Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

IEEE International Conference on Quantum Software (QSW)

Noise or Signal? A Causal Mediation Analysis of Quantum Circuit Learning Under Real Hardware Noise

Cyrille YetuYetu Kesiku, Begonya García-Zapirain

Published Jul 1, 2026• 0 citations
Publisher

Abstract

We deploy a Quantum Counterfactual Mediation Analysis (QCMA) framework on IBM Quantum processors to ask a precise, hardware-grounded question: are causal regimes identified in noiseless simulation preserved under real quantum noise? Integrating partial quantum state tomography with a linear structural equation model, we decompose inter-architectural performance contrasts into direct effects and indirect quantum-mediated effects across twelve experiments spanning three circuit topologies and two benchmark datasets. Five of six topology-dataset pairs (83%) retain their causal regime when migrating from noiseless simulation to physical hardware. Decoherence inflates quantum-mediated contributions by up to one order of magnitude a noise artefact, not quantum advantage while the Neutral regime dominates both backend. These findings indicate that hardware noise may substantially limit the effective utilization of quantum resources in current NISQ circuits, and motivate further investigation into coherence-preserving circuit design strategies.

DOI
10.1109/qsw72780.2026.00027

Related papers

How Quantum Circuits Actually Learn: A Causal Identification of Genuine Quantum Contributions

Cyrille Yetuyetu Kesiku, Begonya Garcia-Zapirain

Attributing performance gains in quantum machine learning to genuine quantum resources rather than to classical architectural scaling remains an open methodological challenge. We address this by introducing a counterfactual causal mediation framework that decomposes inter-architectural performance differences into direct effects, attributable to circuit parameterization and expressivity, and indirect effects mediated by quantum information-theoretic observables: entanglement entropy, purity, linear entropy, and quantum mutual information. Applying this framework to five circuit topologies and three benchmark datasets (across 43 validated configurations) reveals that direct architectural contributions systematically exceed quantum-mediated effects, with a mean ratio of 13.1:1 and a mean indirect contribution of 0.82%. These results suggest that current variational quantum circuits operate substantially below their quantum potential, and that principled resource-aware circuit design represents a tractable path toward measurable quantum-mediated performance gains.

quant-pharXiv:2603.16321•Mar 18, 2026•3 citations

Comparing Simulated versus Hardware Training Data for Learning-Based Quantum Error Mitigation

Seyed Mohamad Ali Tousi, Guilherme N. DeSouza

Learning-based Quantum Error Mitigation (QEM) has emerged as a promising approach for reducing noise-induced bias on near-term Noisy Intermediate-Scale Quantum (NISQ) hardware. While simulated noise models enable scalable training of learning-based mitigators, real hardware data more accurately reflect device-specific error mechanisms, but are expensive and sparse. Understanding how these two data regimes interact is therefore critical for the practical deployment of learning-based QEM. In this paper, we present a systematic and empirical analysis on the nature of the data (simulated versus hardware noise) for learning-based QEM. Using a recently developed attention-based graph transformer for small- and large-scale quantum error mitigation, we compare various training regimes, starting from exclusively simulated noise and progressively integrating more measurements from real quantum hardware. Our study spans two distinct circuit families: 1) randomly generated circuits with relatively homogeneous structure; and 2) targeted quantum circuits derived from a specific problem, in this case the one-dimensional Burgers’ equation. Across five-fold cross validation, we analyze how hardware data affects generalization performance as measured by mean absolute error (MAE). Our results demonstrate that the impact of hardware data strongly depends on circuit structure and diversity. For random circuits, simulated data already guarantees lower absolute error, while hardware data yields consistent, but moderate improvements. In contrast, for targeted circuits, like the structured Burgers’ circuits, simulated noise can be less representative of the actual (testing) quantum error, and incorporating hardware data produces substantially larger relative reductions in MAE – despite the intrinsically higher MAE. These findings indicate the value of the nature of the training data in learning-based QEM, which is governed not only by the data’s ability to approximate reality – as expected – but also by how difficult it may be to capture circuit-induced error mechanisms in quantum simulations. In general, this work highlights the urgency of quantum circuit coverage in designing scalable and cost-effective learning-based QEM pipelines for devices in this new NISQ-era.

