitive used in Fig. 4.","major_comments":[{"comment":"The title and abstract claim distillation 'without measurements', but the only bridges to a fault-tolerant implementation are citations to Refs. [17,59] and an undefined 'coherently-controlled-reset' operation in Fig. 4. The multi-qubit-controlled gates in the CFN are not Clifford for two or more controls, so they cannot be assumed noiseless under the paper's assumption that only logical Clifford gates are free; a cost analysis or an explicit fault-tolerant, measurement-free construction of these gates is missing. Please either provide the Kraus-operator/channel description of the R operation and a resource comparison with the measurement-based baseline, or revise the no-measurement claim to state precisely the hardware assumptions under which it holds.","section":"Coherent feedback network / Fig. 4"},{"comment":"Table I and the CFN gate sequence are described as 'manually crafted' and found by Pauli propagation, but no formal proof or exhaustive machine check is provided that the circuit in Fig. 3 realizes every row of Table I and no spurious corrections. Because the leading-order coefficient 105p^2 depends on exactly which single-qubit errors are corrected, an incomplete syndrome table would change the central quantitative claim. I request an automated exhaustive verification of the table and the CFN circuit (for example, Clifford-tableau propagation of all single-qubit Pauli errors), or an explicit algebraic proof that the circuit implements exactly the listed feedback conditions.","section":"Appendix A / Table I"},{"comment":"The resource comparison that would justify the scheme's practical advantage is absent. The paper states that the CFN may be desirable if it can be executed faster and more reliably than measurements, but no estimate is given for the qubit count, gate count, or time overhead of the CFN relative to a measurement-based 15-to-1 protocol, and the effect of noise in the non-Clifford CFN gates on the output error rate is not modeled. Without such an analysis, the practical significance of replacing O(p^3) with O(p^2) suppression per round is not established, even though the abstract-level scaling claim itself is credible.","section":"Measurement-free distillation circuit analysis"}],"minor_comments":[{"comment":"There is a typo in the coherent-noise paragraph: 'finize-size QEC' should be 'finite-size QEC'.","section":"Coherent noise discussion"},{"comment":"Footnote 1 writes t = floor(d - 1/2); the floor should be applied to (d-1)/2, i.e. floor((d-1)/2).","section":"Footnote 1"},{"comment":"The term 'class A' is used without definition; please state the defining property from Ref. [41] or replace it with an explicit description.","section":"Introduction"},{"comment":"Figure 3 is dense and difficult to read at print size; larger fonts for qubit labels and clearer highlighting of the CFN controls would improve checkability.","section":"Fig. 3"},{"comment":"The sentence that multi-qubit-controlled-X gates can be left out for the noise model of Eq. (1) is slightly confusing in relation to Fig. 4, which also removes CNOT gates; please clarify that the two simplifications apply independently.","section":"Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope, and the central scaling claim is plausible and supported by simulations. The main risk is that the headline 'without measurements' claim currently outruns the presented evidence, since the CFN implementation relies on non-Clifford controlled gates or an undefined reset primitive. If the author can supply the missing implementation details or appropriately qualify the claim, I would be supportive of publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, quick take on Heußen's measurement-free 15-to-1 MSD paper. This is a real construction, not vaporware: the coherent feedback network for the 15-to-1 protocol is explicit, the syndrome table is checkable, and the O(p^2) per-round suppression (from the 105 weight-2 Z errors) is confirmed by both statevector and stabilizer simulations. The author is honest that the design was manual and that the general claim for any acceptance-rate-1 protocol is an outlook. That part holds up.\n\nThe soft spot is the 'without measurements' claim. The CFN on logical qubits requires fault-tolerant multi-qubit-controlled gates (non-Clifford for two or more controls) or a 'coherently-controlled-reset' primitive that is never defined. The cited FT constructions for Bacon-Shor codes may work, but the paper doesn't show they are measurement-free, and the reset operation is non-unitary unless implemented via measurement or ancilla discard—which would reintroduce the very operations the protocol promises to avoid. The author's own hedged wording ('seems convenient', 'may be desirable') is a giveaway. This is not a fatal flaw in the abstract circuit analysis, but it is load-bearing for the title claim. If the CFN costs as much as measurements, the advantage evaporates.\n\nI also note the paper doesn't provide a cost comparison against standard MSD with measurements, and it assumes noise-free logical Clifford gates as usual. Those are standard assumptions in the MSD literature, so I don't hold them against it.