Pith. sign in

REVIEW 3 major objections 5 minor 36 references

Reshaping quantum annealing landscapes with diagonal catalysts

T0 review · 3 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a two-body diagonal catalyst built solely from the coupling signs and magnitudes of an Ising optimization problem—with no knowledge of its solution—can reshape the annealing landscape so that finite-time quantum annea

desk verdict The shell-moment theorem and path-based catalyst construction are genuinely new and worth refereeing, but the headline gains rest on a single fixed Trotter step that the authors themselves leave unchecked. read the letter →

arxiv 2607.14063 v1 pith:6DP7SVHX submitted 2026-07-15 quant-ph

Andrés N. Cáliz , Carlos Ramon-Escandell , Finnley Paolella , Josep Bosch , Jan Nogué , Arnau Riera , Jordi Riu This is my paper · ORCID
classification quant-ph
keywords quantumannealingdiagonalcatalystsHammingshellsIsingspinglassZZcouplingspathsignpropagationshell-momenttheoremapproximateoptimization
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

The paper tries to establish that a solution-blind, diagonal 'catalyst'—a set of extra ZZ couplings computed from the problem's own coupling graph—can make quantum annealing concentrate its final states near the true solution at short sweep times. The key move is a shell-moment theorem: for any two-body Ising cost, the average energy on the Hamming shell at distance d from the solution is a fixed parabola in d, and a diagonal catalyst can only rescale that parabola while independently shrinking its width. Using that lens, the authors build catalysts by propagating signs along self-avoiding paths in the coupling graph, a procedure that uses no information about the answer. On 200 random 3-regular 20-spin instances at sweep time T=6, the median probability within 95% of the ground-state energy rises from 0.067 to 0.324, and the probability within two spin flips of a ground state rises from 0.10 to 0.35; gains persist, weakened, on fully connected models. A sympathetic reader would care because it offers a hardware-native, purely diagonal way to improve approximate optimization on annealers without knowing the solution in advance.

What carries the argument

The load-bearing objects are Hamming shells S_{d,z*}—the sets of configurations at distance d from a reference solution z*—and the exact shell-moment identities (Theorem 1), which give the mean and variance of any field-free two-body Ising operator over a shell. These identities decompose the landscape into a fixed parabolic funnel (μ_2(d) H(z*)) plus within-shell fluctuations; the catalyst's job is to deepen the funnel and narrow the fluctuations. The construction uses open-path sign propagation: along each m-edge self-avoiding path, signs are propagated according to g_{v_{a+1}} = −sgn(J_{v_a v_{a+1}}) g_{v_a}, which always yields a locally satisfying pattern; weighting each vertex by its i

What would settle it

Compute, on the 200 3-regular n=20 instances, the median squared alignment |2φ_p − 1|^2 for the order-4 paths used in the catalyst; if it is at or below the random baseline 1/(m+1), the catalyst has no meaningful estimate of the ground state. Separately, rerun the T=6 protocol at Trotter steps 0.05 and 0.025; if the per-instance paired gains move by more than the quoted interquartile ranges, the reported dynamics are not converged.

Watch

Extended reading notes

Core claim

The central claim is that the energy-versus-Hamming-distance landscape of an Ising problem can be reshaped by a two-body diagonal catalyst assembled from the problem couplings alone, and that this reshaping translates into a many-fold increase in the probability that a finite-time anneal lands at or near the ground state. Theorem 1 pins down what is possible: the shell-mean energy is exactly μ_2(d) times the ground-state energy, with μ_2(d) purely combinatorial, so a diagonal catalyst controls only the overall funnel steepness, while the shell variance is controlled separately. The constructive step propagates signs along self-avoiding paths—each path's propagated pattern is an exact ground

Load-bearing premise

The entire construction leans on the hope that sign-propagated patterns on open paths track the true ground-state alignment better than random (the fidelity condition of Appendix B), and on the numerical simulations' fixed Trotter step Δt=0.1 with no reported step-size convergence check; if either gives way, the reported funneling gains are not established.