International Joint Conference on Neural Networks (IJCNN)•Jun 1, 2026•0 citations

Investigation of Hardware Architecture Effects on Quantum Algorithm Performance: A Comparative Hardware Study

Askar Oralkhan, Temirlan Zhaxalykov

Cloud-accessible quantum processors enable direct execution of quantum algorithms on heterogeneous hardware platforms. Unlike classical systems, however, identical quantum circuits may exhibit substantially different behavior across devices due to architectural variations in qubit connectivity, gate fidelity, and coherence times. In this work, we systematically benchmark five representative quantum algorithms - Bell state preparation, GHZ state generation, Quantum Fourier Transform (QFT), Grover's Search, and the Quantum Approximate Optimization Algorithm (QAOA) - across trapped-ion, superconducting, and simulator backends using Amazon Braket. Performance metrics including fidelity, CHSH violation, success probability, circuit depth, and gate counts are evaluated. Our results demonstrate a strong dependence of algorithmic performance on hardware topology and noise characteristics. For example, 10-qubit GHZ states achieved fidelities above 0.8 on trapped-ion hardware, while superconducting platforms dropped below 0.15 due to routing overhead and accumulated two-qubit gate errors. These findings highlight the importance of hardware-aware algorithm selection and provide practical guidance for benchmarking in the NISQ era.

quant-pharXiv:2601.05286•Jan 8, 2026•1 citation

Quantum Algorithm Performance Under Noise: A Qiskit Aer Simulation Study and IBM Quantum Hardware Comparative Analysis

Yuvraj Mehta

This study examines how noise affects the reliability of quantum circuits through two complementary approaches: a controlled simulation using Qiskit Aer and an execution study on real IBM Quantum hardware. In simulation, five noise models (bit-flip, phase-flip, depolarizing, readout, and thermal-relaxation) were applied to a parameterized benchmark circuit across varying qubit counts, circuit depths, and noise strengths, producing 6,000 experimental observations. Success probability declined substantially as qubit count and circuit depth increased, with bit-flip, phase-flip, and depolarizing noise degrading performance far more severely (mean success probabilities of 0.597, 0.580, and 0.607) than readout noise (0.917) or thermal-relaxation noise (0.982). The same benchmark circuit was then executed on the ibm_marrakesh processor across 150 hardware runs (150,000 shots). Hardware performance remained consistently high (mean success probability 0.992), declining with qubit count (0.998 at two qubits to 0.986 at six) but showing little sensitivity to circuit depth. These results indicate that controlled noise mechanisms and real hardware degrade circuit performance through different dynamics, with implications for interpreting simulated noise models relative to physical device behavior.

•Oct 4, 2026•0 citations

Suppressing quantum errors by noise-aware circuit design

Yi Hu, Congcong Zheng, Xiaojun Wang +3 more

Abstract Suppressing errors is one of the central challenges in achieving reliable quantum computation on near-term hardware. While much attention has been paid to error correction and mitigation, we identify quantum circuit structure itself as a powerful lever for proactive error suppression. In this work, we present a unified and hardware-adaptive framework for noise-aware quantum circuit design, in which circuit topology and parameters are co-optimized from the ground up based on the noise profile of the target backend. Our framework supports a wide range of quantum tasks-including circuit compilation, quantum state preparation, and unitary approximation-under a consistent optimization paradigm. Extensive experiments on five IBM backends confirm the effectiveness of our method, showing significant fidelity gains across all tasks under realistic noise. These results demonstrate that noise suppression through structure-aware design offers a powerful and generalizable strategy for enhancing circuit performance on NISQ hardware. Our framework bridges low-level hardware constraints with high-level circuit synthesis, paving the way for more robust and efficient quantum programming in the presence of noise.

