Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Comparison of schemes for highly loss tolerant photonic fusion based quantum computing

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read By substituting the 6-ring fusion network with the {2,2}-encoded cuboctahedral 'loopy diamond' complex and applying exposure-based adaptivity, the paper reports a 9.0% loss-per-photon threshold — up from 7.5% — using a 32-photon resource…

desk verdict Useful landscape paper with one genuinely new but under-documented threshold; the 9.0% number needs supporting data before it can be trusted. read the letter →

arxiv 2506.11975 v1 pith:KA4PL64L submitted 2025-06-13 quant-ph

Sara Bartolucci , Tom Bell , Hector Bombin , Patrick Birchall , Jacob Bulmer , Christopher Dawson , Terry Farrelly , Samuel Gartenstein
show 13 more authors
Mercedes Gimeno-Segovia Daniel Litinski Yehua Liu Robert Knegjens Naomi Nickerson Andrea Olivo Mihir Pant Ashlesha Patil Sam Roberts Terry Rudolph Chris Sparrow David Tuckett Andrzej Veitia
This is my paper · ORCID
classification quant-ph
keywords fusion-basedquantumcomputingphotoniclossperphotonthresholdadaptivefusionmeasurementdynamicbiasarrangementShorcodeloopydiamondnetworkresourcestatecosting
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the geometry of the fusion network is as important as the local encoding in determining how much photon loss a photonic quantum computer can survive. Its central finding is that a $\{2,2\}$-encoded version of the cuboctahedral 'loopy diamond' fusion complex, combined with exposure-based adaptivity, reaches a 9.0% loss-per-photon threshold where the comparable 6-ring network reaches 7.5%, and its 32-photon resource state is cheaper to prepare. The paper also assembles a unified comparison of published fusion-based schemes under two metrics — loss threshold and preparation cost counted in three-GHZ states — and states the fundamental ceilings: $38.2\% \approx (3-\sqrt{5})/2$ for non-adaptive schemes that salvage 'scrap' information and 50% for adaptive schemes. If these numbers hold, hardware engineers can read off directly how much per-photon loss a given resource-state size and preparation route can tolerate.

What carries the argument

The argument runs on four objects. The fusion network itself — the cuboctahedral fusion complex of [3] engineered into the 8-looped 'loopy diamond' (8-LD) resource state — is the geometry on which the threshold depends. The local encoding is a Shor code $\{n,m\}$ ($n$ X-repetitions and $m$ Z-repetitions, a quantum parity check code), which performs fusions transversally across encoded qubits so that loss of a few physical photons does not erase the logical fusion. Exposure-based adaptivity, from the dynamic bias arrangement of [2], staggers physical fusions and uses feedforward to pick fusion failure bases, and in the high-loss regime single-qubit measurements, such that information protecting against the growth of large erasure clusters is prioritized. The paper's two yardsticks are the loss-per-photon threshold (LPPT), the uniform per-photon loss probability at which the network's decoding breaks, and a preparation-cost count: the number of three-GHZ states needed to build the resource state with destructive Type-II fusions under a one-half success probability and no recycling of failed states.

What would settle it

Run an independent Monte Carlo decoding simulation of the $\{2,2\}$-encoded loopy diamond network under the paper's per-photon loss model and check whether the threshold reproduces 9.0%; separately, derive the non-adaptive scrap-based loss limit from first principles and check whether it equals $(3-\sqrt{5})/2 \approx 38.2\%$, since any discrepancy would overturn the paper's comparison frame.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that the cuboctahedral fusion complex of [3], realized as the 8-loop 'loopy diamond' (8-LD) resource state and encoded with a $\{2,2\}$ Shor code (two X-repetitions and two Z-repetitions), outperforms the standard 6-ring network: with exposure-based adaptivity it reaches a 9.0% loss-per-photon threshold versus 7.5% for the 6-ring at the same encoding, and under the paper's preparation-cost model it needs 1120 three-GHZ states versus 1520. Larger encodings on the same network continue the trend, with $\{4,3\}$ giving 15.4% and $\{7,4\}$ giving 18.8%. The paper places all of these points inside limits it reports as universal: 29.3% for non-adaptive schemes that erase on any loss, $38.2\% \approx (3-\sqrt{5})/2$ for non-adaptive schemes that use the non-stabilizer 'scraps' left after loss, and 50% for adaptive schemes with single-photon measurements. The paper further claims that counting photons in a resource state is a poor footprint metric and that preparation cost in three-GHZ states, though approximate, reverses some naive comparisons.