Editorial extensions

If this is right

  • On sparse 3-regular instances, the near-solution probability mass (within two spin flips of the ground state) at T=6 rises from a median of 0.10 to 0.35, a +239% paired gain on 94% of instances, so annealers become useful as near-optimal solvers even when they miss the exact ground state.
  • The catalyst outperforms the equal peak-coupling control C=H_P by +77% in top-band mass and +102% in near-solution mass at T=6, indicating the distribution of coupling strength matters, not just its total scale.
  • Gains persist, though weaker, on fully connected Sherrington–Kirkpatrick instances (+22% median top-band gain at T=2, with every instance improved in near-solution Hamming mass), so the geometric funneling mechanism is not limited to sparse graphs.
  • The path order m is a tunable design axis: the optimal order decreases with graph degree (m=4,3,2 for 3-,4-,5-regular graphs at longer sweeps), letting practitioners trade reach against pattern fidelity on the fly.
  • Because the final distribution is closer to the solution in Hamming distance even when the ground state is not sampled, the method composes naturally with local post-processing or reverse-annealing warm starts.

Reading between the lines

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

  • Editorial: the shell-moment theorem suggests a static screening tool the paper does not fully exploit—one could search over subgraph families beyond paths (stars, trees, small frustrated loops) for the family that minimizes the scale-invariant shell-overlap metric ov_d before running any dynamics, making catalyst selection a precomputed property of the coupling matrix.
  • Editorial: the near-solution Hamming concentration that persists on fully connected models at all sweep times points to a mechanism distinct from spectral-gap widening; a direct test would compare the instantaneous gap spectra of catalyzed and uncatalyzed protocols on the same instances, which the paper reports only indirectly through final distributions.
  • Editorial: as an explicit testable extension, the same catalyst could be inserted into QAOA as an additional diagonal phase separator—something the paper lists as a possible direction—and one could benchmark fixed-depth QAOA with and without the catalyst on 3-regular MaxCut instances; because the catalyst is built from J alone, this is a drop-in modification.
  • Editorial: the construction's reliance on a fixed Trotter step of 0.1 in the numerical dynamics is the kind of detail a convergence study could settle; if the gains persist at Δt=0.05 and 0.025, the claim becomes robust to discretization.
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. The paper introduces a mathematical framework, built on Hamming shells, for understanding the energy–distance structure of field-free Ising QUBOs. Theorem 1 gives exact shell means and variances for any two-body Ising operator. Using this framework, the authors propose a diagonal ZZ-catalyst constructed only from the coupling matrix J via sign propagation along self-avoiding paths (Eq. (14)), with no information about the ground state. They test the catalyst in state-vector simulations on 200 instances each of 3-, 4-, and 5-regular graphs and on Sherrington–Kirkpatrick models at n=20, reporting large improvements in top-band energy mass and near-solution Hamming mass at sweep times T=2,6,10, relative both to the bare anneal and to an equal-peak-coupling baseline C=H_P.

Significance. If the numerical claims hold, this is a valuable result: the catalyst is two-local, diagonal, solution-free, and implementable with controls native to current annealers. Theorem 1 is exact and its proof in Appendix A is clean and parameter-free. The benchmark design is careful in several respects: paired per-instance comparisons, an equal-budget control, bootstrap confidence intervals, and a public data repository. The central dynamic claim, however, is currently supported only by fixed-step Trotter simulations with no convergence check. In addition, the per-family path order and the omission of path-edge contributions appear to be chosen after seeing results, which introduces a risk of overfitting in the reported gains. These issues are fixable but are load-bearing for the paper's central claim.

