pith. machine review for the scientific record. sign in

arxiv: 2604.14269 · v1 · submitted 2026-04-15 · 🪐 quant-ph

Recognition: unknown

AI-Enabled Decoding of Qubit Loss for Quantum Error-Correcting Codes

Yuqing Wang , Xiaotian Nie , Jiale Dai , Zhongyi Ni , Tao Zhang , Hui Zhai , Linghui Chen
Authors on Pith no claims yet

Pith reviewed 2026-05-10 13:26 UTC · model grok-4.3

classification 🪐 quant-ph
keywords qubit lossquantum error correctionstabilizer codesgraph neural networksyndrome decodingPauli errorsfault-tolerant quantum computation
0
0 comments X

The pith

A graph neural network decoder corrects both Pauli errors and qubit loss locations from syndrome histories with higher accuracy than matching algorithms.

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

The paper introduces a decoder that processes sequences of stabilizer measurements in quantum error-correcting codes to address qubit loss, which disrupts the standard stabilizer structure. The approach uses spatial and temporal patterns in the measurement history to simultaneously fix Pauli errors and locate lost qubits. This yields higher logical accuracy than the minimum-weight perfect matching algorithm and even versions of that algorithm supplied with final-round loss information. The decoder also identifies more than 90 percent of loss events after ten rounds of measurements, which supports practical recovery steps such as qubit reinitialization. The method supplies a scalable route to handling multiple error sources in fault-tolerant quantum computation.

Core claim

The authors develop a spatiotemporal graph neural network that takes syndrome histories as input and performs a dual-head task: it outputs corrections for standard Pauli errors while also identifying the locations of qubit losses. When tested on simulated data, this decoder achieves significantly higher logical accuracy than both the traditional minimum-weight perfect matching algorithm and delayed-erasure variants that incorporate qubit loss information from the final round, and it locates more than 90 percent of loss events after accumulating measurements over the subsequent ten rounds.

What carries the argument

The spatiotemporal Graph Neural Network (STGNN) that extracts spatial and temporal correlations from sequences of stabilizer measurements to output both Pauli error corrections and qubit loss locations.

If this is right

  • The decoder enables more reliable logical operations by addressing qubit loss without requiring separate detection modules.
  • Locating losses after ten rounds of measurements supports qubit reinitialization techniques on platforms such as atom arrays.
  • The method supplies a scalable framework that can manage multiple error types in fault-tolerant quantum computation.
  • Its parallel input structure yields faster inference than recurrent alternatives while preserving performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same architecture could be retrained on syndrome data from other quantum codes to handle loss in surface codes or color codes without redesigning the decoder logic.
  • If the learned separation between loss and Pauli errors holds on hardware, it would reduce the need for dedicated loss-detection circuits in future devices.
  • Extending the input window beyond ten rounds might improve loss localization further, at the cost of additional latency in real-time decoding.

Load-bearing premise

The decoder succeeds only if the patterns present in simulated syndrome histories are representative enough for the network to separate qubit loss events from Pauli errors when the same network is applied to real quantum hardware.

What would settle it

Apply the trained decoder to experimental stabilizer measurement records from a physical quantum processor and check whether the resulting logical error rate is lower than that obtained with minimum-weight perfect matching while also verifying the reported loss detection rate.

Figures

Figures reproduced from arXiv: 2604.14269 by Hui Zhai, Jiale Dai, Linghui Chen, Tao Zhang, Xiaotian Nie, Yuqing Wang, Zhongyi Ni.