Load-bearing premise

The paper's landscape comparison and its stated fundamental limits rest on the 38.2% loss-per-photon threshold for non-adaptive schemes with scraps, a number attributed to an unpublished manuscript and presented without independent verification.

Editorial extensions

If this is right

  • Switching network geometry alone buys tolerance: at the same $\{2,2\}$ encoding and the same exposure-based adaptivity, the loopy diamond network's threshold is 9.0% against the 6-ring's 7.5%.
  • Bigger Shor encodings on the loopy diamond network keep paying off, reaching 15.4% at $\{4,3\}$ and 18.8% at $\{7,4\}$.
  • Preparation cost, not photon count, is the operative footprint: the 32-photon 8-LD state costs 1120 three-GHZ states, less than the 24-photon 6-ring's 1520, so a larger resource state can be the cheaper one.
  • Adaptivity improves the threshold-versus-size tradeoff across the entire surveyed range, which means resource-state savings and loss tolerance can be pursued together.
  • The 'photons per encoded fusion' metric of [17] is misleading because it depends on an arbitrary choice of which concatenation level is 'the encoded qubit' and ignores resource-state size entirely.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the 38.2% scrap-based non-adaptive limit originates in an unpublished manuscript, an independent derivation from first principles would be a valuable check on how much headroom adaptivity really provides beyond non-adaptive schemes.
  • The costing model's suggestion that tree-like (ZX-representable) states are cheaper to prepare points to a design heuristic: future fusion-network search could screen candidate complexes for tree-representability before full threshold simulation.
  • Extending the costing model to recycle fusion-failure states, which the paper notes is possible, would likely narrow the cost gaps between topologies and could reorder the rankings among very large encodings.
  • The 50% adaptive ceiling implies that the highest-loss regime will lean on single-qubit measurements, so hardware that makes fast, low-loss single-qubit measurement available directly unlocks the upper end of the tolerance range.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This manuscript from the PsiQuantum architecture group compiles loss-per-photon thresholds (LPPT) for a range of photonic fusion-based quantum computing (FBQC) schemes under a common resource-state-size metric. The paper reviews earlier randomized, static-bias, and locally adaptive schemes; introduces a new result attributed to the authors, namely a 9.0% LPPT for a {2,2}-encoded 8-qubit 'loopy diamond' fusion network using exposure-based adaptivity (Table II, row 2); compares this with 7.5% for the {2,2}-encoded 6-ring; and frames the landscape against fundamental limits of 29.3%, 38.2%, and 50% loss per photon. An appendix proposes a 3-GHZ-based resource-state costing heuristic and gives a lower bound of (S-2)^2 for the number of 3-GHZ states. The paper contains no simulation code, decoder specification, or numerical data for the new threshold, and it relies on an unpublished companion manuscript for the 38.2% non-adaptive limit.

Significance. The paper's comparative survey is potentially useful: it collects data from [1,2,4,7,17] into a single table and proposes a footprint metric that removes ambiguities in 'photons per encoded fusion'. The authors are candid about the optimistic assumptions in the costing model and explicitly state that the 8-LD and 6-ring schemes are not directly comparable. The central new claim, however, is currently not independently verifiable: the 9.0% threshold is stated without methods or data, and the 38.2% fundamental limit is attributed to an unpublished manuscript. If the missing numerical support were supplied, the comparison would be a valuable reference point for the FBQC community; in its present form the paper is a roadmap or extended abstract rather than a self-contained research article.

