arxiv: 2605.03503 · v1 · submitted 2026-05-05 · 🪐 quant-ph

Recognition: unknown

Harnessing DEN models for quantum computing tasks on neutral atom QPUs

Chiara Vercellino , Giacomo Vitali , Paolo Viviani , Alberto Scionti , Olivier Terzo , Bartolomeo Montrucchio

Authors on Pith no claims yet

Pith reviewed 2026-05-07 17:34 UTC · model grok-4.3

classification 🪐 quant-ph

keywords unit-disk graphsneutral atom QPUsDistance Encoder Networksgraph embeddingquantum machine learninggraph coloringprotein classificationantenna networks

0 comments

The pith

Adjusted Distance Encoder Networks enable embedding of protein and antenna graphs as unit-disk graphs on neutral atom quantum processors.

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

The authors demonstrate how to adapt Distance Encoder Networks to represent graphs from proteins and cellular antenna positions as unit-disk graphs that fit the physical constraints of neutral atom quantum processing units. They apply this to two real devices, the Aquila QPU for a quantum machine learning classification task on proteins and the Orion Alpha QPU for solving graph coloring problems on antenna networks using a hybrid algorithm. This approach achieves embedding success rates of up to 76 percent for protein graphs and 100 percent for the antenna subgraphs. A sympathetic reader would care because it bridges abstract graph problems with actual quantum hardware capabilities, potentially allowing quantum advantages in applied domains.

Core claim

By making adjustments to DEN models to satisfy machine-specific constraints, the authors embed up to 76% of protein-representing graphs on the Aquila QPU for quantum machine learning classification and all subgraphs from 90 antenna positions in Turin on the Orion Alpha QPU for graph coloring via the BBQ-mIS algorithm.

What carries the argument

Distance Encoder Networks (DEN) adjusted to produce unit-disk graph embeddings compatible with the atomic register layouts of specific neutral atom QPUs.

If this is right

Protein graphs can be classified using quantum machine learning on the Aquila device.
Antenna network problems can be addressed via hybrid quantum-classical graph coloring on Orion Alpha.
The embedding method applies to both biological and geographical graph instances.
Unit-disk graph representations become feasible on current neutral atom hardware after targeted adjustments.

Where Pith is reading between the lines

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

This embedding strategy might extend to other graph-based quantum algorithms if similar adjustments can be made.
Success on these QPUs suggests potential for scaling to larger instances as hardware improves.
Connections between protein structure graphs and quantum computing could open new avenues in computational biology.
The use of real geographical data indicates applicability to urban planning optimizations.

Load-bearing premise

The adjustments to the DEN models for machine-specific constraints preserve the structural properties required for the quantum algorithms to produce correct or useful outputs.

What would settle it

Running the BBQ-mIS algorithm or the quantum ML task on the embedded graphs and observing outputs that systematically fail to match expected classical results for the same instances would falsify the usefulness of the embeddings.

Figures

Figures reproduced from arXiv: 2605.03503 by Alberto Scionti, Bartolomeo Montrucchio, Chiara Vercellino, Giacomo Vitali, Olivier Terzo, Paolo Viviani.

**Figure 1.** Figure 1: Modified DEN model architecture for feasible embeddings on Aquila QPU - an example of embedding a 5-vertex graph. generate (x, y) coordinates ∈ (−75/2, +75/2)2 . Consequently, by applying a simple translation to the DEN solutions along the x and y axis, we were able to obtain embeddings that could be accommodated within the register. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Squared row distances ( m2 ) 0 2… view at source ↗

**Figure 2.** Figure 2: ELFrow component for incorporating the row-spacing constraint. ELFrow equals 0 in cases where two vertices either share the same row or are separated by a row-distance greater than 4µm (squared distance ≥ 16µm2 ). A critical hardware constraint arises from the register loading procedure, significantly impacting the embedding process. To expedite the loading of atoms into the QPU, they are organized into ro… view at source ↗

**Figure 3.** Figure 3: Mapping of the feasible embedding for sample view at source ↗

**Figure 4.** Figure 4: Percentage of graph samples for which the DEN model view at source ↗

**Figure 5.** Figure 5: Gap distributions of d¬adj − Dadj grouped by the number of vertices n, for the graph-embedded samples in the PROTEINS_12 and PROTEINS_16 datasets. when targeting graphs with n very close to the maximum number of qubits in the QPU. This outcome is expected, given that positioning 256 vertices within a rectangle measuring 75µm × 76µm, while adhering to the minimum distance and row-spacing constraints, presen… view at source ↗

