{"id":"e668673e-a3a2-462b-ade1-18c8995b2434","arxiv_id":"2510.20489","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The 3D toric code's optimal thresholds under measurement noise are ≈11% (bit-flip) and ≈2% (phase-flip), the latter inferred through an approximate entropy duality from the 4D random-bond Ising model.","lead":"This paper computes the best-possible error thresholds of the 3D toric quantum error-correcting code when measurements are faulty too: about 11% for bit-flip errors and 2% for phase-flip errors. The numbers set theoretical upper bounds that decoder designers and hardware teams can benchmark against.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The new 2% Z-sector threshold rests on an approximate entropy relation applied to a 2-form gauge model with no independent numerical anchor; a failure of Eq. (19) for 2-form theories would leave the headline claim unsupported.","rationale":"The reader's weakest assumption identifies the same load-bearing risk: the 2% threshold is obtained by applying an approximate, replica-based entropy relation to a 2-form gauge model with no numerics and whose mapping derivation is withheld. My stress-test confirms this is the single point on which the paper's central new claim depends. The 11% X threshold is well-supported by self-duality and existing literature, so the paper should not be rejected outright; however, the 2% number is currently a conjecture supported by analogy. The concrete check a direct Monte Carlo study of Eq. (9) along the Nishimori line would settle whether Eq. (19) transfers to 2-form theories, and would also provide the missing numerical anchor. Because the reader already assigned CONDITIONAL, and the concern reinforces rather than alters that verdict, I recommend UNCHANGED. The mapping derivation should be supplied as part of the conditional acceptance, alongside an uncertainty budget for the 28% input.","tokens_in":25979,"tokens_out":11242,"duration_ms":106760,"concrete_test":"Perform a direct Monte Carlo simulation of the 4D RCGM (Eq. 9) along the Nishimori line e^{-2βJ}=p/(1-p), for system sizes L=4,6,8 (and L=10 if feasible) with periodic boundary conditions, measuring the disorder-averaged Wilson surface [⟨W(S)⟩]. Locate the confinement-deconfinement transition via finite-size scaling of the volume-law vs area-law crossover. If the extrapolated p_c differs from ≈2% by more than the statistical uncertainty (e.g., >0.3% absolute), Eq. (19) is not reliable for 2-form theories and the central Z-threshold claim needs revision. As a complementary analytical check, re-derive Eq. (9) from the chain-complex formalism in SI §I.3 to confirm the disordered cube couplings are the correct representation of the faulty-measurement syndrome.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's genuinely new number, p^{Z,M}_th ≈ 2%, is not computed from the 4D random-cube gauge model (RCGM, Eq. 9) directly. It is inferred via the approximate generalized duality H(p_c)+H(ṕ_c)≈1 (Eq. 19), using the 4D random-bond Ising critical point ṕ_c≈28% from Ref. [55]. Eq. 19 is derived in SI §III.2 through the replica trick and the uncontrolled step 'we expect' w≈w̃ (SI Eq. S35). The authors validate Eq. 19 only on ordinary spin/1-form gauge models (e.g., 3D RBIM/RPGM and self-dual 4D RPGM). The RCGM is a 2-form Z2 gauge theory with a 1-form symmetry and volume-law Wilson-surface order, and the paper explicitly concedes 'there are no existing numerics for this model'. Thus the central Z-threshold claim has no direct numerical support: it depends on transferring an approximate relation to a model class on which it has not been tested. The steepness of H(p) near p=0.02 (dH/dp≈5.6 bits per unit p) amplifies any error in Eq. 19 or in the 28% input. Additionally, the mapping from the faulty-measurement Z-error chain complex (SI §I.3) to Eq. (9) is asserted rather than derived in the main text or SI; if that mapping is not exactly the RCGM, the entropy-relation inference is moot. The 11% X threshold is independent and robust, but the '2%' headline and the 'robustness against measurement errors' claim rest on this unverified chain.