major comments (3)
  1. [Main text, 'Some of these new fusion networks are competitively performing...'; Table II, row 2] The new 9.0% LPPT for the {2,2}-encoded 8-LD network is the paper's central numerical claim, but the manuscript provides no code, no decoder specification, no loss-model equations, no threshold-search procedure, and no raw data. The row is attributed to [2,3], yet [2] reports 6-ring networks and [3] defines fusion complexes without computing loss thresholds; neither source alone supports this number. This is a load-bearing reproducibility gap: the reader cannot distinguish a numerical result from a projection. The authors should make the simulation code and data available, or at minimum give a precise algorithmic description of the decoder, the error model, and the threshold-extraction method.
  2. [Main text, fundamental-limits paragraph citing [19]] The non-adaptive limit with scraps, quoted as 38.2% and equal to (3 - sqrt(5))/2, is attributed to an unpublished manuscript by the same group and is asserted without derivation. Because the paper uses this limit to argue that adaptive schemes approach the fundamental bound, the reader cannot assess the framing. The authors should either include a self-contained proof or derivation of the 38.2% limit in an appendix, or replace the reference with a publicly available source before publication.
  3. [Main text, 'improved threshold of 9.0% (vs. 7.5%...)'] The text headlines 'improved threshold of 9.0% (vs. 7.5% for the comparable 6-ring fusion network)' and immediately concedes that the 8-LD state is not directly comparable. The two schemes differ in both unencoded fusion-network geometry and encoded resource-state size (32 vs 24 qubits). The phrase 'comparable' is therefore misleading: the higher threshold could result from the larger state, from the different complex, or from a combination. Please present a controlled comparison, for example at fixed resource-state size or with an analysis that isolates the effect of the fusion complex, or revise the claim to 'higher threshold at larger resource-state size'.
minor comments (5)
  1. [Figure 1 caption] The caption contains a duplicated 'of': 'summary of of optical FBQC schemes' should be 'summary of optical FBQC schemes'.
  2. [Table II] There are spelling errors in the table: 'reuslts' should be 'results' and 'Statis bias arrangement' should be 'Static bias arrangement'.
  3. [Main text, 'This scheme uses a resource state with 8 qubits (8-LD)'] This sentence is ambiguous because Table II lists the {2,2}-encoded 8-LD as 32 qubits. Please clarify that 8-LD denotes the unencoded 8-qubit state before local encoding.
  4. [Main text, 'scraps'] The term 'scraps' is used without a formal definition; the parenthetical 'non-stabilizer information' should be expanded into a sentence or supplied with a reference.
  5. [Appendix A] The claim that the relative cost ordering in Table I 'will remain true irrespective of preparation method details' is stated without sensitivity analysis. Since the authors acknowledge the heuristic is not guaranteed optimal for 6-ring states, a short sensitivity check or a statement of which preparation assumptions would change the ordering would strengthen the claim.

Circularity Check

1 steps flagged · score 4.0 of 10

One load-bearing self-citation to unpublished [19] for the 38.2% limit; the 9.0% threshold is independent but underdocumented.

  1. self citation load bearing [Section 'Loss tolerance of published FBQC schemes', paragraph discussing fundamental limits (page 2, near Fig. 1)]
    "However, in [19] we show even without adaptivity an LPPT of 38.2% (≈ (3 − √ 5)/2) can be achieved by accounting for the non-stabilizer information (known as scraps) which may still available in the presence of loss."

    The 38.2% non-adaptive loss-per-photon threshold is presented as a fundamental limit and used to frame the maximum potential of non-adaptive FBQC. Its only support is [19], which is listed as 'PsiQuantum, Manuscript in preparation' — an unpublished work by the same authors. The derivation is not included in this paper, so within the argument the limit is an input assumption backed solely by self-citation rather than by an independently verifiable derivation.

