{"id":"473844f4-97fc-4077-9eb7-401532305ada","arxiv_id":"2411.12813","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"The paper reports that reusing ancilla qubits for both X and Z stabilizers in alternating sub-rounds cuts rotated surface code qubit overhead by about 25%, with simulations suggesting comparable performance at higher distances.","lead":"This paper proposes reusing the same ancilla qubits for both X and Z stabilizer measurements in rotated surface codes, alternating the two measurement types in sub-rounds to cut ancilla count in half. A smart generalist might read it to see whether a simple time-multiplexing trick can meaningfully reduce the qubit overhead of quantum error correction.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 25% qubit saving depends on an unshown connectivity layout: a single ancilla must measure both an X- and a Z-stabilizer, but the paper never specifies how one ancilla can reach both plaquettes without routing overhead.","rationale":"I agree with the reader: the arithmetic and the d+2 crossover are fine conditional on the shared-ancilla layout. The most load-bearing assumption is not a numerical parameter but the physical realizability of the reuse step. It is the hinge of the entire construction: the 25% saving, the d+2 crossover, and the k-bound all follow from halving the ancilla count. If the ancilla cannot be placed or coupled so that it measures both stabilizer types, those formulas describe a different code than the one simulated. I also see the k-bound coefficient error (the paper gives (2√3-1)d rather than (2/√3-1)d) and the unshown extrapolation to d≥13, but both are secondary: the latter affects only the strength of the 'better logical error rate' claim, while the connectivity issue affects the existence of the scheme. Absence of a circuit is not by itself proof of impossibility, but for a proposal whose central quantitative claim is qubit overhead, specifying the layout is a necessary part of the argument. Since the central claim is unsupported at that load-bearing point, the REJECT verdict is appropriate.","tokens_in":9377,"tokens_out":8244,"duration_ms":87903,"concrete_test":"Give the explicit STIM circuit and planar connectivity graph for distance 3, with one ancilla per X/Z pair, nearest-neighbor CNOTs only, and no extra qubits beyond 9 data + 4 ancillas. Then verify that each data qubit participates in one X and one Z syndrome per round and that no single ancilla fault creates a weight-2 data hook error. If this circuit cannot be written down, the 25% overhead claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is in Sec. III-A: the modified code cuts ancillas from d^2-1 to roughly (d^2-1)/2 by using the same ancillas for X-stabilizers in one sub-round and Z-stabilizers in the next. This requires that one ancilla can couple to all data qubits of both an X-type stabilizer and a Z-type stabilizer. In the rotated surface code those stabilizers live on distinct plaquettes (plus boundary terms), and the standard layout places X-ancillas and Z-ancillas at disjoint positions. The paper gives no circuit, connectivity graph, or routing plan showing how an ancilla reaches the second stabilizer's data qubits; Fig. 1(e) only labels the ancillas 'common.' If reuse is implemented by SWAP-based movement or by long-range couplings, the extra routing qubits or coupling hardware must be counted. Without that, the ~25% total-qubit saving is unverified. The STIM comparison in Sec. IV cannot rescue this: no circuits are included, and the simulation assumes a valid layout exists.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a modified rotated surface code in which the same ancilla qubits are reused for X- and Z-stabilizer measurements by splitting each error-correction round into two sub-rounds. It claims this reduces total qubit overhead by about 25%, and that for distances d ≥ 13 a distance d+2 modified code uses the same or fewer qubits than a distance d original code while achieving a lower logical error rate. The paper provides qubit-count formulas, STIM simulations for distances 3, 5, 7, and 9, and extrapolated logical error rates for larger distances.","tokens_in":9573,"tokens_out":8306,"duration_ms":73771,"significance":"If the ancilla-reuse scheme were physically realizable without additional overhead, a 25% reduction in qubit count for surface codes would be a useful contribution. The qubit-count arithmetic (e.g., the expressions for Q_total and ΔQ) is correct and clearly presented, and the idea of time-multiplexing ancillas is worth exploring. However, the central physical feasibility claim is not established, the simulation comparison appears to under-count errors in the modified scheme, and the key high-distance conclusions are extrapolations without a described method. The paper ships no circuits or code, so the STIM results are not reproducible as reported.","major_comments":[{"comment":"The entire 25% qubit-overhead reduction rests on the assumption that a single physical ancilla can measure both an X-stabilizer and a Z-stabilizer in alternating sub-rounds. However, the paper does not provide a circuit-level implementation, a connectivity graph, or a routing plan showing how the same ancilla couples to the data qubits of two different