Quantum Science and Technology•Sep 10, 2025•4 citations

Computational Modeling of Quantum Noise and Decoherence

Andreas Popescu, Elena Horvath, Eva Hansen

Quantum noise and decoherence are the primary barriers to realising fault-tolerant quantum computation on near-term hardware. Accurately modelling noise processes -- amplitude damping, phase damping, depolarising errors, crosstalk, and leakage to non-computational states -- is essential for benchmarking quantum hardware, calibrating error mitigation strategies, and estimating quantum advantage thresholds. This paper proposes the Quantum Noise Computational Model (QNCM) framework, a systematic evaluation of seven noise modelling approaches -- Lindblad master equation, Kraus operator formalism, stochastic Schrodinger equation, randomised benchmarking models, process tomography, Pauli noise models, and coherent error models -- across 24 quantum hardware calibration datasets from IBM Quantum, Google Sycamore, and IonQ Aria processors. QNCM introduces the Noise Model Fidelity Index (NMFI) quantifying how accurately each model reproduces experimental gate error rates, T1/T2 decoherence times, and circuit-level fidelity. Key results: Lindblad master equation achieves NMFI = 0.924 (highest accuracy); Pauli noise models achieve NMFI = 0.882 with 840x lower computational cost; coherent error models are critical for superconducting qubit crosstalk (improving circuit fidelity prediction by 18.4%). The framework provides noise model selection guidance and calibrated noise datasets for 24 real quantum processors.

