Paper: arXiv 2508.13557

Authors: Gabriele Agliardi, Dimitris Alevras, Vaibhaw Kumar, Roberto Lo Nardo, Gabriele Compostella, Sumit Kumar, Manuel Proissl, Bimal Mehta

Abstract

The efficient and effective construction of portfolios that adhere to real-world constraints is a challenging optimization task in finance. We investigate a concrete representation of the problem with a focus on design proposals of an Exchange Traded Fund. We evaluate the sampling-based CVaR Variational Quantum Algorithm (VQA), combined with a local-search post-processing, for solving problem instances that beyond a certain size become classically hard. We also propose a problem formulation that is suited for sampling-based VQA. Our utility-scale experiments on IBM Heron processors involve 109 qubits and up to 4200 gates, achieving a relative solution error of 0.49%. Results indicate that a combined quantum-classical workflow achieves better accuracy compared to purely classical local search, and that hard-to-simulate quantum circuits may lead to better convergence than simpler circuits. Our work paves the path to further explore portfolio construction with quantum computers.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces a novel quantum algorithm (CVaR-VQA) and modifies a quadratic optimization problem for quantum sampling, involving advanced quantum mechanics and non-trivial mathematical formulation. It is backed by utility-scale experiments on IBM Heron processors with 109 qubits, concrete error metrics (0.49% relative error), and a direct comparison to classical solvers (CPLEX), making it highly implementation-heavy and data-driven.

Research Flowchart

  flowchart TD
  A["Research Goal: Solve large-scale portfolio optimization with constraints using quantum-classical hybrid"] --> B["Data: ETF equity dataset; Real-world constraints"]
  B --> C["Method: Sampling-based CVaR VQA with local search post-processing"]
  C --> D{"Computational Process: Run on IBM Heron<br>109 qubits, 4200 gates"}
  D --> E["Outcomes: 0.49% relative error<br>Better accuracy than classical local search<br>Hard-to-simulate circuits improve convergence"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a sampling-based CVaR Variational Quantum Algorithm (VQA) combined with a classical local-search post‑processing workflow and a problem formulation that avoids qubit overhead for portfolio construction.

Source sentence
We evaluate the sampling-based CVaR Variational Quantum Algorithm (VQA), combined with a local-search post-processing, for solving problem instances that beyond a certain size become classically hard. We also propose a problem formulation that is suited for sampling-based VQA.

Assumes 6

  • Portfolio variables are integer numbers of bond lots (yi) for each candidate bond. model
    Source sentence
    PROBLEM FORMULATION In various industry-relevant scenarios, the problem variables are integer variables yi, where i = 1, ..., N spans all candidate bonds that may be included in the portfolio.
  • Constraints can be incorporated without slack variables, using a rescaling factor s that penalises violations. model
    Source sentence
    It should be noted that during the conversion of the problem to the binary unconstrained form, the inequality constraints are usually handled by the introduction of slack variables, which in turn leads to an increase in the number of binary variables. In this work, we propose an equivalent unconstrained form of the problem as min x ∈{0, 1}n xT Qx + 1T max{0, s ⊙(Ax −b)}, (4) where s ∈Rm is an appropriate rescaling
  • One qubit is used per bond, implying no qubit overhead for constraints. execution
    Source sentence
    II-A, it allows us to introduce problem constraints without any qubit overhead. As a consequence, we use only one qubit per bond.
  • CVaR parameter α is set in the range (0, 1] to improve convergence and robustness to hardware noise. model
    Source sentence
    When setting α in the range (0, 1], then CVaR is known to lead to better convergence in quantum optimization [6], [9] and shows to be more robust against hardware noise [10].
Show 2 more
  • Hardware noise is present but does not prevent the quantum samples from improving over iterations. execution
    Source sentence
    Our results indicate that despite the hardware noise, the raw samples continually improve over an increasing number of iterations and converge towards lower objective values.
  • Local‑search post‑processing can improve the quality of quantum‑generated samples. model
    Source sentence
    Recent works [14]–[16] have shown that a classical post-processing of candidate solutions found by the quantum algo-rithm, based on local search, can often improve the quality of solutions. In this work, we examine, in detail, the effect of local search on the samples obtained from the quantum scheme.

Data

No dataset named in the text.

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Utility‑scale experiments on IBM Heron processors with 109 qubits and up to 4200 gates achieve a relative solution error of 0.49 %. performance
    Source sentence
    Our utility-scale experiments on IBM Heron processors involve 109 qubits and up to 4200 gates, achieving a relative solution error of 0.49%.
  • A combined quantum‑classical workflow (CVaR‑VQA + local search) attains better accuracy than purely classical local search. performance
    Source sentence
    Results indicate that a combined quantum-classical workflow achieves better accuracy compared to purely classical local search, and that hard-to-simulate quantum circuits may lead to better convergence than simpler circuits. Our simulation-based study on problems involving 100+ qubits indicates that the local search alone achieves worse results than in combination with CVaR-VQA.
  • Local search alone yields worse results than when combined with CVaR‑VQA on problems involving 100+ qubits. performance
    Source sentence
    Our simulation-based study on problems involving 100+ qubits indicates that the local search alone achieves worse results than in combination with CVaR-VQA.
  • Hardware runs show that raw quantum samples improve over iterations and converge toward lower objective values despite noise. performance
    Source sentence
    Our results indicate that despite the hardware noise, the raw samples continually improve over an increasing number of iterations and converge towards lower objective values.
  • After local‑search post‑processing, the best optimization gap is 0.49 % (TwoLocal bilinear ansatz), followed by 0.55 % (TwoLocal color) and 0.91 % (BFCD color). performance
    Source sentence
    The best performance after local search is achieved by the TwoLocal bilinear ansatz (0.49% optimization gap), followed by the TwoLocal color (0.55%) and BFCD color (0.91%).

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 16 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.