Figure 1
Figure 1. Figure 1: FIG. 1. Performance of Pauli error v.s. qubit loss in stabilizer measurements using the rotated surface code as an example. A [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Detailed architecture of the STGNN decoder. The network processes input syndrome sequences [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Comparison of decoder performance under qubit loss. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Qubit loss identification performance vs. decision [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Performance analysis of unrecognized qubit loss after [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
read the original abstract

Qubit loss is a major source of error in quantum computation, as it invalidates the algebraic structure of the standard stabilizer formalism for quantum error-correcting codes. On the one hand, it complicates decoding; on the other hand, it introduces stochastic flicker patterns in stabilizers as a hallmark of qubit loss. Here, we develop an artificial-intelligence-enabled decoder based on a spatiotemporal Graph Neural Network (STGNN) architecture to extract spatial and temporal correlations from syndrome histories. Our decoder performs a dual-head task, simultaneously correcting standard Pauli errors and identifying the locations of qubit loss. Our decoder achieves significantly higher logical accuracy than both the traditional minimum-weight perfect matching (MWPM) algorithm and even delayed-erasure MWPM decoders that use qubit loss information from the final round as input. Our decoder can also identify more than 90% of loss locations after accumulating stabilizer measurements over the subsequent ten rounds, thereby facilitating qubit reinitialization, for instance, via the continuous loading technique on the atom array platform. For both tasks, our STGNN performs nearly identically to a modified version of AlphaQubit, but it employs a parallel input structure, giving it an advantage in inference time over modified AlphaQubit's recurrent input structure. This work provides a robust and scalable framework for correcting qubit loss errors, paving the way for more efficient fault-tolerant quantum computation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript introduces a spatiotemporal Graph Neural Network (STGNN) decoder for quantum error-correcting codes that simultaneously corrects Pauli errors and identifies qubit loss locations by extracting correlations from syndrome histories. It reports significantly higher logical accuracy than standard MWPM and delayed-erasure MWPM baselines, over 90% loss identification after accumulating ten rounds of measurements, and performance comparable to a modified AlphaQubit decoder while offering faster inference via its parallel input structure. The work targets platforms such as atom arrays and includes training details, noise models, and numerical comparisons in the full text.

Significance. If the empirical results hold under the reported conditions, this provides a scalable, dual-task ML framework for loss-aware decoding that addresses a key practical error source in near-term quantum hardware. The combination of high loss-detection rates enabling reinitialization and improved logical accuracy over established decoders, together with the inference-time advantage, represents a concrete step toward more efficient fault-tolerant protocols. The manuscript's inclusion of direct baseline comparisons and code-distance specifics strengthens its utility as a reference for decoder development.

minor comments (3)
  1. Abstract: The claim of 'significantly higher logical accuracy' is quantified in the results section; adding a brief parenthetical example (e.g., 'X% improvement at distance d=5') to the abstract would improve immediate readability without altering length.
  2. Methods/Results: The modifications to AlphaQubit are described for the comparison; explicitly listing the architectural differences (recurrent vs. parallel input) in a dedicated paragraph or table would aid reproducibility.
  3. Figures: Ensure all performance plots include error bars derived from multiple random seeds or trials, consistent with the training details already provided in the text.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of our work on the STGNN decoder for simultaneous Pauli error correction and qubit loss identification. The referee correctly notes the performance advantages over MWPM and delayed-erasure MWPM, the >90% loss recovery after ten rounds, and the inference-time benefit relative to modified AlphaQubit. Given the recommendation for minor revision and the absence of any specific major comments, we see no need for changes to the manuscript at this stage.

Circularity Check

0 steps flagged

No significant circularity; empirical ML results are self-contained

full rationale

The paper reports performance of a trained STGNN decoder on simulated syndrome data for Pauli correction and loss detection, with direct numerical comparisons to MWPM baselines. No derivation chain exists that reduces predictions to inputs by construction, no self-definitional parameters, and no load-bearing self-citations for core claims. Training details, noise models, and held-out evaluation are supplied independently of the reported metrics, making the results falsifiable against external benchmarks rather than tautological.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The approach rests on the assumption that qubit loss manifests as detectable stochastic patterns in stabilizer time series and that a graph neural network can be trained to extract those patterns from simulated data.

free parameters (1)
  • STGNN weights and biases
    All network parameters are fitted during supervised training on simulated syndrome histories.
axioms (1)
  • domain assumption Qubit loss produces identifiable flicker patterns in stabilizer measurements over multiple rounds
    Invoked in the abstract as the hallmark that enables the dual-head task.