stabilizer types. In the rotated surface-code layout described in Sec. II, X- and Z-stabilizers occupy disjoint plaquettes and each ancilla is dedicated to a single plaquette; reusing it for the other stabilizer type would require either long-range couplings or SWAP-induced movement, whose qubit and time overhead must be counted against the claimed saving. Without that, the claimed reduction is unverified.","section":"Sec. III-A, Fig. 1(e)"},{"comment":"The noise model is asymmetric between the two codes. The text states that depolarizing error is applied 'before every round' as well as 'after every Clifford gate' in the original code, but for the modified code it only specifies application 'before every second sub-round.' If the modified code is not given the same per-gate depolarizing noise on all gates in both sub-rounds, its error rate is underestimated, and the comparison in Fig. 3 and Table II becomes unfair. The authors should provide the exact noise-injection schedules for both codes and ensure that the per-unit-time physical error rate is identical.","section":"Sec. IV-B"},{"comment":"The logical error rates in Table I for original distances up to d=53 are not simulated; the simulations in Fig. 3 cover only d=3,5,7,9. The paper states that Fig. 4 'projects' the rates to higher distances but does not describe the extrapolation model, its assumptions, or its uncertainty. The central claim that a d+2 modified code has a lower logical error rate than a d original code at the same qubit budget for d≥13 depends entirely on these unvalidated extrapolations and is not a direct simulation result.","section":"Sec. IV-C, Table I, Fig. 4"},{"comment":"The paper's own performance data show that the modified code has mean relative logical error ratios that are all greater than 1, ranging from 1.056 (readout/reset) to 1.435 (gate), i.e., 5% to 44% higher logical error rates at the same distance. The text dismisses these as having 'low practical difference' because the absolute rates are small, but for QEC resource estimation the multiplicative penalty is directly relevant to the number of physical qubits needed to reach a target logical error rate. The assertion that the modified code offers 'similar error correction' should be quantified against a target logical error rate and physical error rate rather than asserted.","section":"Table II"}],"minor_comments":[{"comment":"The example for d=3 says that directly implementing a distance-5 original surface code 'would require 32 qubits instead of 37,' but the paper's own formula Q_original = 2d^2 - 1 gives 49 qubits for d=5; the correct saving compared to 37 qubits is 12 qubits, not 5.","section":"Sec. III-B, Case B"},{"comment":"The sentence 'we used the formula (distance × 3) + 1to determine the number of rounds' has a missing space before 'to,' and the choice of this formula is not justified.","section":"Sec. IV-A"},{"comment":"The sentence 'This technique can be applied broadly across various QEC codes, we focus on rotated surface codes only' is a comma splice; it should be broken into two sentences.","section":"Abstract"},{"comment":"The STIM circuits used for the simulations are not included in the paper or supplementary material; given that the paper's comparisons are the main evidence, the circuits should be made available for reproducibility.","section":"General"}],"recommendation":"reject","confidential_remarks":"To the editor: The manuscript's central claim of a ~25% qubit reduction is not backed by a physical layout, which is a fundamental gap. The simulation comparison appears to under-count errors in the modified scheme, and the key high-distance conclusions rest on unexplained extrapolations. In its current form the paper would not be publishable; substantial additional work is required to demonstrate feasibility and provide fair, reproducible benchmarks."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's core trick—time-multiplexing X and Z stabilizer measurements with shared ancillas—is a natural resource-saving idea, and the qubit-counting arithmetic is straightforward and correct. The 25% total-qubit reduction is real if the layout exists, and the crossover at d=13 for d+2 modified vs d original is a neat observation. That part deserves credit.\n\nThe problems start when you ask how this is actually implemented. The paper never shows a circuit or connectivity graph where one ancilla couples to both an X-stabilizer's data qubits and a Z-stabilizer's data qubits. In a rotated surface code those stabilizers live on different plaquettes, and the standard layout puts X and Z ancillas at disjoint positions. Without a concrete layout or a routing plan, the central 25% saving is unverified. The stress-test note is right: this is load-bearing, and the STIM simulations cannot rescue it because no circuits are included.\n\nThe simulation comparison is also suspect. The depolarizing noise model applies errors before every second sub-round in the modified code, which looks like it undercounts errors relative to the original, where errors are applied each round. The paper's own Table II shows the modified code has logical error rates 5–44% higher across error types. The abstract's \"similar or better\" claim only survives via an undocumented extrapolation to d+2, with no simulations at d≥13.