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the 3D toric code under phenomenological noise with both qubit Pauli errors and faulty syndrome measurements at equal rate. It derives two effective 4D random lattice gauge models: for bit-flip (X) errors a 4D random-plaquette Z2 gauge model (RPGM), which is self-dual and yields a threshold p^{X,M}_th ≈ 11%; for phase-flip (Z) errors a 4D random 2-form (random-cube) Z2 gauge model (RCGM), Eq. (9). Since the RCGM has no existing numerics, the paper infers its threshold p^{Z,M}_th ≈ 2% from the approximate generalized entropy duality H(p_c)+H(ṕ_c)≈1, Eq. (19), using the known 4D random-bond Ising critical point ṕ_c≈28% (Ref. [55]). The abstract and discussion present these as the optimal phenomenological thresholds of the 3D toric code and conclude that the code is robust against measurement errors.","tokens_in":26465,"tokens_out":2781,"duration_ms":27327,"significance":"If the central claim holds, this is the first determination of optimal thresholds of a 3D topological code under measurement errors, providing a benchmark for the growing effort to realize 3D codes. The derivation of a random 2-form gauge theory from a fault-tolerance setting is an interesting contribution in its own right, and the X-sector threshold (11%) is solid, being fixed by self-duality and consistent with known results. However, the headline Z-sector threshold (2%) rests on an approximate duality relation that is validated only on ordinary spin/1-form gauge models and is applied here to a new 2-form model with no independent numerical anchor. The significance of the paper is therefore conditional on closing this gap, either by a direct derivation of Eq. (19) for 2-form models, by numerical verification of the RCGM transition, or by explicitly reframing the 2% value as a conjecture.","major_comments":[{"comment":"The 2% Z threshold is not computed from the RCGM directly but inferred from the approximate generalized duality H(p_c)+H(ṕ_c)≈1. The derivation of this relation uses the uncontrolled approximation w≈w̃ at SI Eq. (S35). The relation is calibrated on spin/1-form gauge models (3D RBIM/RPGM, 4D self-dual RPGM), but the RCGM is a 2-form theory with different symmetry and order parameter. The paper explicitly states there are no existing numerics for this model. Since dH/dp near p=0.02 is about 5.6 per unit p, even a small error in Eq. (19) or in the input ṕ_c≈28% shifts the inferred threshold substantially. The abstract claims p^{Z,M}_th≈2% as an established optimal threshold; this is not supported without either an explicit derivation of Eq. (19) for 2-form theories or numerical evidence for the RCGM phase transition. At minimum, the claim should be softened to a prediction.","section":"Eq. (19) and SI §III.2, Eq. (S35)"},{"comment":"The mapping from the faulty-measurement Z-error sector to the 4D random-cube gauge model is asserted in the main text (Eq. (9) and surrounding discussion) but not derived. SI §I.3 formulates the chain-complex description of measurement errors, but the explicit step from the 1-chain constraint in the 4D spacetime lattice to the 2-form gauge model with six-body cube interactions is not shown. This mapping is load-bearing: if the effective model is not exactly the RCGM, the entropy-relation inference is moot. The authors should provide the explicit derivation of Eq. (9), or at least a detailed outline in the main text.","section":"Eq. (9) and SI §I.3"},{"comment":"The 2% value depends on the external Monte Carlo result ṕ_c≈28% for the 4D random-bond Ising model (Ref. [55]). The statistical uncertainty of that estimate is not propagated, and the sensitivity of p^{Z,M}_th to variations in ṕ_c is not discussed. Given the steepness of H(p) near p=0.02, this is not a negligible effect. The paper should state the uncertainty in the input and, if possible, provide a range for p^{Z,M}_th.","section":"Table 2 and Ref. [55]"}],"minor_comments":[{"comment":"Reference [47] is titled 'Analog information decoding of bosonic quantum low-density parity-check codes', which appears unrelated to the quoted matching-decoder threshold 0.0126 for the 3D toric code. Please verify this citation and, if correct, clarify the connection.","section":"References"},{"comment":"The text says the RCGM is defined on the dual lattice, but Fig. 4(b) shows the dual lattice with plaquette variables and cube interactions; the relationship between the cube index c* and the original lattice could be made more explicit for readability.","section":"Fig. 4 and Eq. (9)"},{"comment":"The entry for the faulty-Z sector lists p_c≈2% and ṕ_c≈28% without error bars or a note that the former is derived from the latter via the approximate relation. A footnote would help the reader distinguish directly computed values from inferred ones.","section":"Table 2"},{"comment":"The symbol p is used both for the physical error rate and as the probability argument of the Shannon entropy H(p). This is standard, but a brief reminder near Eq. (19) would avoid confusion.","section":"General notation"}],"recommendation":"major_revision","confidential_remarks":"The paper has a useful and solid X-sector result and a novel statistical-mechanical model; however, the central Z-sector threshold is not yet established at the level claimed. The authors should be encouraged to either provide a direct derivation of the generalized entropy relation for 2-form gauge theories, perform small-scale numerical tests of the RCGM (e.g., on the Nishimori