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

arXiv:2607.08998

Shadow-Based Noise Fingerprinting of Simulated Quantum Noise Models

Vridhi Jain, Lei Zhang

cs.SEquant-phSubmitted Jul 10, 2026• Announced Jul 13, 2026• 0 citations• Revised Aug 9, 2026 (v3)
v1 · Jul 10, 2026v2 · Jul 27, 2026v3 · Aug 9, 2026
arXivHTMLPDF

Abstract

Accurate noise classification is essential for operating near-term quantum processors, yet existing approaches, such as quantum process tomography, scale exponentially with system size, limiting their practicality for routine calibration. We propose a measurement-efficient noise fingerprinting pipeline that combines structured classical shadow tomography with physics-informed feature engineering to identify noise channels from a fixed set of 3-qubit probe circuits. Each sample is represented by a feature vector constructed from randomized Pauli measurements and derived observables designed to resolve physically similar noise channels that produce overlapping signatures under generic measurement sets. We evaluate random forest, extra trees, and a multilayer perceptron on 10,000 labeled samples spanning ten noise models. The three classifiers achieve comparable performance. In the reported runs, random forest and extra trees perform similarly, achieving approximately 0.736 test accuracy and 0.729-0.730 macro F1, compared with 0.715 accuracy and 0.699 macro F1 for the multilayer perceptron. We further analyze the effect of the noise-strength sampling range and conduct a limited sensitivity check using analogous 2- and 4-qubit probes. Confusion analysis shows that readout error, phase flip, thermal relaxation, and bit flip are classified with high reliability, while most remaining errors occur among channels with similar physical effects.

Comments
4 pages, 3 figures, accepted by the 6th International Workshop on Quantum Software Engineering and Technology at IEEE Quantum Week

Related papers

Artifact of Live Noise Fingerprints as an Empirical Semantics for Noisy Quantum Channels