\n\nThere is also a clear mathematical error in the k-bound. Solving Q_modified,d+k ≤ Q_original,d gives k ≤ (2/√3 − 1)d ≈ 0.1547d, not (2√3 − 1)d ≈ 2.464d. The claimed bound is off by more than a factor of ten, and the table's k=4 for d=27 satisfies the correct bound but does not validate the stated one.\n\nSo the paper is a mix: a sensible resource-accounting idea with correct counting in the simple cases, but the performance claim is not supported by the evidence, the layout is missing, and the k-bound is wrong. The flaws are specific and correctable in principle, but as written the central promise does not hold up.\n\nWho is this for? Researchers thinking about QEC resource trade-offs might find the time-multiplexing framing worth discussing, and the d+2 crossover is a useful sanity check. But the paper needs a full rewrite, with a real circuit layout, a fair noise comparison, and a corrected bound.\n\nIf I were the editor, I would send it to peer review, not because it is ready, but because the underlying idea is not crank-ish and the errors are concrete enough that a referee could guide the authors toward a testable version. Desk rejection feels slightly harsh for a paper with a clear and correct qubit-counting core, even if the main claim currently fails.","headline":"A clean resource-counting idea undermined by a missing connectivity layout, a likely unfair noise model, and a wrong inequality solution.","tokens_in":10113,"tokens_out":2972,"would_cite":false,"duration_ms":30863,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Pp"],"model":"deepseek-v4-flash","headline":"Reusing one ancilla set for both X- and Z-stabilizer measurements cuts rotated surface-code qubit overhead by about 25%.","keywords":["quantum error correction","surface code","ancilla reuse","syndrome measurement","qubit overhead","logical error rate","stabilizer circuits"],"falsifier":"Try to construct an explicit stabilizer measurement circuit for a distance-3 rotated surface code where each ancilla couples to the data qubits of both an X- and a Z-stabilizer using only nearest-neighbor connectivity; if this requires swaps or extra routing qubits that exceed the saved ancillas, the 25% saving disappears. Alternatively, simulate the modified code with noise that includes slow mid-circuit resets and time-correlated errors, and check whether the logical error rate penalty exceeds the threshold advantage claimed.","tokens_in":9113,"feed_emoji":"⚛️","tokens_out":5292,"duration_ms":49261,"temperature":0.7,"pith_summary":"This paper tries to reduce the qubit overhead of rotated surface codes by reusing the same ancilla qubits for both X- and Z-stabilizer measurements. The trick is to split each error-correction round into two sub-rounds, measuring X-stabilizers first and Z-stabilizers second on the same ancillas after a reset. That halves the number of ancilla qubits and cuts total qubit count by roughly 25%. The authors argue this preserves error-correcting ability: logical error rates stay close to the original code, and for distances d ≥ 13 a distance d+2 modified code uses the same or fewer qubits than a distance d original code while achieving a lower logical error rate.","feed_headline":"Recycled ancillas cut surface-code qubit overhead by 25%","feed_subtitle":"Halving ancilla qubits by measuring X- and Z-stabilizers on the same qubits, one after the other.","key_machinery":"Ancilla recycling by alternating sub-rounds: each full round becomes two half-rounds, the first measuring only X-stabilizers and the second only Z-stabilizers on the same physical ancillas, with a reset in between. This replaces the standard simultaneous X/Z measurement that needs dedicated ancillas per stabilizer. The counting identity Q_modified(d) = $d^{2}$ + ($d^{2}$−1)/2 versus Q_original(d) = $2d^{2}$−1 drives the claimed 25% total-qubit saving, and the cross-over condition Q_modified(d+2) ≤ Q_original(d) gives d ≥ 13.","core_discovery":"Reusing ancilla qubits across X- and Z-stabilizer measurements via time-division multiplexing reduces the ancilla count from $d^{2}$−1 to about ($d^{2}$−1)/2, lowering the total qubit count from $2d^{2}$−1 to about ($3d^{2}$−1)/2. The paper claims the modified code maintains nearly identical thresholds (within about 2% across four error models) and logical error rates that are only slightly higher than the original, with the relative penalty depending on error type. It further claims that for d ≥ 13, upgrading to a modified code of distance d+2 requires the same or fewer qubits than the original distance-d code while delivering a lower logical error rate, so the resource saving converts into better error suppression.","pith_inferences":["The temporal separation of X and Z measurements likely doubles the syndrome-extraction time per round, which could affect error rates under time-dependent noise; the paper's noise model does not directly capture this delay.","The central saving depends on a single ancilla physically coupling to the data qubits of both an X-stabilizer and a Z-stabilizer; without a concrete circuit layout or connectivity graph, a hardware implementation might need extra routing or swap layers that erode the 25% gain.","For small distances (d < 13), the logical error rate