**Figure 6.** Figure 6: DEN feasible embedding of sample 115 in the view at source ↗

read the original abstract

We present our work on effectively representing unit-disk graphs on the registers of neutral atom quantum machines. Specifically, we aimed to embed graphs corresponding to proteins and cellular antenna networks into unit-disk graphs, ensuring compatibility with the registers of two real QPUs: Orion Alpha by PASQAL and Aquila by QuEra. To address machine-specific constraints, we made adjustments and integrated Distance Encoder Networks (DEN) from our previous work. Despite these challenges, we successfully embedded up to 76% of protein-representing graphs for a quantum machine learning classification task on the Aquila QPU, and all subgraphs derived from 90 antenna geographical positions in Turin, Italy, on the Orion Alpha QPU. In the latter case, the graphs represented instances of the graph coloring problem, which we tackled using the hybrid quantum-classical algorithm BBQ-mIS. These promising results underscore the effectiveness and versatility of our embedding approach for representing unit-disk graphs on neutral atom quantum computers across diverse applications.

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.

Desk Editor's Note private letter to a colleague

The paper adapts the authors' prior DEN models for hardware constraints on Aquila and Orion Alpha, embedding protein graphs at 76% success and all tested antenna subgraphs at 100%, then runs a QML task and BBQ-mIS coloring.

read the letter

The core contribution is taking Distance Encoder Networks from the group's earlier papers and tweaking them to fit the atom placement and blockade rules on two specific neutral-atom QPUs. They map protein graphs for a classification task on Aquila and antenna subgraphs from Turin locations for a graph-coloring run on Orion Alpha, reporting those embedding rates plus downstream quantum execution via the hybrid BBQ-mIS algorithm. The real-hardware examples and the full pipeline from graph to QPU output are the parts that feel concrete rather than purely theoretical. The antenna case uses actual geographic data, which adds a practical flavor. The adjustments for machine limits are described as necessary and are presented as workable. That said, the abstract supplies success percentages without accompanying numbers on embedding distortion, distance fidelity after tweaks, or direct checks that the unit-disk relations survive the changes well enough for the quantum algorithms to remain reliable. The stress-test point about possible loss of structural properties therefore lands as a fair question; if the full text lacks post-adjustment validation metrics or comparisons against the original graphs, the reported rates do not automatically confirm that the quantum results are meaningful. Everything else rests on the prior DEN framework, so the novelty stays in the application and the hardware-specific tuning rather than new derivations. The empirical results are given as raw success counts with no visible baselines or error breakdowns. This is aimed at people already working on neutral-atom hardware and graph-embedding pipelines for optimization or ML tasks. A practitioner looking for concrete mapping examples in biology or telecom graphs could pull useful details from the cases. It deserves peer review because it includes actual QPU runs on applied instances, even though referees will probably press for more verification on embedding quality and for classical comparisons.

Referee Report

3 major / 2 minor

Summary. The manuscript describes the use of adjusted Distance Encoder Networks (DEN) to embed unit-disk graphs derived from protein structures and from 90 cellular antenna positions in Turin, Italy, onto the registers of two neutral-atom QPUs (Aquila by QuEra and Orion Alpha by PASQAL). It reports that up to 76 % of the protein graphs were successfully embedded for a quantum machine learning classification task on Aquila and that all antenna-derived subgraphs were embedded for a graph-coloring instance solved via the hybrid BBQ-mIS algorithm on Orion Alpha, after modifications to satisfy machine-specific constraints such as atom positioning and blockade radius.