Anonymous for Peer Review

Abstract Quantum computations execute on noisy hardware whose effective noise is opaque and varies across platforms and over time. Complete descriptions of the underlying noise channel are often too detailed for reasoning, comparison, or analysis. We therefore study the semantic problem of constructing tractable abstractions of noisy quantum channels: given a finite set of observations, how can we summarise channel behaviour while preserving properties of interest? We characterise the effective noise of a quantum channel as an abstract domain of noise fingerprints. A noise fingerprint is an observational abstraction of a quantum channel over a finite experiment interface. We formalise the relation between concrete channels and fingerprints through an approximate Galois connection and derive four abstraction properties: soundness, conciseness, efficiency, and adaptiveness. We provide an executable instantiation of the abstraction function using the classical-shadow estimator, together with a partial concretisation function that reconstructs channels from fingerprints via semidefinite programming. We evaluate this abstraction against the formal properties: the soundness bound is shown to be tight, compression over the concrete domain is exponential with near-lossless retention, sample complexity is (poly)logarithmic, and incremental updates track channel drift at the optimal rate. Additional experiments show that fingerprints preserve physically meaningful noise signatures and distinguish cross-platform behaviours. Together, this work brings abstract interpretation to the empirical semantics of noisy quantum channels, providing a lightweight, formally grounded abstraction for characterising, comparing, and monitoring noisy quantum executions. Context Characterising noisy quantum channels remains challenging: full process tomography scales exponentially with system size, while commonly used noise metrics provide limited semantic guarantees. This artifact accompanies a paper that introduces live noise fingerprints as a reusable semantic abstraction for noisy quantum channels. A noise fingerprint is defined as the matrix of deviations between observed and ideal Pauli expectation values over a fixed experimental interface. We show that these fingerprints form an abstract domain for completely positive trace-preserving (CPTP) maps through an approximate Galois connection to the concrete channel space. The framework is evaluated against four formal requirements: Soundness — a Lipschitz bound relating abstract distance to diamond-norm distance. Conciseness — substantial compression relative to the underlying channel representation. Estimation efficiency — practical estimation from finite measurements. Adaptiveness — support for incremental updates under channel drift. The concrete instantiation uses the Huang–Kueng–Preskill classical-shadow estimator with median-of-means aggregation. Every theoretical claim is machine-verified. The artifact includes exact SDP-based diamond norms for small systems, two independently validated large-system ground truths (completely bounded norm multiplicativity for single-site channel pairs and the exact Pauli-channel ℓ₁ formula), and a product-factorised classical-shadow sampler proven distribution-identical to the full sampler. All four research questions are evaluated for n = 1–8 qubits across 12 experimental configurations, combining three channel families with hardware-informed profiles derived from IBM Boston and Quantinuum H2 calibrations under both Qiskit and Cirq simulation variants. All stochastic measurements are reported as mean ± standard deviation, and every figure includes error bars. New in this version. The artifact now includes a self-contained worked example, worked_example.ipynb, which runs the abstraction end to end on small quantum programs: from OpenQASM source, through the gates and the noise a backend inserts around them, to the observation interface a program's test defines, the resulting fingerprints, and the decision they support. It is written for readers approaching the work from programming languages and software engineering rather than quantum information, and it is the shortest path into the artifact: it requires only NumPy, runs in seconds, and needs no SDP solver, quantum SDK, or hardware. The artifact also characterises the operational limits of the approach, including detectability thresholds at realistic hardware noise rates, readout-error-induced bias floors, explicit interface-blindness constructions and their informationally complete remedies, and drift regimes where warm-started updates remain beneficial. Artifact Evaluation file This artifact accompanies the paper "Live Noise Fingerprints as an Empirical Semantics for Noisy Quantum Channels" and provides the complete implementation, experiments, and evaluation pipeline used to validate the proposed noise-fingerprint abstraction. The package includes a pure Python/NumPy implementation of noisy quantum channels, classical-shadow fingerprint estimation, semidefinite-programming-based concretisation, machine-checked tests, a worked-example notebook, and scripts to reproduce every figure and quantitative claim reported in the paper. Evaluators can start from the notebook for a program-level walkthrough of the abstraction, verify correctness using an automated verification target, and regenerate all experimental results, including scalability, convergence, variance, detectability, and warm-start analyses. Precomputed datasets and figures are included to enable immediate inspection, while the full experimental pipeline can be rerun from scratch without requiring GPUs, quantum hardware, or network access after dependency installation. The artifact is released under the MIT License and is designed for reuse by allowing new quantum programs, hardware noise profiles, qubit counts, and abstraction properties to be incorporated with minimal modification. Package Contents worked_example.ipynb — a self-contained, executed walkthrough of the abstraction on small programs. It covers: an OpenQASM front end and the ideal denotation of a program; the realisation of that program on a backend, with the backend's noise inserted after every gate and rendered in the printed circuit; the observation interface, derived from what a program's test asserts rather than chosen freely; fingerprints and the equivalence decision, over five benchmark programs and five backend configurations, with a diagnostic per program naming the observation that carries the difference; interface blindness and its repair by refinement; prefix sweeping to localise which gate two realisations first disagree on; a control panel for running the whole chain on the reader's own .qasm file; measurement-reuse accounting, and numerical checks of three properties from the paper. Outputs are stored in the notebook, so it can be read without execution. src/ Pure NumPy implementation of: Quantum states and channels Noise fingerprints Literal classical-shadow estimators (full and product-factorised) Optional readout-error models Exact diamond-norm computation via Watrous SDPs Validated large-system diamond-norm ground truths General SDP-based concretisation procedures Configurable noise models Registry of all 12 experimental configurations (profiles.py) experiments/ Experiment drivers for all research questions: rq1_sweep.py — Soundness rq2_sweep.py — Conciseness rq3_sweep.py — Estimation Efficiency rq4_sweep.py — Adaptiveness Additional components include audited single-size reproductions, a stress-test suite (realistic_limits.py), a seed-sweep audit framework, and a figure-generation pipeline with shared publication styling. results/ — JSON outputs containing every measured data point for each research question. figures/ — Results figures, one file per figure. tests/ — Forty-one machine-checked tests, including validation of exact diamond-norm ground truths, verification of the product-factorised sampler, and consistency checks required before experimental execution. Additional files — README.md (overview, results summary, and reproduction guide), Dockerfile (containerised environment), requirements.txt (pinned dependencies). Installation The artifact requires Python 3.10 or later. Install dependencies: pip install -r requirements.txt Required packages: NumPy, SciPy, CVXPY, scikit-learn, matplotlib. Opening the worked-example notebook additionally requires jupyter; the notebook itself imports only NumPy and can also be read as-is, since its outputs are stored. No quantum SDK is required; all simulations use a pure NumPy density-matrix implementation. A containerised environment is also provided: docker build . Getting started: the worked example jupyter notebook notebooks/worked_example.ipynb Approximate runtime: under one minute; NumPy only. This is the recommended entry point. It takes a program in OpenQASM, shows what a backend does to it, derives the observation interface from the program's test, computes the fingerprints, and decides whether the two backends are interchangeable for that test — the abstraction of the paper, on objects small enough to check by hand. Expected results Two backends implementing the same physical noise by different means are judged interchangeable on all five programs, with deviations at machine precision (≈ 10⁻¹⁶). A leftover coherent phase rotation, of the kind an optimisation pass can introduce, is flagged on four of the five programs and correctly cleared on the fifth, whose test reads only populations and therefore cannot be affected by it. Judged on a population-only interface, the same phase rotation is invisible; refining the interface to the observables the program's test actually reads restores it. The prefix sweep localises a fault confined to the two-qubit gate at exactly that gate, and reports the earliest divergence when the

