{"id":"58fe7169-a05d-4a40-9dc7-db657b6c4eb7","arxiv_id":"2506.13619","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Blocklet concatenation yields fusion-based quantum computing protocols with constant-sized resource states, erasure thresholds up to 19.1%, and footprint per logical qubit scaling better than surface codes.","lead":"This paper introduces a new way to build error-correcting quantum computing protocols for photonic machines, called blocklet concatenation, which reuses the same small resource states at every level of the code. The resulting fusion networks show higher tolerance to lost photons than standard surface-code approaches while using fewer resources per logical qubit as the code distance grows.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table 1's footprint scaling rests entirely on the unproven distance-scaling conjecture of Appendix A; if the true distance grows more slowly than c·d^L, the claimed O(d)-to-O(d^1.77) advantages over surface codes are optimistic.","rationale":"I read the paper in good faith as a construction of fault-tolerant fusion networks via concatenation, supported by erasure-only Monte Carlo simulations. The strongest claims are the thresholds and, especially, the footprint scalings in Table 1. The thresholds are measured crossing points and are internally consistent with the simulations, though they lack error bars. The footprint scalings, however, are not measured: they are derived from the distance-scaling conjecture of Appendix A, which the author explicitly flags as an open problem. This is the same weakest assumption identified by the reader. It is genuinely load-bearing because every row's footprint exponent is a function of the assumed distance growth: for [5,1,3], exponent log(5)/log(3) with prefactor 5/9; for [6,4,2], exponent log(1.5)/log(2). If the true distance grows less favorably, these exponents understate the resources needed for a target distance, and the central advantage over surface codes is weakened or lost. The subthreshold fits are helpful evidence but cannot rule out lower-weight error strings that only appear at larger L. The proposed check computes the exact minimum distance at L=3, the first level where the product-code prefactor matters, and would either falsify the conjecture or substantially increase confidence in it. Because this is an addressable open problem rather than a demonstrated contradiction, the reader's CONDITIONAL verdict remains appropriate; no change in verdict is needed.","tokens_in":22423,"tokens_out":18310,"duration_ms":184353,"concrete_test":"Construct the full parity-check matrix of the L=3 protocols for the [7,1,3] and [5,1,3] families, including all fundamental, product, and encoded checks, and solve the minimum-distance problem for the resulting classical code exactly with an ILP or SAT solver: minimize the Hamming weight of a nonzero binary vector that satisfies all checks and has odd overlap with a logical operator. If the optimum is below c·d^3 (21 for [7,1,3], 15 for [5,1,3]), the Appendix A conjecture is false and the Table 1 footprints are optimistic; if the optimum equals or exceeds c·d^3, the conjecture survives its first nontrivial concatenation level and the footprint scaling is on substantially firmer ground.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claims that blocklet protocols outperform surface codes are the footprint-per-logical-qubit scalings in Table 1. These scalings are not measured directly; they are computed from the Appendix A conjecture that an [n,k,d]^L protocol has distance c·d^L. For example, the [5,1,3] row's 0.47·d^1.46 footprint follows from n/k=5, d=3, and c=5/9, and the [6,4,2] row's d^0.58 follows from n/k=1.5 and c=1. The only support offered is the construction of weight-dprod·d^(L-2) error strings plus subthreshold Monte Carlo fits up to L=4 or 5. The author explicitly states in Appendix A and in the Conclusion that a proof of the conjecture remains an open problem, and Table 1 itself carries a footnote saying the footprint scaling is based on this conjecture. If the true distance grows more slowly than d^L, then reaching a target distance requires more concatenation steps, and the footprint exponents in Table 1 would be replaced by larger values. The threshold numbers themselves come from finite-size crossings and would not be directly invalidated, but the 'low-overhead' conclusion, which is a primary basis for the claim that these protocols are promising replacements for surface codes, is only as secure as this conjecture.