penalty is relatively larger, so the resource saving may not justify the performance loss unless qubit count is the dominant constraint.","Combining this ancilla-recycling scheme with other overhead-reduction techniques such as flag qubits or multiplexed readout could push total qubit counts even lower, though the paper does not explore that combination."],"forward_implications":["For any rotated surface code distance d, the ancilla count is halved and the total qubit count drops by approximately 25%.","At distances d ≥ 13, a modified code of distance d+2 fits within the same qubit budget as the original distance-d code and yields a lower logical error rate.","The error-correction threshold is essentially preserved, with a mean difference of about 1.86% across the error models tested.","The approach is presented as generalizable to other quantum error-correcting codes that use rounds and stabilizer circuits, not just surface codes.","Under a fixed qubit budget, the code distance can be extended by up to roughly k ≤ (2√3 − 1)d while staying within the original number of qubits."],"supporting_citations":[{"why":"Introduces the topological surface code and the error-correction framework the paper builds on.","marker":"[4]"},{"why":"Provides the rotated surface code construction and its noise behavior, which the modified code directly alters.","marker":"[5]"},{"why":"Establishes practical surface code designs and qubit overhead estimates for rotated and unrotated variants.","marker":"[6]"},{"why":"Supplies the stabilizer-circuit simulator used to obtain all logical error rates for original and modified codes.","marker":"[7]"},{"why":"Provides the tableau-based simulation method underlying the simulator, enabling fast stabilizer circuit evaluation.","marker":"[8]"},{"why":"Implements the minimum-weight perfect matching algorithm used to decode syndrome measurements.","marker":"[11]"},{"why":"Provides the Python decoding package used to compute logical error rates from the simulator output.","marker":"[12]"}],"fun_headline_variants":["Ancilla reuse halves stabilizer qubits in surface codes","Time-shared ancillas slash surface-code qubit count","One ancilla set does X and Z, cutting overhead 25%","Recycled ancillas: same error correction, fewer qubits","Surface codes need 25% fewer qubits with shared ancillas"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"One physical ancilla can be reset and then couple to the data qubits of both an X-stabilizer and a Z-stabilizer without additional overhead, yet the paper does not provide a circuit layout or connectivity graph showing this is actually realizable.","fun_headline_variants_meta":{"raw":{"variants":["Ancilla reuse halves stabilizer qubits in surface codes","Time-shared ancillas slash surface-code qubit count","One ancilla set does X and Z, cutting overhead 25%","Recycled ancillas: same error correction, fewer qubits","Surface codes need 25% fewer qubits with shared ancillas"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000209,"raw_usage":{"total_tokens":1416,"prompt_tokens":964,"completion_tokens":452,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":363}},"tokens_in":580,"tokens_out":452,"duration_ms":4868,"temperature":1.0,"reasoning_tokens":363,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:10:56.161884+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Try to construct an explicit stabilizer measurement circuit for a distance-3 rotated surface code where each ancilla couples to the data qubits of both an X- and a Z-stabilizer using only nearest-neighbor connectivity; if this requires swaps or extra routing qubits that exceed the saved ancillas, the 25% saving disappears. Alternatively, simulate the modified code with noise that includes slow mid-circuit resets and time-correlated errors, and check whether the logical error rate penalty exceeds the threshold advantage claimed.","supporting_citations":[{"cited_title":"Pymatching: A python package for decoding quantum codes with minimum-weight perfect matching,","cited_arxiv_id":null,"evidence_quote":"Provides the Python decoding package used to compute logical error rates from the simulator output."},{"cited_title":"Topological quantum memory,","cited_arxiv_id":null,"evidence_quote":"Introduces the topological surface code and the error-correction framework the paper builds on."},{"cited_title":"Low-distance surface codes under realistic quantum noise,","cited_arxiv_id":null,"evidence_quote":"Provides the rotated surface code construction and its noise behavior, which the modified code directly alters."},{"cited_title":"Surface codes: Towards practical large-scale quantum computation,","cited_arxiv_id":null,"evidence_quote":"Establishes practical surface code designs and qubit overhead estimates for rotated and unrotated variants."},{"cited_title":"Stim: a fast stabilizer circuit simulator,","cited_arxiv_id":null,"evidence_quote":"Supplies the stabilizer-circuit simulator used to obtain all logical error rates for original and modified codes."},{"cited_title":"Blossom v: a new implementation of a minimum cost perfect matching algorithm,","cited_arxiv_id":null,"evidence_quote":"Implements the minimum-weight perfect matching algorithm used to decode syndrome measurements."}],"review_version":1}