line), or explicitly present the 2% value as a duality-based conjecture. With such changes, the paper would be suitable for publication. The citation issue for Ref. [47] should also be corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the one new result, the phase-flip phenomenological threshold around 2%, is not computed from the random-cube gauge model directly. It is inferred via the approximate entropy relation H(p_c)+H(ṕ_c)≈1 applied to an external 4D RBIM critical point. The authors are honest that there are no numerics for their RCGM, but the headline still presents 2% as a settled benchmark. The X-sector 11% is well-known Nishimori self-duality, so the novelty really is the Z-sector claim and the RCGM itself.\n\nWhat the paper does well: it gives a clean statistical-mechanical derivation for the 3D toric code with faulty measurements, identifies the 2-form nature of the Z-error model, and the entropy relation passes calibration on known pairs — H(0.233)+H(0.033)≈0.99 is a nice check. The comparisons with decoder-dependent thresholds (1.26% matching, 9.65% overlapping-window) are useful framing for decoder designers.\n\nThe soft spots matter. The 2% figure rests on three legs: the uncontrolled “we expect” step w≈w̃ in SI Eq. S35; the mapping from the Z-error chain complex to Eq. (9), whose derivation is deferred even in the SI; and the external 4D RBIM critical point ≈28% with no uncertainty propagated. Because H(p) is steep at p=0.02, small errors in any of these inputs change the output noticeably. There is also an internal inconsistency: the introduction says the perfect-measurement Z threshold is 0.11, while the results section and Table 2 say 3.3%. And the discussion claims “exact value” for a relation the paper itself labels approximate. These are fixable, not fatal.\n\nMy overall read: the X-sector threshold is solid and the RCGM mapping is a legitimately interesting contribution, but the 2% Z threshold is a conjecture dressed as a result. It may well be right — the entropy relation has no known counterexample in this family — but it lacks direct numerical or derived support as written.\n\nWho is this for? People benchmarking decoders for 3D codes, and anyone working on higher-form gauge theories from QEC. It deserves a serious referee who can push on the derivation of Eq. (9) and S35. I would not cite the 2% number in my own work yet, but I would follow the RCGM mapping.","headline":"The paper's only genuinely new number, p^Z,M≈2%, is a plausible but unanchored inference through an approximate duality applied to an unsimulated model, while the X-sector result is established; worth peer review, not desk rejection.","tokens_in":26992,"tokens_out":2498,"would_cite":false,"duration_ms":25452,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Pp","05.50.+q","75.10.Nr"],"model":"deepseek-v4-flash","headline":"The 3D toric code keeps about 2% phase-flip and 11% bit-flip thresholds even when syndrome measurements fail at the same rate.","keywords":["3D toric code","fault-tolerant threshold","measurement errors","random 2-form Z2 gauge theory","Nishimori line","generalized duality","statistical-mechanical mapping","topological quantum error correction"],"falsifier":"Run a Monte Carlo simulation of the four-dimensional random-cube Z2 gauge model on the Nishimori line, locate its critical error rate p_c, and check whether H(p_c)+H(0.28) equals 1 within error bars; a substantial deviation would invalidate the generalized duality for two-form models and remove the support for the claimed 2% threshold.","tokens_in":25788,"feed_emoji":"⚛️","tokens_out":5435,"duration_ms":46937,"temperature":0.7,"pith_summary":"The paper asks how well the 3D toric code — a leading code for fault-tolerant non-Clifford gates — performs when syndrome measurements are themselves noisy, not just the qubits. It claims that under equal-rate qubit and measurement errors the code has optimal thresholds of roughly 11% for bit-flip errors and 2% for phase-flip errors, only modest declines from the perfect-measurement values of about 23% and 3.3%. The overall threshold is set by the lower phase-flip value, about 2%. The authors reach these numbers not by simulating the 4D disordered models directly but through a generalized duality that relates both error sectors to known random spin models. If right, realistic measurement noise is not a major obstacle for 3D toric-code-based quantum computing, and decoder designers gain concrete upper-bound benchmarks.","feed_headline":"3D toric code keeps a 2% threshold under faulty readout","feed_subtitle":"Bit-flip tolerance drops from 23% to 11%, phase-flip from 3.3% to 2%, a mild dent from realistic noise.","key_machinery":"The engine is the generalized Kramers-Wannier duality with quenched disorder, expressed along the Nishimori line as