•Jul 5, 2026•0 citations

Shadow tomography with noisy readouts

H. Chau Nguyen

Shadow tomography is a scalable technique to characterise the quantum state of a quantum computer or quantum simulator. The protocol is based on the transformation of the outcomes of random measurements into the so-called classical shadows, which can later be transformed into samples of expectation values of the observables of interest. By construction, classical shadows are intrinsically sensitive to readout noise. In fact, the complicated structure of the readout noise due to crosstalk appears to be detrimental to its scalability. We show that classical shadows accept much more flexible constructions beyond the standard ones, which can eventually be made more conformable with readout noise. With this construction, we show that readout errors in classical shadows can be efficiently mitigated by randomly flipping the qubit before, and the classical outcome bit after the measurement, referred to as $X$-twirling. That a single $X$-gate is sufficient for mitigating readout noise for classical shadows is in contrast to Clifford-twirling, where the implementation of random Clifford gates is required.

arXiv:2310.17328•Oct 26, 2023•1 citation

Shadow Engineering of Quantum Processes

Tian-Ci Tian, De-Tao Jiang, Wei-Ming Zhu +3 more

Characterizing quantum processes is essential for hardware benchmarking, error diagnosis, and algorithm verification. While recent work [PRX QUANTUM \textbf{4}, 040337 (2023)] extended classical shadows from quantum state to quantum process, enabling efficient single-channel $\mathcal{E}$ property prediction, its applicability to composite processes $f(\mathcal{E}_1, \mathcal{E}_2,\cdots, \mathcal{E}_k)$ remains unexplored. We introduce shadow engineering, a framework encoding the classical shadows of processes into sparse transfer matrices to predict $f(\mathcal{E}_1, \mathcal{E}_2,\cdots, \mathcal{E}_k)$ properties with proven polynomial sample complexity, matching single-channel efficiency while exponentially lower than quantum process tomography. Crucially, this approach repurposes existing $\mathcal{E}_m$-shadow data without physical execution of $f(\mathcal{E}_1, \mathcal{E}_2,\cdots, \mathcal{E}_k)$, enabling flexible quantum process characterization with minimal hardware overhead. We demonstrate the framework's effectiveness and practicality on a superconducting quantum processor for typical applications such as error mitigation and Hamiltonian dynamical simulation. This framework unlocks new capabilities for predicting complex quantum behaviors without physical re-execution, with immediate applications in near-term device calibration and quantum simulation.