\n\nOverall: this is a worthwhile contribution for the MSD and measurement-free QEC community. It deserves serious peer review, not desk rejection. The referee should push for a concrete FT implementation of the CFN or an explicit bound on its cost, and for a definition of the reset primitive. I'd bring it to reading group, and I'd cite it for the explicit construction, though I wouldn't rely on the no-measurement advantage without seeing the FT implementation.","headline":"A genuine, simulation-backed coherent feedback construction for 15-to-1 MSD, but the 'without measurements' claim depends on unverified fault-tolerant multi-controlled gates or an undefined reset primitive.","tokens_in":18584,"tokens_out":3448,"would_cite":true,"duration_ms":30084,"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":"The paper claims that the 15-to-1 magic state distillation protocol can be made deterministic and fully measurement-free by replacing syndrome measurement and post-selection with a coherent feedback network, at the cost of reducing…","keywords":["magic state distillation","measurement-free quantum error correction","coherent feedback network","15-to-1 protocol","[[15,1,3]] code","post-selection","fault-tolerant quantum computation","transversal T gate"],"falsifier":"Simulate one round of the noisy circuit under the twirled noise model and check the leading behavior: if the output error rate does not follow $105p^2 + O(p^3)$ for small $p$, for instance if it shows a $p^3$ floor from the feedback gates themselves, then the deterministic correction is not achieving the claimed suppression. A hardware test would count the total error contribution of the correction network implemented fault-tolerantly and compare the end-to-end output fidelity against measurement-based 15-to-1 distillation at the same physical error rate.","tokens_in":17582,"feed_emoji":"🔮","tokens_out":5664,"duration_ms":52142,"temperature":0.7,"pith_summary":"This paper claims that the standard 15-to-1 magic state distillation protocol can be run deterministically, with no individual qubit measurements and no post-selection, by replacing the syndrome measurement and classically conditioned correction with a coherent feedback network built from unitary encoding and decoding circuits. Magic state distillation is the standard way to produce the high-fidelity non-Clifford resource states needed for fault-tolerant universal quantum computing, so removing the measurement bottleneck makes the subroutine available on platforms where fast mid-circuit measurement and feed-forward are slow or unavailable. The cost is that one round suppresses noise from $\\mathcal{O}(p)$ to $\\mathcal{O}(p^2)$ rather than $\\mathcal{O}(p^3)$, but repeated rounds still suppress noise exponentially below a threshold near $p \\approx 1\\%$. If correct, this gives a measurement-free, synchronized distillation routine whose runtime is fixed in advance, and the same construction applies to any distillation protocol that succeeds with probability 1 when noise-free.","feed_headline":"Distill magic states without measurements or post-selection","feed_subtitle":"A coherent feedback network replaces syndrome measurement, trading cubic per-round noise suppression for quadratic.","key_machinery":"The load-bearing object is the coherent feedback network (CFN): a sequence of multi-qubit-controlled gates whose controls are wired to the syndrome qubits emerging from the unitary decoding circuit and whose target is the output message qubit. Pauli propagation rules are used to compile a syndrome look-up table for every single-qubit Pauli error, and the CFN is then crafted so that the message qubit is flipped exactly for syndromes that would otherwise corrupt it, and left alone for syndromes that are harmless or impossible. This converts the classical feed-forward step of magic state distillation into a purely coherent operation, which is what eliminates measurement and post-selection. In a simplified variant, the multi-qubit-controlled gates are replaced by coherently-controlled-reset operations.","core_discovery":"The paper's central claim is that rejection is not necessary: a [[15,1,3]] code whose decoding circuit leaves syndrome information on 14 ancillary qubits can, instead of post-selecting on the trivial syndrome, feed that syndrome coherently into a network of multi-qubit-controlled gates that flip the message qubit exactly when a harmful error has occurred. Because every single-qubit Pauli error is either corrected or harmless, the output state is accepted every round; the leading uncorrectable errors are the weight-2 Z errors, giving a per-round output error rate $105p^2 + O(p^3)$ under twirled noise. The paper shows analytically and by simulation that repeated rounds give exponential suppression, with three rounds taking $p=10^{-3}$ input noise to about $10^{-10}$ output noise. The author presents this as a deterministic alternative to textbook measurement-based magic state distillation, at the price of a lower suppression order per round.","pith_inferences":["If