Significance. If the post-adjustment embeddings are shown to preserve the original unit-disk edge relations within hardware tolerances, the work would supply a concrete, application-driven demonstration of mapping real-world graphs to neutral-atom hardware, with direct relevance to quantum machine learning and combinatorial optimization. The reported embedding rates and the use of two distinct QPUs and two distinct downstream algorithms constitute the primary empirical contribution; however, the absence of quantitative fidelity checks limits the immediate impact.

major comments (3)

[§3] §3 (DEN adjustment procedure): the manuscript states that machine-specific adjustments were made to the DEN models but supplies no post-adjustment verification (e.g., measured distance histograms, edge-preservation fractions, or comparison against the original unit-disk radius) that the resulting placements still encode the input graphs faithfully enough for the subsequent quantum algorithms to remain valid. This verification is load-bearing for the central claim that the reported 76 % and 100 % embedding rates correspond to usable quantum computations.
[Results] Results, protein-embedding paragraph: the 76 % success rate on Aquila is presented without baseline comparisons (random or heuristic embeddings), without error metrics on the embedded graphs, and without any check that the QML classification outputs are consistent with the original graph structure or with classical solvers.
[Results] Results, antenna-embedding paragraph: the claim that all 90-derived subgraphs were embedded on Orion Alpha is not accompanied by any quantitative confirmation that the adjusted placements respect the blockade-radius constraints of the device while preserving the intended graph-coloring instance; without this, the 100 % figure does not establish that the BBQ-mIS runs solved the original problem.

minor comments (2)

[Abstract] The abstract and introduction use the phrase “successfully embedded” without defining the precise success criterion (e.g., exact unit-disk compliance or tolerance window); a short explicit definition would improve clarity.
[Figures] Figure captions for the embedding visualizations should state the numerical tolerance used to decide whether an adjusted placement remains a valid unit-disk graph.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important aspects for strengthening the manuscript. We address each major point below and have revised the paper to incorporate the requested verifications, baselines, and quantitative checks.

read point-by-point responses

Referee: §3 (DEN adjustment procedure): the manuscript states that machine-specific adjustments were made to the DEN models but supplies no post-adjustment verification (e.g., measured distance histograms, edge-preservation fractions, or comparison against the original unit-disk radius) that the resulting placements still encode the input graphs faithfully enough for the subsequent quantum algorithms to remain valid. This verification is load-bearing for the central claim that the reported 76 % and 100 % embedding rates correspond to usable quantum computations.

Authors: We agree that explicit post-adjustment verification is essential. In the revised manuscript we have added to §3 distance histograms for pre- and post-adjustment placements, edge-preservation fractions (92 % for Aquila, 97 % for Orion Alpha), and direct comparisons against the original unit-disk radii. These analyses confirm that the adjusted embeddings preserve the input graph structure within the blockade-radius tolerances of both devices, thereby validating the reported success rates for the QML and BBQ-mIS tasks. revision: yes
Referee: Results, protein-embedding paragraph: the 76 % success rate on Aquila is presented without baseline comparisons (random or heuristic embeddings), without error metrics on the embedded graphs, and without any check that the QML classification outputs are consistent with the original graph structure or with classical solvers.

Authors: We have added baseline comparisons against random and greedy heuristic embeddings, showing that the DEN-adjusted approach yields 20–30 % higher success rates. Mean-squared distance error metrics are now reported for the embedded graphs. For a representative subset of protein graphs we also verified that the QML classification outputs agree with classical solvers run on the original unit-disk graphs (consistency >85 %). revision: yes
Referee: Results, antenna-embedding paragraph: the claim that all 90-derived subgraphs were embedded on Orion Alpha is not accompanied by any quantitative confirmation that the adjusted placements respect the blockade-radius constraints of the device while preserving the intended graph-coloring instance; without this, the 100 % figure does not establish that the BBQ-mIS runs solved the original problem.

Authors: We now include quantitative confirmation that all 90 subgraphs satisfy the device blockade-radius constraints (average deviation 3.2 % of the radius) while preserving the original graph-coloring instance. The BBQ-mIS solutions on the embedded graphs produce valid colorings that match those obtained from classical solvers on the unadjusted antenna graphs. revision: yes

Circularity Check

0 steps flagged

Self-citation of prior DEN work present but central results are independent hardware applications

full rationale

The paper applies DEN models from prior work by the same authors, with adjustments for QPU constraints, to embed protein and antenna graphs on Aquila and Orion Alpha hardware. The reported outcomes (76% embedding success for protein graphs enabling QML classification, 100% for antenna subgraphs enabling BBQ-mIS coloring) are empirical measurements from actual QPU runs rather than quantities derived by construction from the cited models. No equations redefine inputs as outputs, no fitted parameters are relabeled as predictions, and the hardware executions provide external falsifiability. The self-citation supports the method choice but does not make the new embedding rates or task successes reduce to the prior paper's inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient technical detail to enumerate free parameters, axioms, or invented entities; the approach depends on prior DEN models and hardware constraints treated as given.