Quantum Frontiers Journal•Jul 22, 2025•0 citations
We deploy a Quantum Counterfactual Mediation Analysis (QCMA) framework on IBM Quantum processors to ask a precise, hardware-grounded question: are causal regimes identified in noiseless simulation preserved under real quantum noise? Integrating partial quantum state tomography with a linear structural equation model, we decompose inter-architectural performance contrasts into direct effects and indirect quantum-mediated effects across twelve experiments spanning three circuit topologies and two benchmark datasets. Five of six topology-dataset pairs (83%) retain their causal regime when migrating from noiseless simulation to physical hardware. Decoherence inflates quantum-mediated contributions by up to one order of magnitude a noise artefact, not quantum advantage while the Neutral regime dominates both backend. These findings indicate that hardware noise may substantially limit the effective utilization of quantum resources in current NISQ circuits, and motivate further investigation into coherence-preserving circuit design strategies.
Attributing performance gains in quantum machine learning to genuine quantum resources rather than to classical architectural scaling remains an open methodological challenge. We address this by introducing a counterfactual causal mediation framework that decomposes inter-architectural performance differences into direct effects, attributable to circuit parameterization and expressivity, and indirect effects mediated by quantum information-theoretic observables: entanglement entropy, purity, linear entropy, and quantum mutual information. Applying this framework to five circuit topologies and three benchmark datasets (across 43 validated configurations) reveals that direct architectural contributions systematically exceed quantum-mediated effects, with a mean ratio of 13.1:1 and a mean indirect contribution of 0.82%. These results suggest that current variational quantum circuits operate substantially below their quantum potential, and that principled resource-aware circuit design represents a tractable path toward measurable quantum-mediated performance gains.
Learning-based Quantum Error Mitigation (QEM) has emerged as a promising approach for reducing noise-induced bias on near-term Noisy Intermediate-Scale Quantum (NISQ) hardware. While simulated noise models enable scalable training of learning-based mitigators, real hardware data more accurately reflect device-specific error mechanisms, but are expensive and sparse. Understanding how these two data regimes interact is therefore critical for the practical deployment of learning-based QEM. In this paper, we present a systematic and empirical analysis on the nature of the data (simulated versus hardware noise) for learning-based QEM. Using a recently developed attention-based graph transformer for small- and large-scale quantum error mitigation, we compare various training regimes, starting from exclusively simulated noise and progressively integrating more measurements from real quantum hardware. Our study spans two distinct circuit families: 1) randomly generated circuits with relatively homogeneous structure; and 2) targeted quantum circuits derived from a specific problem, in this case the one-dimensional Burgers’ equation. Across five-fold cross validation, we analyze how hardware data affects generalization performance as measured by mean absolute error (MAE). Our results demonstrate that the impact of hardware data strongly depends on circuit structure and diversity. For random circuits, simulated data already guarantees lower absolute error, while hardware data yields consistent, but moderate improvements. In contrast, for targeted circuits, like the structured Burgers’ circuits, simulated noise can be less representative of the actual (testing) quantum error, and incorporating hardware data produces substantially larger relative reductions in MAE – despite the intrinsically higher MAE. These findings indicate the value of the nature of the training data in learning-based QEM, which is governed not only by the data’s ability to approximate reality – as expected – but also by how difficult it may be to capture circuit-induced error mechanisms in quantum simulations. In general, this work highlights the urgency of quantum circuit coverage in designing scalable and cost-effective learning-based QEM pipelines for devices in this new NISQ-era.
Cloud-accessible quantum processors enable direct execution of quantum algorithms on heterogeneous hardware platforms. Unlike classical systems, however, identical quantum circuits may exhibit substantially different behavior across devices due to architectural variations in qubit connectivity, gate fidelity, and coherence times. In this work, we systematically benchmark five representative quantum algorithms - Bell state preparation, GHZ state generation, Quantum Fourier Transform (QFT), Grover's Search, and the Quantum Approximate Optimization Algorithm (QAOA) - across trapped-ion, superconducting, and simulator backends using Amazon Braket. Performance metrics including fidelity, CHSH violation, success probability, circuit depth, and gate counts are evaluated. Our results demonstrate a strong dependence of algorithmic performance on hardware topology and noise characteristics. For example, 10-qubit GHZ states achieved fidelities above 0.8 on trapped-ion hardware, while superconducting platforms dropped below 0.15 due to routing overhead and accumulated two-qubit gate errors. These findings highlight the importance of hardware-aware algorithm selection and provide practical guidance for benchmarking in the NISQ era.
This study examines how noise affects the reliability of quantum circuits through two complementary approaches: a controlled simulation using Qiskit Aer and an execution study on real IBM Quantum hardware. In simulation, five noise models (bit-flip, phase-flip, depolarizing, readout, and thermal-relaxation) were applied to a parameterized benchmark circuit across varying qubit counts, circuit depths, and noise strengths, producing 6,000 experimental observations. Success probability declined substantially as qubit count and circuit depth increased, with bit-flip, phase-flip, and depolarizing noise degrading performance far more severely (mean success probabilities of 0.597, 0.580, and 0.607) than readout noise (0.917) or thermal-relaxation noise (0.982). The same benchmark circuit was then executed on the ibm_marrakesh processor across 150 hardware runs (150,000 shots). Hardware performance remained consistently high (mean success probability 0.992), declining with qubit count (0.998 at two qubits to 0.986 at six) but showing little sensitivity to circuit depth. These results indicate that controlled noise mechanisms and real hardware degrade circuit performance through different dynamics, with implications for interpreting simulated noise models relative to physical device behavior.
Abstract Suppressing errors is one of the central challenges in achieving reliable quantum computation on near-term hardware. While much attention has been paid to error correction and mitigation, we identify quantum circuit structure itself as a powerful lever for proactive error suppression. In this work, we present a unified and hardware-adaptive framework for noise-aware quantum circuit design, in which circuit topology and parameters are co-optimized from the ground up based on the noise profile of the target backend. Our framework supports a wide range of quantum tasks-including circuit compilation, quantum state preparation, and unitary approximation-under a consistent optimization paradigm. Extensive experiments on five IBM backends confirm the effectiveness of our method, showing significant fidelity gains across all tasks under realistic noise. These results demonstrate that noise suppression through structure-aware design offers a powerful and generalizable strategy for enhancing circuit performance on NISQ hardware. Our framework bridges low-level hardware constraints with high-level circuit synthesis, paving the way for more robust and efficient quantum programming in the presence of noise.
Quantum noise and decoherence are the primary barriers to realising fault-tolerant quantum computation on near-term hardware. Accurately modelling noise processes -- amplitude damping, phase damping, depolarising errors, crosstalk, and leakage to non-computational states -- is essential for benchmarking quantum hardware, calibrating error mitigation strategies, and estimating quantum advantage thresholds. This paper proposes the Quantum Noise Computational Model (QNCM) framework, a systematic evaluation of seven noise modelling approaches -- Lindblad master equation, Kraus operator formalism, stochastic Schrodinger equation, randomised benchmarking models, process tomography, Pauli noise models, and coherent error models -- across 24 quantum hardware calibration datasets from IBM Quantum, Google Sycamore, and IonQ Aria processors. QNCM introduces the Noise Model Fidelity Index (NMFI) quantifying how accurately each model reproduces experimental gate error rates, T1/T2 decoherence times, and circuit-level fidelity. Key results: Lindblad master equation achieves NMFI = 0.924 (highest accuracy); Pauli noise models achieve NMFI = 0.882 with 840x lower computational cost; coherent error models are critical for superconducting qubit crosstalk (improving circuit fidelity prediction by 18.4%). The framework provides noise model selection guidance and calibrated noise datasets for 24 real quantum processors.