pith-pipeline@v0.9.0 · 5558 in / 1225 out tokens · 31804 ms · 2026-05-10T13:26:32.461473+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

39 extracted references · 8 canonical work pages

  1. [1]

    Preskill, Quantum computing in the nisq era and be- yond, Quantum2, 79 (2018)

    J. Preskill, Quantum computing in the nisq era and be- yond, Quantum2, 79 (2018)

    2018

  2. [2]

    P. W. Shor, Scheme for reducing decoherence in quantum computer memory, Physical review A52, R2493 (1995)

    1995

  3. [3]

    A. R. Calderbank and P. W. Shor, Good quantum error-correcting codes exist, Physical Review A54, 1098 (1996)

    1996

  4. [4]

    M. A. Nielsen and I. L. Chuang,Quantum computation and quantum information(Cambridge university press, 2010)

    2010

  5. [5]

    B. M. Terhal, Quantum error correction for quantum memories, Reviews of Modern Physics87, 307 (2015)

    2015

  6. [6]

    E. T. Campbell, B. M. Terhal, and C. Vuillot, Roads towards fault-tolerant universal quantum computation, Nature549, 172 (2017)

    2017

  7. [7]

    Gottesman,Stabilizer codes and quantum error cor- rection(California Institute of Technology, 1997)

    D. Gottesman,Stabilizer codes and quantum error cor- rection(California Institute of Technology, 1997)

    1997

  8. [8]

    L. Hu, Y. Ma, W. Cai, X. Mu, Y. Xu, W. Wang, Y. Wu, H. Wang, Y. P. Song, C.-L. Zou, S. M. Girvin, L.-M. Duan, and L. Sun, Quantum error correction and univer- sal gate set operation on a binomial bosonic logical qubit, Nature Physics15, 503–508 (2019)

    2019

  9. [9]

    Google Quantum AI and Collaborators, Suppressing quantum errors by scaling a surface code logical qubit, Nature614, 676 (2023)

    2023

  10. [10]

    Google Quantum AI and Collaborators, Quantum error correction below the surface code threshold, Nature638, 920 (2025)

    2025

  11. [11]

    Putterman, K

    H. Putterman, K. Noh, C. T. Hann, G. S. MacCabe, S. Aghaeimeibodi, R. N. Patel, M. Lee, W. M. Jones, H. Moradinejad, R. Rodriguez, N. Mahuli, J. Rose, J. C. Owens, H. Levine, E. Rosenfeld, P. Reinhold, L. Mon- celsi, J. A. Alcid, N. Alidoust, P. Arrangoiz-Arriola, J. Barnett, P. Bienias, H. A. Carson, C. Chen, L. Chen, H. Chinkezian, E. M. Chisholm, M.-H....

    2025

  12. [12]

    Postler, S

    L. Postler, S. Heußen, I. Pogorelov, M. Rispler, T. Feld- ker, M. Meth, C. D. Marciniak, R. Stricker, M. Ring- bauer, R. Blatt, P. Schindler, M. M¨ uller, and T. Monz, Demonstration of fault-tolerant universal quantum gate operations, Nature605, 675 (2022)

    2022

  13. [13]

    Y. Hong, E. Durso-Sabina, D. Hayes, and A. Lucas, En- tangling four logical qubits beyond break-even in a non- local code, Physical Review Letters133, 180601 (2024)

    2024

  14. [14]

    Demonstration of Logical Qubits and Repeated Error Correction with Better-than- Physical Error Rates

    A. Paetznick, M. P. da Silva, C. Ryan-Anderson, J. M. Bello-Rivas, J. P. Campora, A. Chernoguzov, J. M. Dreil- ing, C. Foltz, F. Frachon, J. P. Gaebler, T. M. Gatter- man, L. Grans-Samuelsson, D. Gresh, D. Hayes, N. He- witt, C. Holliman, C. V. Horst, J. Johansen, D. Luc- chetti, Y. Matsuoka, M. Mills, S. A. Moses, B. Neyenhuis, A. Paz, J. Pino, P. Siegfr...

    work page arXiv 2024

  15. [15]