the approximate Shannon-entropy relation H(p_c)+H(ṕ_c)≈1, where H is binary entropy and p_c and ṕ_c are the critical points of a model and its dual. This lets the authors infer the phase transition of the difficult new 4D random 2-form Z2 gauge model — spins on plaquettes, six-body random cube couplings, Wilson-surface order parameter — from the known 4D random-bond Ising transition. The self-dual 4D random-plaquette gauge model carries the bit-flip argument, while the random-cube gauge model carries the phase-flip argument.","core_discovery":"The central claim is that the 3D toric code's optimal phenomenological thresholds are p^{X,M}_th ≈ 11% and p^{Z,M}_th ≈ 2% when qubit and measurement errors occur at the same rate. The bit-flip sector maps to a 4D random-plaquette Z2 gauge model, which is self-dual, and self-duality pins the threshold to the Shannon entropy point H ≈ 1/2. The phase-flip sector maps to a new 4D random 2-form (random-cube) Z2 gauge theory with plaquette variables and six-body cube interactions; the paper argues this model is dual to the 4D random-bond Ising model, whose known critical point near 28% yields p^{Z,M}_th ≈ 2% through the approximate relation H(p_c)+H(ṕ_c)≈1. This makes the 3D toric code robust to","pith_inferences":["Editorial inference: if the generalized entropy duality extends beyond the models tested in the paper, the same derivation could yield analytic thresholds for 3D color codes and fracton codes under faulty measurements, likely generating further higher-form gauge models.","Editorial inference: the small phase-flip drop (3.3% to 2%) suggests the Z sector of 3D codes is inherently less measurement-sensitive than 2D surface codes, so biased-noise engineering may improve overall 3D thresholds more than it does in 2D.","Editorial inference: the decisive test is a direct Monte Carlo simulation of the 4D random-cube gauge model along the Nishimori line; if its critical point does not satisfy H(p_c)+H(0.28)≈1 within uncertainty, the 2% figure would need revision, with the error inherited from the unpropagated uncertainty in the 28% input."],"forward_implications":["If correct, the 3D toric code has an overall phenomenological threshold near 2%, set by phase-flip errors, with bit-flip errors tolerated up to about 11%.","The bit-flip threshold exceeds values reported for neural-network (~7%) and single-shot (~3%) decoding approaches and approaches a recent overlapping-window decoder result (~9.65%), indicating practical decoders still have room to improve.","Measurement noise damages the 3D toric code far less than the 2D toric code, whose threshold drops from about 11% to 3.3%.","The mapping produces a new disordered 2-form gauge theory from a quantum error-correcting code, directly connecting fault tolerance with higher-form gauge theory.","Because these are optimal thresholds, they serve as hard upper-bound benchmarks: no decoder can exceed them under the same noise model."],"fun_headline_variants":["3D toric code: 2% phase-flip threshold under noisy readout","3D toric code resists measurement errors: 2% threshold","Optimal thresholds for 3D toric code under faulty measurements","3D toric code robust: bit-flip 11%, phase-flip 2%","Noisy readout barely hurts 3D toric code thresholds"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The 2% phase-flip threshold rests on transferring the approximate entropy relation H(p_c)+H(ṕ_c)≈1, validated on ordinary one-form spin and gauge models, to the new four-dimensional random two-form gauge model — a step the paper flags as approximate, with no explicit dual mapping shown and no existing numerics for the model.","fun_headline_variants_meta":{"raw":{"variants":["3D toric code: 2% phase-flip threshold under noisy readout","3D toric code resists measurement errors: 2% threshold","Optimal thresholds for 3D toric code under faulty measurements","3D toric code robust: bit-flip 11%, phase-flip 2%","Noisy readout barely hurts 3D toric code thresholds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000509,"raw_usage":{"total_tokens":2361,"prompt_tokens":833,"completion_tokens":1528,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":577,"completion_tokens_details":{"reasoning_tokens":1436}},"tokens_in":577,"tokens_out":1528,"duration_ms":16604,"temperature":1.0,"reasoning_tokens":1436,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T08:26:46.834403+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a Monte Carlo simulation of the four-dimensional random-cube Z2 gauge model on the Nishimori line, locate its critical error rate p_c, and check whether H(p_c)+H(0.28) equals 1 within error bars; a substantial deviation would invalidate the generalized duality for two-form models and remove the support for the claimed 2% threshold.","supporting_citations":[],"review_version":1}