quant-pharXiv:2606.12035•Jun 11, 2026•0 citations

A Systematic Study of Noise Effects in Hybrid Quantum-Classical Machine Learning

Bhavna Bose, Muhammad Faryad

Near-term quantum machine learning (QML) models operate in environments wherein noise is unavoidable, arising from both imperfect classical data acquisition and the limitations of noisy intermediate-scale quantum (NISQ) hardware. Although most existing studies have focused primarily on quantum circuit noise in isolation, the combined influence of corrupted classical inputs and quantum hardware noise has received comparatively little attention. In this work, we present a systematic experimental study of the robustness of a variational quantum classifier under realistic multi-level noise conditions. Using the Titanic dataset as a benchmark, a range of dataset-level noise models-including speckle noise, impulse noise, quantization noise, and feature dropout are applied to classical features prior to quantum encoding using a ZZ feature map. In parallel, hardware-inspired quantum noise channels such as depolarizing noise, amplitude damping, phase damping, Pauli errors, and readout errors are incorporated at the circuit level using the Qiskit Aer simulator. The experimental results indicate that noise in classical input data can significantly intensify the effects of quantum decoherence, resulting in less stable training and noticeably lower classification accuracy. Together, these observations emphasize the importance of designing and evaluating quantum machine learning pipelines with noise in mind, and highlight the need to consider classical and quantum noise simultaneously when assessing QML performance in the NISQ era

quant-pharXiv:2604.11541•Apr 14, 2026•0 citations

Scalable Tomography of Coherent Noise via Stabilizer-Code Polarization

Long Huang

Coherent errors accumulate linearly rather than diffusively, making them invisible to Pauli-twirl-based characterization. This poses a central challenge for quantum error correction, yet its characterization typically requires multi-qubit process tomography whose cost grows exponentially with system size. We show that measurement-free error correction projects arbitrary few-qubit coherent noise onto a fixed single-logical-qubit effective channel, enabling tomography at constant cost. Exploiting the polarization effects of bit-flip and phase-flip codes, we construct a concatenated effective-channel fingerprint and prove, by a parity-based contradiction argument, that the observation map is globally injective over a prescribed coherent-noise family; hence the underlying noise parameters are uniquely recovered without any sparsity or low-rank assumptions. We further corroborate the theoretical guarantee by a synthetic numerical test: over 500 randomly sampled points in the compact parameter domain, the $64\times11$ Jacobian of the concatenated fingerprint maintains full column rank with a strictly positive smallest singular value in every case, confirming that all 11 parameter directions remain first-order identifiable throughout the region. The recovered coherent noise model can be fed into decoding algorithms; relative to independent noise priors, we expect this to reduce the logical error rate for surface-code decoders. The protocol offers a scalable route to coherent-noise identification whenever the noise admits a moderate-dimensional parameterization.

quant-pharXiv:2610.19779•Oct 15, 2026•0 citations

Demonstration of robust and efficient quantum property learning with shallow shadows

Hong-Ye Hu, Andi Gu, Swarnadeep Majumder +7 more

Extracting information efficiently from quantum systems is crucial for quantum information processing. Classical shadows enable predicting many properties of arbitrary quantum states using few measurements. While random single-qubit measurements are experimentally friendly and suitable for learning low-weight Pauli observables, they perform poorly for nonlocal observables. Introducing a shallow random quantum circuit before measurements improves sample efficiency for high-weight Pauli observables and low-rank properties. However, in practice, these circuits can be noisy and bias the measurement results. Here, we propose the robust shallow shadows, which employs Bayesian inference to learn and mitigate noise in postprocessing. We analyze noise effects on sample complexity and the optimal circuit depth. We provide theoretical guarantees for the success of error mitigation under a wide class of noise processes. Experimental validation on a superconducting quantum processor confirms the advantage of our method, even in the presence of realistic noise, over single-qubit measurements for predicting diverse state properties, such as fidelity and entanglement entropy. Our protocol thus offers a scalable, robust, and sample-efficient method for quantum state characterization on near-term quantum devices. Fast and reliable characterisation of quantum systems is a key part of quantum technologies development. Here, the authors propose and demonstrate a way to embed noise characterisation in classical shadow estimation of nonlocal properties, enabling an efficient way to extract information from noisy quantum systems.

