Where the Quantum Lives in D-Wave Hybrid Portfolio Optimization: An Operational Decomposition Audit

Luis Lozano

arxiv: 2605.17623 · v2 · pith:IEVGHJR3new · submitted 2026-05-17 · 🪐 quant-ph · math.OC· q-fin.PM

Where the Quantum Lives in D-Wave Hybrid Portfolio Optimization: An Operational Decomposition Audit

Luis Lozano This is my paper

Pith reviewed 2026-05-20 12:22 UTC · model grok-4.3

classification 🪐 quant-ph math.OCq-fin.PM

keywords hybrid quantum-classical optimizationportfolio optimizationconstraint-native solverspenalty encodingmean-variance optimizationquantum contribution auditdeterminism in solvers

0 comments

The pith

A hybrid quantum-classical portfolio optimizer derives its performance on constrained problems mostly from classical decomposition rather than quantum sampling.

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

The paper audits the actual quantum contribution inside a hybrid solver applied to cardinality-constrained mean-variance-turnover portfolio problems. It establishes that the solver reaches solutions matching those proven optimal by a classical mixed-integer benchmark on every instance where that benchmark finishes, yet the quantum processing unit receives only a small fraction of the allotted wall-clock time. The great majority of the effort occurs in classical steps that decompose the problem, assemble subproblems, and reassemble feasible solutions. Two further structural findings appear: penalty encodings of the cardinality constraint create a dense rank-one term that fully connects the logical graph regardless of the original covariance sparsity, and the constraint-native path returns exactly the same solution for every tested time budget and every repeated call. These observations together locate the reported advantage inside the classical pipeline.

Core claim

On cardinality-constrained mean-variance-turnover portfolio instances ranging from 10 to 640 assets, the constraint-native hybrid service reproduces the proven optima of a classical mixed-integer quadratic programming anchor on all instances where that anchor reaches proven optimality, while the quantum processing unit is accessed for only a negligible share of the total 5-second wall-clock budget. The remainder of the computation consists of classical decomposition, subproblem assembly, and feasibility-aware reassembly. Encoding the cardinality constraint through a penalty term adds a dense rank-one update that renders the logical graph fully connected irrespective of the original problem’s

What carries the argument

The constraint-native hybrid interface, which embeds constraints directly into the solver rather than converting them to penalty terms, thereby preserving sparsity and enabling a clean separation between classical pipeline steps and the small quantum contribution.

If this is right

Penalty-encoded paths lose the intended density benchmark axis because the rank-one cardinality term forces full connectivity.
The constraint-native service produces identical solutions at every wall-clock budget from 5 to 300 seconds and across repeated calls.
Hybrid performance claims on this problem class require explicit reporting of the classical-versus-quantum time split.
Direct comparisons against exact classical solvers and simulated annealing isolate the source of any reported advantage.

Where Pith is reading between the lines

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

Audits that separate classical pipeline time from quantum access time could be applied to other hybrid solvers used in finance and combinatorial optimization.
The observed determinism suggests the service may be driven by fixed classical heuristics rather than stochastic sampling, a distinction worth testing on different problem classes.
Benchmark design for quantum advantage should treat constraint-native and penalty-encoded routes as separate methodological families rather than interchangeable.

Load-bearing premise

The instances on which the classical benchmark proves optimality are representative of the full test collection and the fixed wall-clock budget measures practical performance without favoring any solver’s internal heuristics.

What would settle it

A larger collection of instances or a different time budget in which the quantum processing unit time fraction rises substantially while the service still matches or improves upon the classical benchmark solutions would indicate a larger quantum role.

Figures

Figures reproduced from arXiv: 2605.17623 by Luis Lozano.

**Figure 2.** Figure 2: Formulation choice for D-Wave portfolio optimization. The constraint-native CQM [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: D-Wave hybrid solver architecture. The solver iterates between classical decomposition [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

**Figure 4.** Figure 4: (a) Mean chain-break fraction and (b) embedding overhead (physical / logical qubits) [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 5.** Figure 5: Chain-break fraction vs N split by covariance-density family (diagonal, block, dense), on Pegasus (left) and Zephyr (right). The three curves nearly overlap at each N, confirming that penalty encoding collapses the density axis [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

**Figure 6.** Figure 6: Mean objective value vs problem size across solver families (averaged over 3 density [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗

**Figure 7.** Figure 7: Relative gap to Gurobi optimal vs N. CQM (green) is indistinguishable from the optimal baseline; BQM (red) and SA (orange) diverge as penalty dilution intensifies. 5 Discussion 5.1 Formulation Choice Dominates Solver Choice On the tested instance families, the constraint-native CQM formulation produces lower objective values than the penalty-encoded BQM formulation at every N > 10, across all three density… view at source ↗

**Figure 8.** Figure 8: Direct QPU on real Fama–French 49 equity data: chain-break fraction vs [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

**Figure 9.** Figure 9: Budget response curves for hybrid BQM (dashed) and CQM (solid) across problem [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗

**Figure 10.** Figure 10: Stochastic validation: box plots of objective values across 10 repeated runs for CQM [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗

read the original abstract

We audit the operational decomposition of D-Wave's hybrid quantum-classical portfolio-optimization service on cardinality-constrained mean-variance-turnover instances spanning N=10 to 640, with the constraint-native LeapHybridCQM interface, the penalty-encoded LeapHybridBQM interface, and Gurobi MIQP and simulated-annealing classical anchors. We report all three SDK timing fields (t_run, t_charge, t_QPU) and define a candidate four-metric audit protocol for hybrid quantum-classical solvers. Three findings. First, the LeapHybridCQM service matches Gurobi's proven optimum on all 54 head-to-head instances at N <= 120, but the mean QPU access time is 0.034 seconds out of the 5-second nominal wall-clock budget -- 0.68% of the nominal budget, approximately 0.72% of measured run time -- and the remaining ~99% is the service's classical decomposition and feasibility-aware reassembly. Second, in a CPU-only matched-wall-clock counterfactual, TabuSampler on the penalty-encoded BQM reaches final exact-K objectives within mean absolute delta 0.001 of hybrid CQM on 24 tested instances; this does not ablate the LeapHybridCQM pipeline internals, but it shows that these objective levels are reproducible by a classical heuristic at the same wall-clock budget. Third, the cardinality penalty contributes a dense rank-one term that fully connects the encoded logical graph independent of the input covariance density, an effect we prove as a structural theorem; the resulting density-axis collapse explains the BQM degradation observed in the empirical comparison. Out-of-sample on Fama-French 49 industry portfolios, the QPU-selected portfolios deliver a mean Sharpe ratio of 1.94 versus 2.22 for the 1/N baseline. The practical implication is that reported D-Wave hybrid wins on this problem class are constraint-native classical pipelines, not quantum-sampling wins.

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

D-Wave hybrid portfolio results on these problems come from classical decomposition with only 0.7 percent QPU time.

read the letter

D-Wave's hybrid service on these cardinality-constrained portfolio problems delivers its results through classical decomposition and reassembly, with the QPU accounting for only about 0.7% of the 5-second budget. The paper backs this with direct measurements. On the 54 instances where Gurobi finds and proves the optimum, the hybrid matches it exactly. The mean QPU access time is 0.034 seconds, leaving the rest to the classical parts of the service. They also show that the service returns the same solution every time, even when the wall-clock budget changes from 5 to 300 seconds or when called repeatedly. The structural point about the rank-one penalty making the logical graph fully connected is a useful addition to the earlier distinction between constraint-native and penalty methods. These observations rest on straightforward comparisons to Gurobi MIQP and simulated annealing, without relying on any self-referential math. One limitation is that the 54 instances are those Gurobi can certify, which may not include the harder problems in the full set. On those, the relative contribution or the performance edge could look different. A single fixed 5-second budget also leaves room for the possibility that the classical heuristics are particularly effective in that window. This is worth reading for people who track quantum optimization benchmarks in finance. It gives a clearer view of what is actually being measured in hybrid runs. I recommend sending it for peer review. The empirical audit is grounded enough to be useful even if the authors need to expand on the harder instances.

Referee Report

1 major / 2 minor

Summary. The paper audits D-Wave's LeapHybridCQM hybrid solver on cardinality-constrained mean-variance-turnover portfolio problems (N=10 to 640). It reports that the service matches Gurobi MIQP proven optima on all 54 instances where Gurobi certifies optimality, yet mean QPU access time is only 0.034 s out of a 5 s wall-clock budget (~0.7 %). The remaining time is classical decomposition and reassembly; the authors conclude the observed performance is a constraint-native classical pipeline with minor QPU contribution rather than a quantum-sampling advantage. Supporting observations include the dense rank-one term induced by cardinality penalties (collapsing sparsity benchmarks for penalty-encoded paths) and deterministic identical solutions across budgets (5–300 s) and repeats. Comparisons to Gurobi, simulated annealing, and penalty-encoded hybrids are provided.