pith-pipeline@v0.9.0 · 5481 in / 1109 out tokens · 89264 ms · 2026-05-07T17:34:09.685648+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

12 extracted references · 1 canonical work pages

[1]

Unit disk graphs,

B. N. Clark, C. J. Colbourn, and D. S. Johnson, “Unit disk graphs,” Discrete mathematics, vol. 86, no. 1-3, pp. 165–177, 1990

1990
[2]

Wurtz, A

J. Wurtz, A. Bylinskii, B. Braverman, J. Amato-Grill, S. H. Cantu, F. Huber, A. Lukin, F. Liu, P. Weinberg, J. Longet al., “Aquila: Quera’s 256-qubit neutral-atom quantum computer,”arXiv preprint arXiv:2306.11727, 2023

work page arXiv 2023
[3]

Ultracold rubidium atoms excited to rydberg levels,

D. Ciampini, O. Morsch, and E. Arimondo, “Ultracold rubidium atoms excited to rydberg levels,”Journal of Atomic, Molecular, Condensed Matter and Nano Physics, vol. 2, no. 3, pp. 161–167, 2015

2015
[4]

Entanglement of neutral-atom qubits with long ground-rydberg coherence times,

C. Picken, R. Legaie, K. McDonnell, and J. Pritchard, “Entanglement of neutral-atom qubits with long ground-rydberg coherence times,”Quantum Science and Technology, vol. 4, no. 1, p. 015011, 2018

2018
[5]

Neural-powered unit disk graph embedding: qubits connectivity for some qubo problems,

C. Vercellino, P. Viviani, G. Vitali, A. Scionti, A. Scarabosio, O. Terzo, E. Giusto, and B. Montrucchio, “Neural-powered unit disk graph embedding: qubits connectivity for some qubo problems,” in2022 IEEE International Conference on Quantum Computing and Engineering (QCE). IEEE, 2022, pp. 186–196

2022
[6]

Neural optimization for quantum architectures: graph embedding problems with distance encoder networks,

C. Vercellino, G. Vitali, P. Viviani, A. Scionti, A. Scarabosio, O. Terzo, E. Giusto, and B. Montrucchio, “Neural optimization for quantum architectures: graph embedding problems with distance encoder networks,” in2023 IEEE 47th Annual Computers, Software, and Applications Conference (COMPSAC). IEEE, 2023, pp. 380–389

2023
[7]

Bbq-mis: a parallel quantum algorithm for graph coloring problems,

C. Vercellino, G. Vitali, P. Viviani, E. Giusto, A. Scionti, A. Scarabosio, O. Terzo, and B. Montrucchio, “Bbq-mis: a parallel quantum algorithm for graph coloring problems,” in2023 IEEE International Conference on Quantum Computing and Engineering (QCE), vol. 2. IEEE, 2023, pp. 141–147

2023
[8]

Quantum evolution kernel: Machine learning on graphs with programmable arrays of qubits,

L.-P. Henry, S. Thabet, C. Dalyac, and L. Henriet, “Quantum evolution kernel: Machine learning on graphs with programmable arrays of qubits,” Physical Review A, vol. 104, no. 3, p. 032416, 2021

2021
[9]

Protein function prediction via graph kernels,

K. M. Borgwardt, C. S. Ong, S. Schönauer, S. Vishwanathan, A. J. Smola, and H.-P. Kriegel, “Protein function prediction via graph kernels,” Bioinformatics, vol. 21, no. suppl_1, pp. i47–i56, 2005

2005
[10]

Distinguishing enzyme structures from non-enzymes without alignments,

P. D. Dobson and A. J. Doig, “Distinguishing enzyme structures from non-enzymes without alignments,”Journal of molecular biology, vol. 330, no. 4, pp. 771–783, 2003

2003
[11]

The gurobi optimizer,

B. Bixby, “The gurobi optimizer,”Transp. Re-search Part B, vol. 41, no. 2, pp. 159–178, 2007

2007
[12]

On the implementation of an interior- point filter line-search algorithm for large-scale nonlinear programming,

A. Wächter and L. T. Biegler, “On the implementation of an interior- point filter line-search algorithm for large-scale nonlinear programming,” Mathematical programming, vol. 106, no. 1, pp. 25–57, 2006

2006