major comments (3)
  1. [Supplemental Material, 'Numerical Dynamics and Normalization'] All quantitative results, including the headline '+351% median top-band mass' and '99% improved' at T=6, are produced with a fixed first-order Trotter step Δt=0.1. The Supplemental explicitly states that no adaptive ODE tolerance is used and that 'the same instances can be rerun at smaller Δt for step-size checks,' but no such check is reported. With n=20 and peak-normalized O(n^2) ZZ terms, the product-formula error is not negligible by construction, and diabatic annealing distributions can be sensitive to small perturbations. I request a convergence study (e.g., Δt=0.05, 0.02, 0.01) on at least a subset of instances for the main comparisons, reporting how the median paired gains and improved fractions change. Without this, the quantitative claims are not yet supported.
  2. [Main text, Results and Eq. (14)] The path order m is fixed per family (m=4,3,2) because 'the order that maximizes the median paired gain ... decreases with degree.' This is a post hoc choice made after inspecting the simulation results. Moreover, Eq. (14) omits path-edge contributions because this 'empirically improves funneling.' These choices are not derived from Theorem 1 or from any prespecified rule. As reported, the gains may overstate the performance of a predetermined catalyst. The authors should either provide a predictive rule for choosing m from J alone, report results for all orders for each family, or separate order selection from evaluation using a held-out design.
  3. [Appendix B and Eq. (12)] The construction's starting assumption is that sign propagation along paths gives patterns g^(p) that track the global ground state. Appendix B shows that a path contributes to the funnel only when (2φ_p−1)^2 > 1/(m+1), and that for independent random agreement the expected value equals the threshold. The actual fidelity distribution φ_p on the benchmark instances is never reported. Without evidence that the path patterns are sufficiently aligned, the empirical gains could in principle arise from the added density of a catalyst whose patterns are not tracking the solution. I ask for the distribution of |2φ_p−1| on the tested instances, and ideally a randomized-pattern control with the same weights and support, to confirm that the structure, not merely the added couplings, is responsible for the gains.
minor comments (5)
  1. [Introduction, Eq. (3)] The transverse-field scale Γ is introduced informally. Please state its relation to the driver coefficient in Eq. (2), e.g., Γ=1 there, or define it through the Hamiltonian norm.
  2. [Fig. 1] The caption mentions 'seed H = 4' but 'seed' is not defined. Please clarify or remove the term.
  3. [Abstract] The phrase 'ground-state patterns of small frustration-free subproblems' is slightly misleading: a path is a tree and hence is automatically frustration-free. Consider saying 'small tree subproblems' or 'sign patterns on tree subgraphs.'
  4. [Supplemental Material, 'Relative-Improvement Statistics'] Instances with zero baseline probability are omitted from ratio summaries. This can bias the reported improved fraction and median gain when the baseline is occasionally zero. Please report how many instances are omitted for each protocol and threshold.
  5. [Appendix B, Eq. (B2)] The sign convention in (B2) is easy to misread. Please spell out the expansion of −M_p^2 after discarding the constant, so that the reader can verify that a negative C_id(z*) corresponds to deepening the funnel.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the catalyst is constructed from J alone and validated against independently computed final-state metrics.

full rationale

The derivation chain is not circular. Theorem 1 (Eqs. 7-10) is an exact combinatorial identity proven in Appendix A from hypergeometric shell averages; it is not used to define the success metrics. The catalyst C^(m) is built by Eq. (12) sign propagation on open paths and Eq. (14), using only coupling signs, weights, and graph structure; the text explicitly states 'the catalyst is constructed from J alone', while ground-state sets are 'used only for diagnostics, through Hamming distances and final-state binning'. The reported gains in P0.95 and P(delta<=2) come from state-vector simulations with the final Hamiltonian unchanged (Eq. 15), so no fitted parameter is renamed as a prediction. The heuristic assumption that path patterns track the global ground state is openly framed as an empirical trade-off (Appendix B), not as a derived consequence. The only self-citations, [32] and [36], concern an extension and the simulation tool, and are not load-bearing for the central claim. Two non-circular limitations exist: the path order m=4,3,2 is chosen after observing per-family gains ('The order that maximizes the median paired gain in top-band mass at the longer sweep times (T=6,10) decreases with degree'), which is post-selection rather than derivation-by-definition; and the fixed Trotter step dt=0.1 is acknowledged with reruns possible but no convergence study is reported ('the same instances can be rerun at smaller dt for step-size checks'). These affect reliability and generalization, not circularity. The paper is self-contained against the quantities it claims to improve, so the circularity score is 0.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a few explicit assumptions: exact shell moments, path-fidelity heuristic, Trotter propagation, and instance representativeness. No new particles, forces, dimensions, or conservation laws are introduced; the ZZ catalysts are ordinary two-body diagonal Hamiltonians. The main selected/fitted object is the path order m, plus reporting thresholds.