Significance. If the measurements hold, the work supplies concrete, reproducible anchors—0.034 s QPU time, 54/54 optimality matches, and identical solutions across budgets and 10 repeats—against Gurobi MIQP and simulated annealing baselines. These data usefully decompose hybrid performance and extend Sakuler et al.'s constraint-native versus penalty-encoded distinction with operational timing detail. The structural result on the rank-one penalty term and the determinism observation are clear strengths that ground the audit.

major comments (1)

[Experimental results and instance selection] The optimality-matching claim is restricted to the 54 instances where Gurobi proves optimality; the manuscript should either demonstrate that these instances are representative of the full test set or report solution quality (e.g., duality gaps or objective values relative to Gurobi bounds) on the complementary instances to support generalization to the broader problem class.

minor comments (2)

Exact Gurobi parameter settings (e.g., MIP gap tolerance, time limit) and simulated annealing configuration should be stated explicitly to permit full reproduction of the baselines.
The two structural results (rank-one penalty term and determinism) are mentioned in the abstract; labeling them as numbered subsections or theorems in the main text would improve navigation.

Simulated Author's Rebuttal

1 responses · 0 unresolved

Thank you for the detailed review and the helpful comment on our experimental results. We address the concern about instance selection and generalization below.

read point-by-point responses

Referee: [Experimental results and instance selection] The optimality-matching claim is restricted to the 54 instances where Gurobi proves optimality; the manuscript should either demonstrate that these instances are representative of the full test set or report solution quality (e.g., duality gaps or objective values relative to Gurobi bounds) on the complementary instances to support generalization to the broader problem class.

Authors: We agree that the optimality-matching results are presented only for the subset of instances where Gurobi certifies optimality. These 54 instances were selected because they allow direct verification against proven optima, which is central to our audit of solution quality. To demonstrate representativeness, we note that they cover the entire range of problem sizes (N = 10 to 640) and both sparse and dense covariance structures used in the full test set. For the complementary instances, where Gurobi does not prove optimality within the time limit, we will add to the revised manuscript a table reporting the objective values obtained by LeapHybridCQM alongside the best primal and dual bounds from Gurobi. This will allow readers to assess the solution quality relative to the available bounds and support broader generalization. We believe this addition addresses the concern without altering the core conclusions of the paper. revision: yes

Circularity Check

0 steps flagged

No circularity: central claims rest on direct wall-clock measurements and external optimality comparisons

full rationale

The paper's derivation consists of empirical observations: exact matching of Gurobi-proven optima on 54 instances, mean QPU access time of 0.034 s within a fixed 5 s budget (0.7 % quantum fraction), and identical solutions across budgets and repeats. These quantities are measured against external solvers and hardware interfaces rather than fitted to or defined by any internal parameter of the present work. The rank-one penalty term and determinism property are direct consequences of the problem encoding and service behavior, not self-referential equations. The citation to Sakuler et al. supplies prior context on constraint-native interfaces but is not invoked as a uniqueness theorem or load-bearing premise that forces the current decomposition; the operational audit is independently verified here against Gurobi, simulated annealing, and penalty-encoded baselines. The chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The audit relies on standard mixed-integer quadratic programming assumptions and the external optimality of Gurobi; no new free parameters, axioms, or invented entities are introduced beyond the problem formulation itself.

axioms (1)

domain assumption Gurobi MIQP provides proven global optima on the tested cardinality-constrained mean-variance-turnover instances
Invoked when declaring that the hybrid service matches the proven optimum on all 54 instances

pith-pipeline@v0.9.0 · 5843 in / 1390 out tokens · 40790 ms · 2026-05-20T12:22:20.557217+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

mean QPU access time is only 0.034 seconds out of a 5-second wall-clock budget, roughly 0.7 percent of the run. The remaining ∼99 percent is the service’s classical decomposition, sub-problem assembly, and feasibility-aware reassembly
IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

the cardinality penalty contributes a dense rank-one term that makes the penalty-encoded logical graph fully connected regardless of the original covariance density

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.