full rationale

This paper is primarily a comparative survey, and its central new numerical claim is the 9.0% loss-per-photon threshold for the {2,2}-encoded 8-LD (loopy diamond) network. That threshold is obtained by applying the self-authored adaptivity methods of [2] to the self-authored fusion complex of [3], but it is a new computation rather than a fit of an input, and the paper compares it against external benchmarks from [4], [7], and [17]. The main circularity concern is the 38.2% non-adaptive limit with scraps, which is attributed solely to the unpublished self-citation [19] and is not derived in the manuscript; this makes the framing 'fundamental limit' reliant on a self-citation to an unverified source. However, that 38.2% limit is not the basis of the 9.0% threshold, and the wide Table II comparison draws on independent groups. The 9.0% result itself is underdocumented here (no simulation code, decoder specification, or detailed loss model is given), which is a reproducibility issue rather than a circular reduction. Weighing these factors, there is one load-bearing self-citation to unpublished work, but the central threshold claim retains independent content, so the circularity score is 4.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central comparison depends on LPPT values from multiple sources and on a simplified costing model. The only genuinely new theoretical quantity, the 38.2% non-adaptive loss threshold, is attributed to an unpublished self-cited manuscript, so it is treated as an assumption here.

assumptions (3)
  • domain assumption The loss per photon threshold (LPPT) error model applies equal loss probability to every photon in each resource state.
    This is the standard error model used to compare FBQC schemes, but it may not capture all realistic loss sources.
  • domain assumption Resource state costing assumes each Type-II fusion succeeds with probability 1/2, uses 1/p overhead for p-probability steps, and discards fusion-failure states.
    These assumptions are stated in Appendix A and are acknowledged as optimistic.
  • ad hoc to paper The non-adaptive loss threshold with scraps is 38.2% approximately (3-sqrt(5))/2, as claimed in the unpublished manuscript [19] by the authors.
    This fundamental limit is cited from an unpublished PsiQuantum manuscript and is not independently verifiable from this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Comparison of schemes for highly loss tolerant photonic fusion based quantum computing." pith.science (2026). https://pith.science/paper/KA4PL64L

@misc{pith2026250611975,
  author       = {Pith},
  title        = {Pith review of: Comparison of schemes for highly loss tolerant photonic fusion based quantum computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KA4PL64L}},
  note         = {Machine review of arXiv:2506.11975}
}
read the original abstract

We summarize the performance of recently-proposed methods for achieving fault tolerant fusions-based quantum computation with high tolerance to qubit loss, specifically aimed at photonic implementations.

Figures

Figures reproduced from arXiv: 2506.11975 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Single-photon-boosted type-I fusion gates

    quant-ph 2026-03 unverdicted novelty 7.0 of 10

    A type-I fusion gate, boosted with four single-photon ancillas and passive linear optics, reaches 3/4 success probability via a distillation protocol.

  2. Quantifying Pauli Errors in Single-Photon Resource-State Generation

    quant-ph 2026-08 accept novelty 6.0 of 10

    A scheme that extracts Pauli error rates for emitter-generated photonic resource states from first-order coherence and cross-correlation measurements.

Reference graph

Works this paper leans on

25 extracted references · 20 canonical work pages · cited by 2 Pith papers

  1. [17]

    S.-H. Lee, S. Omkar, Y . S. Teo, and H. Jeong, npj Quantum Information 9, 39 (2023)

    work page 2023

  2. [2]

    For the {2,2}-encoded 6-ring resource state this in- creases the LPPT to 7.5%

    termed exposure based adaptivity optimizes the choice of fusion bases for physical fusions (which may include single qubit measurements in the high loss regime) in a topological fusion network such that information which is more impor- tant for suppressing the growth of large erasure clusters is pri- oritized. For the {2,2}-encoded 6-ring resource state t...

    work page

  3. [3]

    Increasing error tolerance in quantum computers with dynamic bias arrangement,

    H. Bombin, C. Dawson, N. Nickerson, M. Pant, and J. Sullivan, “Increasing error tolerance in quantum computers with dynamic bias arrangement,” (2023), arXiv:2303.16122 [quant-ph]

    arXiv 2023

  4. [19]

    Hilaire, Y

    P. Hilaire, Y . Castor, E. Barnes, S. E. Economou, and F. Grosshans, PRX Quantum 4, 040322 (2023)

    work page 2023

  5. [1]