the coherent feedback network's multi-qubit-controlled gates are implemented by measurement-based gadgets on a given architecture, the advertised advantage is lost; the fair comparison is end-to-end physical resource cost, not just the absence of explicit measurements.","A fully measurement-free pipeline (quantum error correction plus distillation) may accumulate coherent errors that would otherwise be reset by mid-circuit measurements; the paper's own suspicion is that in-situ randomized compiling would then be needed, which is a testable design requirement.","The same unitary-decoding-plus-correction idea could be applied to entanglement distillation, where deterministic acceptance would remove similar post-selection bottlenecks.","The manual syndrome-table compilation could be automated and extended to larger codes or other magic state protocols, which would likely be necessary before the method competes with optimized measurement-based factories."],"forward_implications":["Deterministic rounds make distillation time a priori fixed, so magic state production can be synchronized with logical clock cycles.","Platforms without fast mid-circuit measurement and feed-forward can still run magic state distillation, using only unitary gates and reset operations.","Noise suppression per round drops from $\\mathcal{O}(p^3)$ to $\\mathcal{O}(p^2)$, but concatenating rounds suppresses errors exponentially below threshold $p \\lesssim 1\\%$.","The construction transfers to any magic state distillation protocol with acceptance rate 1 in the noiseless limit, not just the 15-to-1 scheme.","It removes the overhead of waiting for accepted rounds and re-initializing failed rounds during routing of magic states."],"supporting_citations":[{"why":"Defines the 15-to-1 distillation protocol and the twirled noise model that the paper adapts.","marker":"[7]"},{"why":"Provide the [[15,1,3]] Reed-Muller and gauge color code with a transversal T gate used as the distillation code.","marker":"[55, 56]"},{"why":"Supply the measurement-free correction approaches that the coherent feedback network applies to distillation.","marker":"[14, 15]"},{"why":"Give fault-tolerant constructions of multi-qubit-controlled gates on Bacon-Shor codes, used to argue the correction network is realizable.","marker":"[17, 59]"},{"why":"Provides the simulation tooling behind the numerical results.","marker":"[60]"},{"why":"Shows how unitary encoding and decoding circuits can be constructed systematically for CSS codes.","marker":"[54]"}],"fun_headline_variants":["Coherent feedback makes magic state distillation deterministic","No measurement, no post-selection: magic states via feedback","Deterministic magic distillation with coherent feedback","Drop measurement, use feedback for magic state distillation","Magic states without measurement: feedback does the work"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The scheme is measurement-free only if the special multi-qubit-controlled gates used to apply corrections can themselves be built fault-tolerantly without measurements; the paper points to existing constructions but does not show they are measurement-free or practical on any specific machine.","fun_headline_variants_meta":{"raw":{"variants":["Coherent feedback makes magic state distillation deterministic","No measurement, no post-selection: magic states via feedback","Deterministic magic distillation with coherent feedback","Drop measurement, use feedback for magic state distillation","Magic states without measurement: feedback does the work"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000204,"raw_usage":{"total_tokens":1404,"prompt_tokens":971,"completion_tokens":433,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":587,"completion_tokens_details":{"reasoning_tokens":361}},"tokens_in":587,"tokens_out":433,"duration_ms":4511,"temperature":1.0,"reasoning_tokens":361,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:39:20.414970+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate one round of the noisy circuit under the twirled noise model and check the leading behavior: if the output error rate does not follow $105p^2 + O(p^3)$ for small $p$, for instance if it shows a $p^3$ floor from the feedback gates themselves, then the deterministic correction is not achieving the claimed suppression. A hardware test would count the total error contribution of the correction network implemented fault-tolerantly and compare the end-to-end output fidelity against measurement-based 15-to-1 distillation at the same physical error rate.","supporting_citations":[{"cited_title":"Bravyi and A","cited_arxiv_id":null,"evidence_quote":"Defines the 15-to-1 distillation protocol and the twirled noise model that the paper adapts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the simulation tooling behind the numerical results."},{"cited_title":"Pogorelov, et al., Compact ion-trap quantum computing demonstrator, PRX Quantum 2, 020343 (2021)","cited_arxiv_id":null,"evidence_quote":"Shows how unitary encoding and decoding circuits can be constructed systematically for CSS codes."}],"review_version":1}