    Bluvstein, S

    D. Bluvstein, S. J. Evered, A. A. Geim, S. H. Li, H. Zhou, T. Manovitz, S. Ebadi, M. Cain, M. Kali- nowski, D. Hangleiter, J. P. Bonilla Ataides, N. Maskara, I. Cong, X. Gao, P. Sales Rodriguez, T. Karolyshyn, G. Semeghini, M. J. Gullans, M. Greiner, V. Vuleti´ c, and M. D. Lukin, Logical quantum processor based on reconfigurable atom arrays, Nature626, 58 (2024)

    2024

  16. [16]

    Bluvstein, A

    D. Bluvstein, A. A. Geim, S. H. Li, S. J. Evered, J. P. Bonilla Ataides, G. Baranes, A. Gu, T. Manovitz, M. Xu, M. Kalinowski, S. Majidy, C. Kokail, N. Maskara, E. C. Trapp, L. M. Stewart, S. Hollerith, H. Zhou, M. J. Gul- lans, S. F. Yelin, M. Greiner, V. Vuleti´ c, M. Cain, and M. D. Lukin, A fault-tolerant neutral-atom architecture for universal quantu...

    2025

  17. [17]

    N.-C. Chiu, E. C. Trapp, J. Guo, M. H. Abobeih, L. M. Stewart, S. Hollerith, P. L. Stroganov, M. Kalinowski, A. A. Geim, S. J. Evered, S. H. Li, X. Lyu, L. M. Peters, D. Bluvstein, T. T. Wang, M. Greiner, V. Vuleti´ c, and M. D. Lukin, Continuous operation of a coherent 3,000- qubit system, Nature646, 1075 (2025)

    2025

  18. [18]

    Scholl, A

    P. Scholl, A. L. Shaw, R. B.-S. Tsai, R. Finkelstein, J. Choi, and M. Endres, Erasure conversion in a high- fidelity rydberg quantum simulator, Nature622, 273 (2023)

    2023

  19. [19]

    S. Ma, G. Liu, P. Peng, B. Zhang, S. Jandura, J. Claes, A. P. Burgers, G. Pupillo, S. Puri, and J. D. Thompson, High-fidelity gates and mid-circuit erasure conversion in an atomic qubit, Nature622, 279 (2023)

    2023

  20. [20]

    A. G. Fowler, Coping with qubit leakage in topological codes, arXiv preprint arXiv:1308.6642 (2013)

    work page arXiv 2013

  21. [21]

    K. C. Miao, M. McEwen, J. Atalaya, D. Kafri, L. P. 7 Pryadko, A. Bengtsson, A. Opremcak, K. J. Satzinger, Z. Chen, P. V. Klimov, C. Quintana, R. Acharya, K. An- derson, M. Ansmann, F. Arute, K. Arya, A. Asfaw, J. C. Bardin, A. Bourassa, J. Bovaird, L. Brill, B. B. Buckley, D. A. Buell, T. Burger, B. Burkett, N. Bush- nell, J. Campero, B. Chiaro, R. Collin...

    2023

  22. [22]

    Perrin, S

    H. Perrin, S. Jandura, and G. Pupillo, Quantum error correction resilient against atom loss, Quantum9, 1884 (2025)

    2025

  23. [23]

    Baranes, M

    G. Baranes, M. Cain, J. P. B. Ataides, D. Bluvstein, J. Sinclair, V. Vuleti´ c, H. Zhou, and M. D. Lukin, Lever- aging qubit loss detection in fault-tolerant quantum al- gorithms, Physical Review X16, 011002 (2026)

    2026

  24. [24]

    Ghosh, A

    J. Ghosh, A. G. Fowler, J. M. Martinis, and M. R. Geller, Understanding the effects of leakage in superconducting quantum-error-detection circuits, Physical Review A88, 062329 (2013)

    2013

  25. [25]

    B. M. Varbanov, F. Battistel, B. M. Tarasinski, V. P. Ostroukh, T. E. O’Brien, L. DiCarlo, and B. M. Terhal, Leakage detection for a transmon-based surface code, npj Quantum Information6, 102 (2020)