arXiv:2402.17911•Nature Communications•Mar 26, 2025•28 citations
Accurate noise classification is essential for operating near-term quantum processors, yet existing approaches, such as quantum process tomography, scale exponentially with system size, limiting their practicality for routine calibration. We propose a measurement-efficient noise fingerprinting pipeline that combines structured classical shadow tomography with physics-informed feature engineering to identify noise channels from a fixed set of 3-qubit probe circuits. Each sample is represented by a feature vector constructed from randomized Pauli measurements and derived observables designed to resolve physically similar noise channels that produce overlapping signatures under generic measurement sets. We evaluate random forest, extra trees, and a multilayer perceptron on 10,000 labeled samples spanning ten noise models. The three classifiers achieve comparable performance. In the reported runs, random forest and extra trees perform similarly, achieving approximately 0.736 test accuracy and 0.729-0.730 macro F1, compared with 0.715 accuracy and 0.699 macro F1 for the multilayer perceptron. We further analyze the effect of the noise-strength sampling range and conduct a limited sensitivity check using analogous 2- and 4-qubit probes. Confusion analysis shows that readout error, phase flip, thermal relaxation, and bit flip are classified with high reliability, while most remaining errors occur among channels with similar physical effects.
Abstract Quantum computations execute on noisy hardware whose effective noise is opaque and varies across platforms and over time. Complete descriptions of the underlying noise channel are often too detailed for reasoning, comparison, or analysis. We therefore study the semantic problem of constructing tractable abstractions of noisy quantum channels: given a finite set of observations, how can we summarise channel behaviour while preserving properties of interest? We characterise the effective noise of a quantum channel as an abstract domain of noise fingerprints. A noise fingerprint is an observational abstraction of a quantum channel over a finite experiment interface. We formalise the relation between concrete channels and fingerprints through an approximate Galois connection and derive four abstraction properties: soundness, conciseness, efficiency, and adaptiveness. We provide an executable instantiation of the abstraction function using the classical-shadow estimator, together with a partial concretisation function that reconstructs channels from fingerprints via semidefinite programming. We evaluate this abstraction against the formal properties: the soundness bound is shown to be tight, compression over the concrete domain is exponential with near-lossless retention, sample complexity is (poly)logarithmic, and incremental updates track channel drift at the optimal rate. Additional experiments show that fingerprints preserve physically meaningful noise signatures and distinguish cross-platform behaviours. Together, this work brings abstract interpretation to the empirical semantics of noisy quantum channels, providing a lightweight, formally grounded abstraction for characterising, comparing, and monitoring noisy quantum executions. Context Characterising noisy quantum channels remains challenging: full process tomography scales exponentially with system size, while commonly used noise metrics provide limited semantic guarantees. This artifact accompanies a paper that introduces live noise fingerprints as a reusable semantic abstraction for noisy quantum channels. A noise fingerprint is defined as the matrix of deviations between observed and ideal Pauli expectation values over a fixed experimental interface. We show that these fingerprints form an abstract domain for completely positive trace-preserving (CPTP) maps through an approximate Galois connection to the concrete channel space. The framework is evaluated against four formal requirements: Soundness — a Lipschitz bound relating abstract distance to diamond-norm distance. Conciseness — substantial compression relative to the underlying channel representation. Estimation efficiency — practical estimation from finite measurements. Adaptiveness — support for incremental updates under channel drift. The concrete instantiation uses the Huang–Kueng–Preskill classical-shadow estimator with median-of-means aggregation. Every theoretical claim is machine-verified. The artifact includes exact SDP-based diamond norms for small systems, two independently validated large-system ground truths (completely bounded norm multiplicativity for single-site channel pairs and the exact Pauli-channel ℓ₁ formula), and a product-factorised classical-shadow sampler proven distribution-identical to the full sampler. All four research questions are evaluated for n = 1–8 qubits across 12 experimental configurations, combining three channel families with hardware-informed profiles derived from IBM Boston and Quantinuum H2 calibrations under both Qiskit and Cirq simulation variants. All stochastic measurements are reported as mean ± standard deviation, and every figure includes error bars. New in this version. The artifact now includes a self-contained worked example, worked_example.ipynb, which runs the abstraction end to end on small quantum programs: from OpenQASM source, through the gates and the noise a backend inserts around them, to the observation interface a program's test defines, the resulting fingerprints, and the decision they support. It is written for readers approaching the work