free parameters (2)
  • path order m per graph family = m=4 (3-regular), m=3 (4-regular), m=2 (5-regular)
    Main comparisons use a degree-specific order selected after scanning orders 2–5 on the same 200-instance ensembles. All orders are shown, but the headline gains correspond to the selected order.
  • diagnostic thresholds = ρ0=0.95 (also 0.90/0.98); δ≤2
    Reported gains are threshold-dependent; the authors show robustness across ρ0 values and for δ≤2, but the headline numbers use these specific thresholds.
assumptions (5)
  • domain assumption Trotter product formula with Δt=0.1 accurately approximates continuous-time Schrödinger evolution
    Load-bearing for all dynamic results; no convergence study is reported (Supplemental Material, Numerical Dynamics and Normalization).
  • ad hoc to paper Path sign-propagation g^(p) from Eq. (12) is a good proxy for global ground-state alignment on the full graph
    Core heuristic of the catalyst construction; Appendix B gives only a statistical order–fidelity trade-off, not an instance-level guarantee.
  • ad hoc to paper Omitting path-edge contributions improves funneling
    Stated in the main text after Eq. (13) as an empirical observation, with no theoretical account.
  • domain assumption 200 random instances at n=20 per graph family are representative of the relevant QA regime
    All quantitative claims are ensemble statistics on this size and sample count.
  • domain assumption Adiabatic/diabatic transition phenomenology, including Eq. (3) for avoided-crossing gaps
    Used to motivate the Hamming-shell competition picture; not directly tested in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reshaping quantum annealing landscapes with diagonal catalysts." pith.science (2026). https://pith.science/paper/6DP7SVHX

@misc{pith2026260714063,
  author       = {Pith},
  title        = {Pith review of: Reshaping quantum annealing landscapes with diagonal catalysts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6DP7SVHX}},
  note         = {Machine review of arXiv:2607.14063}
}
read the original abstract

Quantum annealing is often limited by population trapped in local minima many spin flips from the solution. We introduce a mathematical framework to understand the connection between energy and Hamming distance in optimization problems. Using this, we build ZZ-catalysts from ground-state patterns of small frustration-free subproblems that make configurations far from the solution less energetically competitive. On sparse problems they multiply the near-solution probability at short sweeps, with gains persisting on fully-connected models and tunable via subproblem choice.

Figures

Figures reproduced from arXiv: 2607.14063 by the authors.

Figure 1
Figure 1. (a). The shell means are therefore correctly or￾dered for every instance, and any overlap between shells arises from their within-shell spread. Applying Eq. (7) to the combined two-body operator HP + C gives ⟨HP + C⟩d,z⋆ = µ2(d) [HP (z ⋆ ) + C(z ⋆ )] . (11) The theorem thus delimits what a two-body diagonal cat￾alyst can and cannot change. The shape of the shell-mean curve is fixed by µ2(d), while the catalyst contr… view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 7 linked inside Pith

  1. [1]

    Kadowaki and H

    T. Kadowaki and H. Nishimori, Quantum annealing in the transverse Ising model, Phys. Rev. E58, 5355 (1998)

    1998

  2. [2]

    Farhi, J

    E. Farhi, J. Goldstone, S. Gutmann, and M. Sipser, Quantum computation by adiabatic evolution (2000), arXiv:quant-ph/0001106

    arXiv 2000

  3. [3]

    G. E. Santoro, R. Martoˇ n´ ak, E. Tosatti, and R. Car, The- ory of quantum annealing of an Ising spin glass, Science 295, 2427 (2002)