    2020

  26. [26]

    Torlai and R

    G. Torlai and R. G. Melko, Neural decoder for topological codes, Physical review letters119, 030501 (2017)

    2017

  27. [27]

    Varsamopoulos, B

    S. Varsamopoulos, B. Criger, and K. Bertels, Decoding small surface codes with feedforward neural networks, Quantum Science and Technology3, 015004 (2017)

    2017

  28. [28]

    Bausch, A

    J. Bausch, A. W. Senior, F. J. H. Heras, T. Edlich, A. Davies, M. Newman, C. Jones, K. Satzinger, M. Y. Niu, S. Blackwell, G. Holland, D. Kafri, J. Atalaya, C. Gidney, D. Hassabis, S. Boixo, H. Neven, and P. Kohli, Learning high-accuracy error decoding for quantum pro- cessors, Nature635, 834 (2024)

    2024

  29. [29]

    A. W. Senior, T. Edlich, F. J. H. Heras, L. M. Zhang, O. Higgott, J. S. Spencer, T. Applebaum, S. Black- well, J. Ledford, A. ˇZemgulyt˙ e, A. ˇZ´ ıdek, N. Shutty, A. Cowie, Y. Li, G. Holland, P. Brooks, C. Beattie, M. Newman, A. Davies, C. Jones, S. Boixo, H. Neven, P. Kohli, and J. Bausch, A scalable and real-time neural decoder for topological quantum ...

    work page arXiv 2025

  30. [30]

    G. Hu, W. Ouyang, C.-Y. Lu, C. Lin, and H.-S. Zhong, Efficient and universal neural-network decoder for stabilizer-based quantum error correction, arXiv preprint arXiv:2502.19971 (2025)

    work page arXiv 2025

  31. [31]

    A. S. Maan and A. Paler, Machine learning message- passing for the scalable decoding of qldpc codes, npj Quantum Information11, 78 (2025)

    2025

  32. [32]

    J. Blue, H. Avlani, Z. He, L. Ziyin, and I. L. Chuang, Ma- chine learning decoding of circuit-level noise for bivariate bicycle codes, arXiv preprint arXiv:2504.13043 (2025)

    work page arXiv 2025

  33. [33]

    H. Cao, F. Pan, D. Feng, Y. Wang, and P. Zhang, Gener- ative decoding for quantum error-correcting codes, arXiv preprint arXiv:2503.21374 (2025)

    work page arXiv 2025

  34. [34]

    Zhang, Z

    K. Zhang, Z. Yi, S. Guo, L. Kong, S. Wang, X. Zhan, T. He, W. Lin, T. Jiang, D. Gao, Y. Zhang, F. Liu, F. Zhang, Z. Ji, F. Chen, and J. Chen, Learning to decode in parallel: Self-coordinating neural network for real-time quantum error correction, arXiv preprint arXiv:2601.09921 (2026)

    work page arXiv 2026

  35. [35]

    S. B. Bravyi and A. Y. Kitaev, Quantum codes on a lattice with boundary, arXiv preprint quant-ph/9811052 (1998)

    work page Pith review arXiv 1998

  36. [36]

    Dennis, A

    E. Dennis, A. Kitaev, A. Landahl, and J. Preskill, Topological quantum memory, Journal of Mathematical Physics43, 4452 (2002)

    2002

  37. [37]

    A. G. Fowler, M. Mariantoni, J. M. Martinis, and A. N. Cleland, Surface codes: Towards practical large-scale quantum computation, Physical Review A—Atomic, Molecular, and Optical Physics86, 032324 (2012)

    2012

  38. [38]

    Gidney, Stim: a fast stabilizer circuit simulator, Quan- tum5, 497 (2021)

    C. Gidney, Stim: a fast stabilizer circuit simulator, Quan- tum5, 497 (2021)

    2021

  39. [39]

    Bian Que was a famous doctor in ancient China who excelled at diagnosing a person’s illness through the non- invasive method of observing, listening, questioning, and pulse-taking, which is very similar to inferring errors through stabilizer measurements