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces 'blocklet concatenation,' a construction for fault-tolerant quantum protocols based on repeatedly concatenating a stabilizer code with itself via its own syndrome readout circuit, interpreted as an encoded Bell pair (blocklet). The construction yields families of fusion-based quantum computing (FBQC) networks with constant-sized resource states, and the paper presents specific families based on [4,2,2], [5,1,3], [6,4,2], and [7,1,3] codes, with Monte Carlo erasure-threshold simulations, logical-operation prescriptions, a hierarchical decoder, and a photonic interleaving implementation sketch. The headline results are erasure thresholds up to 19.1% and footprint-per-logical-qubit scalings O(d) to O(d^1.77) that would improve on surface-code fusion networks.","tokens_in":22773,"tokens_out":6675,"duration_ms":65193,"significance":"If the conjectured distance scaling holds, this is an important conceptual advance: it demonstrates a systematic way to obtain high erasure thresholds with constant-size resource states and sub-quadratic footprint scaling, breaking the surface-code paradigm. The work is carefully presented, the ZX/Pauli-web description is rigorous, and the Monte Carlo methodology with an optimal erasure decoder is standard and appropriate. The author is transparent about the main caveat (the unproven distance-scaling conjecture) and about the decoder being a proof of principle. However, because the advertised advantage over surface codes rests on that conjecture, the result is currently conditional.","major_comments":[{"comment":"The footprint-per-logical-qubit scalings in Table 1 and the abstract (O(d), O(d^1.46), O(d^0.58)) are derived from the unproven distance-scaling conjecture of Appendix A, namely that the code distance of an [n,k,d]^L protocol equals d_prod·d^(L-2). The manuscript explicitly acknowledges this in the Table 1 footnote and in the Conclusion ('proving (or disproving) the distance scaling conjecture also remains an open problem'), yet the central claim that blocklet protocols are 'promising candidates to replace surface codes' rests directly on these scalings. The numerical support in Fig. 25 (subthreshold fits) is consistent with the conjecture but does not establish it; if the true distance grows more slowly, all three footprint exponents would increase and the advantage over the d^2 surface-code footprint could disappear. The paper should either provide a proof or a rigorous lower bound, or clearly condition the overhead claims on the conjecture in the abstract and conclusion.","section":"Appendix A, Table 1"},{"comment":"The erasure thresholds in Table 1 are reported as the physical error rate at which the logical error-rate curves of the two largest simulated protocols intersect, without statistical uncertainties or finite-size scaling analysis. Given that the headline comparison is a threshold of 13.8% versus 12.7% for the surface code, the absence of error bars makes it impossible to assess whether the difference is significant. I recommend reporting confidence intervals from the Monte Carlo sampling and, preferably, a scaling-collapse or fit-based threshold extraction to confirm convergence of the crossing point with system size.","section":"Sec. 2.2, Fig. 8"}],"minor_comments":[{"comment":"The notation 'dp' is used in the caption of Fig. 8 and in Sec. 2.2 (e.g., 'dp = 16') before being formally defined; please define dp = c·d^L at first use in Sec. 2.","section":"Sec. 2.2"},{"comment":"The statement 'we also observe that the logical error rates indeed scale no worse than O(p^{dp}) below threshold' is ambiguous; the intended meaning is presumably that the logical error rate decays at least as fast as p^{dp}.","section":"Sec. 2.2"},{"comment":"The hierarchical decoding section would benefit from a pseudo-code description or explicit equations for the conversion matrices C_i and for the conditioning/normalization step; the current text is sufficient for a proof of principle but hard to reproduce exactly.","section":"Sec. 4"},{"comment":"The discussion of interleaving modules would benefit from a figure showing the resource-state labels and delay lengths for a larger L; Fig. 23 is for L=3 with n=4 and is clear, but the generalization to arbitrary n and L is stated only in a formula.","section":"Sec. 5"},{"comment":"In Table 1, the references to surface code protocols as 'cubic [1]' and 