    Bartolucci, P

    S. Bartolucci, P. Birchall, H. Bombin, H. Cable, C. Dawson, M. Gimeno-Segovia, E. Johnston, K. Kieling, N. Nickerson, M. Pant, F. Pastawski, T. Rudolph, and C. Sparrow, Nature Communications 14 (2023), 10.1038/s41467-023-36493-1

    doi:10.1038/s41467-023-36493-1 2023

  6. [4]

    Fault-tolerant com- plexes,

    H. Bombin, C. Dawson, T. Farrelly, Y . Liu, N. Nickerson, M. Pant, F. Pastawski, and S. Roberts, “Fault-tolerant com- plexes,” (2023), arXiv:2308.07844 [quant-ph]

    arXiv 2023

  7. [5]

    T. J. Bell, L. A. Pettersson, and S. Paesani, PRX Quantum 4, 020328 (2023)

    work page 2023

  8. [6]

    Paesani and B

    S. Paesani and B. J. Brown, Physical Review Letters 131, 120603 (2023)

    work page 2023

Show all 25 references
  1. [7]

    Sahay, J

    K. Sahay, J. Claes, and S. Puri, Physical Review Letters 131, 120604 (2023)

    2023

  2. [8]

    Pankovich, A

    B. Pankovich, A. Kan, K. H. Wan, M. Ostmann, A. Neville, S. Omkar, A. Sohbi, and K. Br ´adler, Physical Review Letters 133, 050604 (2024)

    2024

  3. [9]

    D. E. Browne and T. Rudolph, Phys. Rev. Lett. 95, 010501 (2005), arXiv:quant-ph/0405157

    2005 arXiv

  4. [10]

    Varnava, D

    M. Varnava, D. E. Browne, and T. Rudolph, Physical review letters 97, 120501 (2006)

    2006

  5. [11]

    Azuma, K

    K. Azuma, K. Tamaki, and H.-K. Lo, Nature communications 6, 1 (2015)

    2015

  6. [12]

    Y . Li, P. C. Humphreys, G. J. Mendoza, and S. C. Benjamin, Physical Review X 5, 041007 (2015)

    2015

  7. [13]

    S.-W. Lee, K. Park, T. C. Ralph, and H. Jeong, Physical review letters 114, 113603 (2015)

    2015

  8. [14]

    Ewert, M

    F. Ewert, M. Bergmann, and P. van Loock, Physical review letters 117, 210501 (2016)

    2016

  9. [15]

    S.-W. Lee, T. C. Ralph, and H. Jeong, Physical Review A 100, 052303 (2019)

    2019

  10. [16]

    Hilaire, E

    P. Hilaire, E. Barnes, S. E. Economou, and F. Grosshans, Phys- ical Review A 104, 052623 (2021)

    2021

  11. [18]

    W. Song, N. Kang, Y .-S. Kim, and S.-W. Lee, Phys. Rev. Lett. 133, 050605 (2024), arXiv:2408.01041 [quant-ph]

    2024 arXiv

  12. [20]

    PsiQuantum, Manuscript in preparation

  13. [21]

    Bomb ´ın, M

    H. Bomb ´ın, M. Pant, S. Roberts, and K. I. Seetharam, PRX Quantum 5, 010302 (2024)

    2024

  14. [22]

    Bartolucci, P

    S. Bartolucci, P. Birchall, D. Bonneau, H. Cable, M. Gimeno- Segovia, K. Kieling, N. Nickerson, T. Rudolph, and C. Spar- row, arXiv preprint arXiv:2109.13760 (2021)

    2021 arXiv

  15. [23]

    Lee and H

    S.-H. Lee and H. Jeong, Quantum 7, 1212 (2023)

    2023

  16. [24]

    Kieling, D

    K. Kieling, D. Gross, and J. Eisert, JOSA B 24, 184 (2007)

    2007

  17. [25]

    number of 3GHZ states

    D. Gross, K. Kieling, and J. Eisert, Physical Review A—Atomic, Molecular, and Optical Physics 74, 042343 (2006). Appendix A: Resource state costing As a simple footprint metric we count the “number of 3GHZ states” that would be required to make the target re- source state usin...

    2006

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.