    2002

  4. [4]

    Albash and D

    T. Albash and D. A. Lidar, Adiabatic quantum compu- tation, Rev. Mod. Phys.90, 015002 (2018)

    2018

  5. [5]

    Hauke, H

    P. Hauke, H. G. Katzgraber, W. Lechner, H. Nishi- mori, and W. D. Oliver, Perspectives of quantum anneal- ing: methods and implementations, Rep. Prog. Phys.83, 054401 (2020)

    2020

  6. [6]

    Lucas, Ising formulations of many NP problems, Front

    A. Lucas, Ising formulations of many NP problems, Front. Phys.2, 5 (2014)

    2014

  7. [7]

    Jansen, M.-B

    S. Jansen, M.-B. Ruskai, and R. Seiler, Bounds for the adiabatic approximation with applications to quan- tum computation, Journal of Mathematical Physics48, 102111 (2007)

    2007

  8. [8]

    J¨ org, F

    T. J¨ org, F. Krzakala, J. Kurchan, and A. C. Maggs, Sim- ple glass models and their quantum annealing, Phys. Rev. Lett.101, 147204 (2008)

    2008

Show all 36 references
  1. [9]

    A. P. Young, S. Knysh, and V. N. Smelyanskiy, First- order phase transition in the quantum adiabatic algo- rithm, Phys. Rev. Lett.104, 020502 (2010)

    2010

  2. [10]

    Altshuler, H

    B. Altshuler, H. Krovi, and J. Roland, Anderson local- ization makes adiabatic quantum optimization fail, Proc. Natl. Acad. Sci. U.S.A.107, 12446 (2010)

    2010

  3. [11]

    M. H. S. Amin, Effect of local minima on adiabatic quan- tum optimization, Phys. Rev. Lett.100, 130503 (2008)

    2008

  4. [12]

    M. V. Berry, Transitionless quantum driving, J. Phys. A: Math. Theor.42, 365303 (2009)

    2009

  5. [13]

    Gu´ ery-Odelin, A

    D. Gu´ ery-Odelin, A. Ruschhaupt, A. Kiely, E. Tor- rontegui, S. Mart ´ ınez-Garaot, and J. G. Muga, Short- cuts to adiabaticity: Concepts, methods, and applica- tions, Rev. Mod. Phys.91, 045001 (2019)

    2019

  6. [14]

    Sels and A

    D. Sels and A. Polkovnikov, Minimizing irreversible losses in quantum systems by local counterdiabatic driving, Proc. Natl. Acad. Sci. U.S.A.114, E3909 (2017)

    2017

  7. [15]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, Quantum adi- abatic evolution algorithms with different paths (2002), arXiv:quant-ph/0208135

    2002 arXiv

  8. [16]

    Crosson, T

    E. Crosson, T. Albash, I. Hen, and A. P. Young, De- signing Hamiltonians for quantum adiabatic optimiza- tion, Quantum4, 334 (2020)

    2020

  9. [17]

    Seki and H

    Y. Seki and H. Nishimori, Quantum annealing with an- tiferromagnetic fluctuations, Phys. Rev. E85, 051112 (2012)

    2012

  10. [18]

    Hormozi, E

    L. Hormozi, E. W. Brown, G. Carleo, and M. Troyer, 6 Nonstoquastic Hamiltonians and quantum annealing of an Ising spin glass, Phys. Rev. B95, 184416 (2017)

    2017

  11. [19]

    Albash, Role of nonstoquastic catalysts in quantum adiabatic optimization, Phys

    T. Albash, Role of nonstoquastic catalysts in quantum adiabatic optimization, Phys. Rev. A99, 042334 (2019)

    2019

  12. [20]

    Takada, S

    K. Takada, S. Sota, S. Yunoki, B. Pokharel, H. Nishimori, and D. A. Lidar, Phase transitions in the frustrated Ising ladder with stoquastic and nonstoquastic catalysts, Phys. Rev. Research3, 043013 (2021)

    2021

  13. [21]