'cuboctahedral [7]' are unclear; please specify the resource-state size and the exact fusion network variant.","section":"Table 1"},{"comment":"There are minor typos, e.g., 'more granular control control' in the Conclusion.","section":"Sec. 6"}],"recommendation":"major_revision","confidential_remarks":"The distance-scaling conjecture is the crux of the paper's main quantitative claim. I would not require a full proof for acceptance, but the paper should clearly state in the abstract and conclusion that the footprint scalings are conjectural, and ideally include a robustness analysis under alternative distance scalings (e.g., if the distance grows as d^L / poly(L)). The thresholds would also be more convincing with error bars. The paper is otherwise well within the scope of the journal and is a strong candidate if these points are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the blocklet construction is genuinely new, and the simulation evidence is worth taking seriously. Send it to peer review, but make sure the referees see the distance conjecture as a load-bearing open problem rather than a minor appendix.\n\nWhat is actually new: the observation that a CSS code's syndrome readout circuit is itself a transversal operation, so you can concatenate the code with itself while keeping the resource state constant, is clean and original. It produces fusion networks with constant-sized resource states and varying, non-local connectivity. The Monte Carlo evidence for high erasure thresholds is direct and credible: 13.8% for an 8-qubit state, 19.1% for a 10-qubit state, 11.5% for a 12-qubit state, using an optimal erasure decoder. Those numbers compare favorably with the 12–12.7% of surface-code fusion networks. The paper is also honest about its limits: it flags the distance conjecture as open, reports a hierarchical decoder that underperforms by 13%, and does not overclaim the CBQC relevance.\n\nThe soft spots are real but not fatal. First, the footprint-per-logical-qubit scalings in Table 1—O(d), O(d^1.46), O(d^0.58)—rest entirely on the Appendix A conjecture that the protocol distance is c·d^L. The author states plainly that a proof is open, and the subthreshold fits are consistent but not proof. If the true distance grows more slowly, the advantage over surface codes shrinks. This is the main quantitative claim of the paper, so it deserves emphasis, not a footnote. Second, the thresholds are crossing points of the two largest simulated protocol sizes, with no error bars or finite-size uncertainty analysis. A shift of a percentage point or two would still leave the main qualitative conclusion intact, but the reported values look more precise than the evidence supports. Third, no code or data is released, which makes independent verification harder than it should be for a paper whose results are simulation-driven.\n\nThe citation pattern looks fine, and the construction is clearly distinct from foliation, fault-tolerant complexes, and tensor-network codes. The manuscript is well organized and the diagrams are effective.\n\nWho this is for: anyone working on photonic FBQC or on alternatives to surface codes in low-overhead fault-tolerant architectures. It deserves a serious referee, and the referee should push for error bars, a more systematic numerical test of the distance conjecture (even up to moderately larger L), and ideally code release.","headline":"Genuinely new blocklet concatenation construction with strong simulated thresholds, but the footprint scalings that beat surface codes rest on an unproven distance conjecture and the threshold numbers lack error bars.","tokens_in":23173,"tokens_out":1242,"would_cite":true,"duration_ms":14273,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68","81P70"],"pacs":["03.67.Pp","03.67.Lx"],"model":"deepseek-v4-flash","headline":"Blocklet concatenation turns a code's syndrome-readout circuit into a transversal gate, yielding fault-tolerant fusion networks with constant-sized resource states and erasure thresholds up to 19.1%.","keywords":["blocklet concatenation","fusion-based quantum computing","erasure threshold","code concatenation","transversal gates","photonic quantum computing","Pauli webs","fault-tolerant protocols"],"falsifier":"Exhaustively search small concatenated protocols, such as $[7,1,3]^3$ or $[5,1,3]^3$, for undetectable error strings of weight below $c\\cdot d^L$; finding one, or observing subthreshold logical-error-rate exponents inconsistent with $c\\cdot d^L$, would refute the distance conjecture and invalidate the $O(d)$, $O(d^{1.46})$, and $O(d^{0.58})$ footprint claims.","tokens_in":22226,"feed_emoji":"🧩","tokens_out":7712,"duration_ms":67155,"temperature":0.7,"pith_summary":"The paper introduces blocklet concatenation, a way to build fault-tolerant quantum-computing protocols by repeatedly concatenating a stabilizer code with itself while keeping the resource state fixed. In fusion-based quantum computing (FBQC), this yields fusion networks built from constant-sized resource states whose erasure thresholds can exceed those of comparable surface-code networks: the paper reports 13.8%, 19.1%, and 11.5% thresholds for 8-, 10-, and 12-qubit resource states, versus 12.0–12.7% for the surface-code baselines. The paper argues these protocols are candidates to replace surface codes in photonic FBQC because their footprint per logical qubit grows as $O(d)$, $O(d^{1.46})$, or $O(d^{0.58})$ instead of $d^2$. That footprint advantage depends on a distance-scaling conjecture that the paper states openly. The paper also provides techniques for logical operations, decoding, and photonic implementation, making the construction a full-stack recipe rather than a mere code family.","feed_headline":"Blocklet code beats surface-code thresholds in fusion computing","feed_subtitle":"Constant-size resource states deliver 13.8–19.1% erasure thresholds with sub-quadratic footprint scaling.","key_machinery":"The blocklet is the central object: a $2n$-qubit resource state that is a Bell pair with both ends encoded in the same $[n,k,d]$ code, represented as a box in a ZX diagram whose Pauli webs split into cups, caps, and membranes. The concatenation prescription replaces each blocklet with two layers of $n$ blocklets connected transversally, so every check and logical operator gets re-encoded while the resource state itself is unchanged. The mechanism is carried by encoded Pauli webs: cups from one layer and membranes from the other reassemble into encoded cups, caps, and product checks, which is what makes the hierarchy of checks and the distance growth possible.","core_discovery":"The central discovery is that the syndrome-readout circuit of an $[n,k,d]$ stabilizer code is itself a transversal operation of that code (for CSS codes, and specially for the five-qubit code), so executing that readout transversally across $n$ code blocks implements one level of code concatenation using only the original low-weight measurements. The corresponding $2n$-qubit resource state, called a blocklet, is an encoded Bell pair; chaining blocklets and applying the same prescription repeatedly produces $[n,k,d]^L$ protocols whose resource state stays identical at every level while the code distance grows. Encoded cups, caps, and membranes—Pauli webs of the blocklet—combine to form hierarchical checks and logical operators. Simulated erasure thresholds for the $[4,2,2]$, $[5,1,3]$, and $[6,4,2]$ families are 13.8%, 19.1%, and 11.5%, and the paper conjectures distances $c\\cdot d^L$ (Appendix A) that give footprint scalings $O(d)$, $O(d^{1.46})$, and $O(d^{0.58})$.","pith_inferences":["If the distance conjecture is proved, the $O(d)$ footprint of the $[4,2,2]$ family would make blocklet protocols competitive with or better than surface codes on total photonic overhead; if it fails, the threshold advantages alone may not justify the non-local connectivity.","The author notes that any phaseless ZX diagram has cup, cap, and membrane Pauli webs, so the blocklet construction could in principle be searched systematically over stabilizer states to find better threshold–footprint trade-offs than the four families reported.","Because XX and ZZ erasure thresholds can be made asymmetric (e.g., 21% versus 6.7% for $[4,1,2]$), orienting the fusion network to the dominant loss channel, or applying the paper's linking procedure, could extract further effective threshold gains in hardware where correlated erasures are controlled.","The $O(\\log d)$ delay-line count for photonic implementation suggests the practical overhead of the non-local fusion network grows slowly enough that blocklet protocols remain implementable at scales where surface-code fusion networks would need $d^2$ footprint."],"forward_implications":["$[4,2,2]$ blocklet protocols use an 8-qubit resource state, have a 13.8% erasure threshold, and $O(d)$ footprint per logical qubit, exceeding the 12.0–12.7% thresholds and $d^2$ footprint of the surface-code baselines under the distance