    Feinstein, L

    N. Feinstein, L. Fry-Bouriaux, S. Bose, and P. A. War- burton, Effects ofXXcatalysts on quantum annealing spectra with perturbative crossings, Phys. Rev. A110, 042609 (2024)

    2024

  14. [22]

    Ghosh, L

    R. Ghosh, L. A. Nutricati, N. Feinstein, P. A. Warburton, and S. Bose, Exponential speed-up of quantum annealing vian-local catalysts (2024), arXiv:2409.13029

    2024

  15. [23]

    L. A. Nutricati, R. Ghosh, N. Feinstein, S. Bose, and P. A. Warburton, Enhancing the energy gap of random graph problems viaXX-catalysts in quantum annealing (2024), arXiv:2409.16350

    2024 arXiv

  16. [24]

    R. J. Banks, N. Feinstein, R. Ghosh, S. Bose, and P. A. Warburton, Gadgets for simulating a non- nativeXXinteraction in quantum annealing (2025), arXiv:2503.16663

    2025 arXiv

  17. [25]

    Albash and M

    T. Albash and M. Kowalsky, Diagonal catalysts in quan- tum adiabatic optimization, Phys. Rev. A103, 022608 (2021)

    2021

  18. [26]

    Hattori and S

    T. Hattori and S. Tanaka, Improving the efficiency of quantum annealing with controlled diagonal catalysts, Phys. Rev. A113, 022433 (2026)

    2026

  19. [27]

    See Supplemental Material forReshaping quantum an- nealing landscapes with diagonal catalysts

  20. [28]

    M. W. Johnson, M. H. S. Amin, S. Gildert, T. Lanting, F. Hamze, N. Dickson, R. Harris, A. J. Berkley, J. Jo- hansson, P. Bunyk,et al., Quantum annealing with man- ufactured spins, Nature473, 194 (2011)

    2011

  21. [29]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, A quan- tum approximate optimization algorithm (2014), arXiv:1411.4028

    2014 arXiv

  22. [30]

    E. J. Crosson and D. A. Lidar, Prospects for quantum en- hancement with diabatic quantum annealing, Nat. Rev. Phys.3, 466 (2021)

    2021

  23. [31]

    Feinstein, I

    N. Feinstein, I. Shalashilin, S. Bose, and P. A. Warbur- ton, Robustness of diabatic enhancement in quantum an- nealing, Quantum Sci. Technol.10, 025011 (2025)

    2025

  24. [32]

    Werner, M

    M. Werner, M. Jonsson, A. Garc ´ ıa-S´ aez, A. Riera, and T. Albash, Leveraging Landau–Zener–St¨ uckelberg in- terference for accelerating diabatic quantum annealing (2026), arXiv:2606.09706

    2026 arXiv

  25. [33]

    Ohkuwa, H

    M. Ohkuwa, H. Nishimori, and D. A. Lidar, Reverse an- nealing for the fully connectedp-spin model, Physical Review A98, 022314 (2018), arXiv:1806.02542 [quant- ph]

    2018 arXiv

  26. [34]

    Munoz-Bauza and D

    H. Munoz-Bauza and D. A. Lidar, Scaling advantage in approximate optimization with quantum annealing, Phys. Rev. Lett.134, 160601 (2025)

    2025

  27. [35]

    Paw lowski, P

    J. Paw lowski, P. Tarasiuk, J. Tuziemski, L. Pawela, and B. Gardas, Toward quantum scaling advantage in approx- imate optimization, Phys. Rev. Appl.26, 014024 (2026)

    2026

  28. [36]

    Reshaping quantum annealing landscapes with diagonal catalysts

    V. Vasileiadis, A. Azzam, L. Mortimer, N. Padilla Sirera, and D. Arcos, QiliSDK (2026). Appendix A: Proof of Theorem 1 Proof.—Fix a reference configurationz ⋆ and define the agreement variablesx i =z iz⋆ i ∈ {±1}, together with the aligned couplings ˆWij =W ijz⋆ i z⋆ j . Any t...

    2026

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.