conjecture.","$[5,1,3]$ blocklet protocols reach a 19.1% erasure threshold with 10-qubit resource states, and their footprint scales as about $0.47 d^{1.46}$.","$[6,4,2]$ blocklet protocols use 12-qubit resource states, have an 11.5% threshold, and about $(1/6) d^{0.58}$ footprint, while encoding $k^L$ logical qubits per block.","Blocklet protocols support universal computation: logical GHZ-state preparation, Hadamard and state injection for CSS codes, special logical operations for the $[5,1,3]$ code, and selective addressing of individual logical qubits for $k>1$.","The hierarchical decoder decodes a distance-$d$ protocol in $O(\\log d)$ parallel steps; for $[5,1,3]$ it gives a 16.7% erasure crossing versus 19.1% for the optimal decoder and a 1.7% Pauli-error crossing, and the protocols can be implemented with interleaving modules using $O(\\log d)$ switchable delay lines."],"supporting_citations":[{"why":"Defines fusion-based quantum computation and the resource-state/fusion-network model that the paper's protocols target.","marker":"[1]"},{"why":"Provides the ZX-calculus Pauli-web framework used to define blocklets, checks, and logical membranes.","marker":"[5]"},{"why":"Introduces foliation, the alternative conversion of stabilizer circuits to fusion networks that blocklet concatenation is contrasted with.","marker":"[6]"},{"why":"Gives the surface-code fusion-network baselines, including the cuboctahedral 8-qubit variant with 12.7% threshold, that blocklet protocols are compared against.","marker":"[7]"},{"why":"Supports using resource-state size as a cost proxy when comparing blocklet and surface-code protocols.","marker":"[24]"},{"why":"Active-volume architecture used to show how logical operations such as GHZ preparation and fusions can be compiled with blocklet protocols.","marker":"[25]"},{"why":"Interleaving modular architecture used for the photonic implementation with $O(\\log d)$ switchable delay lines.","marker":"[26]"},{"why":"Source of the $[5,1,3]$ code whose blocklet family achieves the 19.1% erasure threshold.","marker":"[27, 28]"}],"fun_headline_variants":["Blocklet codes beat surface-code thresholds in fusion quantum computing","Constant-size resource states achieve 19% erasure thresholds in fusion codes","Blocklet concatenation gives low-overhead fault tolerance for fusion quantum computing","Fusion quantum computing benefits from constant-size blocklet resource states","Blocklet protocol: high erasure thresholds with constant-size resource states"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claimed footprint advantage over surface codes rests on the unproved conjecture that the code distance of an $[n,k,d]^L$ blocklet protocol is $c\\cdot d^L$ with $c = d_{\\mathrm{prod}}/d^2$, supported by subthreshold simulation data but not proven.","fun_headline_variants_meta":{"raw":{"variants":["Blocklet codes beat surface-code thresholds in fusion quantum computing","Constant-size resource states achieve 19% erasure thresholds in fusion codes","Blocklet concatenation gives low-overhead fault tolerance for fusion quantum computing","Fusion quantum computing benefits from constant-size blocklet resource states","Blocklet protocol: high erasure thresholds with constant-size resource states"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001775,"raw_usage":{"total_tokens":7028,"prompt_tokens":1003,"completion_tokens":6025,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":619,"completion_tokens_details":{"reasoning_tokens":5935}},"tokens_in":619,"tokens_out":6025,"duration_ms":40812,"temperature":1.0,"reasoning_tokens":5935,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:57:40.088821+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhaustively search small concatenated protocols, such as $[7,1,3]^3$ or $[5,1,3]^3$, for undetectable error strings of weight below $c\\cdot d^L$; finding one, or observing subthreshold logical-error-rate exponents inconsistent with $c\\cdot d^L$, would refute the distance conjecture and invalidate the $O(d)$, $O(d^{1.46})$, and $O(d^{0.58})$ footprint claims.","supporting_citations":[{"cited_title":"Bartolucci, P","cited_arxiv_id":null,"evidence_quote":"Defines fusion-based quantum computation and the resource-state/fusion-network model that the paper's protocols target."},{"cited_title":"Bombin, D","cited_arxiv_id":null,"evidence_quote":"Provides the ZX-calculus Pauli-web framework used to define blocklets, checks, and logical membranes."}],"review_version":2}