from programming languages and software engineering rather than quantum information, and it is the shortest path into the artifact: it requires only NumPy, runs in seconds, and needs no SDP solver, quantum SDK, or hardware. The artifact also characterises the operational limits of the approach, including detectability thresholds at realistic hardware noise rates, readout-error-induced bias floors, explicit interface-blindness constructions and their informationally complete remedies, and drift regimes where warm-started updates remain beneficial. Artifact Evaluation file This artifact accompanies the paper "Live Noise Fingerprints as an Empirical Semantics for Noisy Quantum Channels" and provides the complete implementation, experiments, and evaluation pipeline used to validate the proposed noise-fingerprint abstraction. The package includes a pure Python/NumPy implementation of noisy quantum channels, classical-shadow fingerprint estimation, semidefinite-programming-based concretisation, machine-checked tests, a worked-example notebook, and scripts to reproduce every figure and quantitative claim reported in the paper. Evaluators can start from the notebook for a program-level walkthrough of the abstraction, verify correctness using an automated verification target, and regenerate all experimental results, including scalability, convergence, variance, detectability, and warm-start analyses. Precomputed datasets and figures are included to enable immediate inspection, while the full experimental pipeline can be rerun from scratch without requiring GPUs, quantum hardware, or network access after dependency installation. The artifact is released under the MIT License and is designed for reuse by allowing new quantum programs, hardware noise profiles, qubit counts, and abstraction properties to be incorporated with minimal modification. Package Contents worked_example.ipynb — a self-contained, executed walkthrough of the abstraction on small programs. It covers: an OpenQASM front end and the ideal denotation of a program; the realisation of that program on a backend, with the backend's noise inserted after every gate and rendered in the printed circuit; the observation interface, derived from what a program's test asserts rather than chosen freely; fingerprints and the equivalence decision, over five benchmark programs and five backend configurations, with a diagnostic per program naming the observation that carries the difference; interface blindness and its repair by refinement; prefix sweeping to localise which gate two realisations first disagree on; a control panel for running the whole chain on the reader's own .qasm file; measurement-reuse accounting, and numerical checks of three properties from the paper. Outputs are stored in the notebook, so it can be read without execution. src/ Pure NumPy implementation of: Quantum states and channels Noise fingerprints Literal classical-shadow estimators (full and product-factorised) Optional readout-error models Exact diamond-norm computation via Watrous SDPs Validated large-system diamond-norm ground truths General SDP-based concretisation procedures Configurable noise models Registry of all 12 experimental configurations (profiles.py) experiments/ Experiment drivers for all research questions: rq1_sweep.py — Soundness rq2_sweep.py — Conciseness rq3_sweep.py — Estimation Efficiency rq4_sweep.py — Adaptiveness Additional components include audited single-size reproductions, a stress-test suite (realistic_limits.py), a seed-sweep audit framework, and a figure-generation pipeline with shared publication styling. results/ — JSON outputs containing every measured data point for each research question. figures/ — Results figures, one file per figure. tests/ — Forty-one machine-checked tests, including validation of exact diamond-norm ground truths, verification of the product-factorised sampler, and consistency checks required before experimental execution. Additional files — README.md (overview, results summary, and reproduction guide), Dockerfile (containerised environment), requirements.txt (pinned dependencies). Installation The artifact requires Python 3.10 or later. Install dependencies: pip install -r requirements.txt Required packages: NumPy, SciPy, CVXPY, scikit-learn, matplotlib. Opening the worked-example notebook additionally requires jupyter; the notebook itself imports only NumPy and can also be read as-is, since its outputs are stored. No quantum SDK is required; all simulations use a pure NumPy density-matrix implementation. A containerised environment is also provided: docker build . Getting started: the worked example jupyter notebook notebooks/worked_example.ipynb Approximate runtime: under one minute; NumPy only. This is the recommended entry point. It takes a program in OpenQASM, shows what a backend does to it, derives the observation interface from the program's test, computes the fingerprints, and decides whether the two backends are interchangeable for that test — the abstraction of the paper, on objects small enough to check by hand. Expected results Two backends implementing the same physical noise by different means are judged interchangeable on all five programs, with deviations at machine precision (≈ 10⁻¹⁶). A leftover coherent phase rotation, of the kind an optimisation pass can introduce, is flagged on four of the five programs and correctly cleared on the fifth, whose test reads only populations and therefore cannot be affected by it. Judged on a population-only interface, the same phase rotation is invisible; refining the interface to the observables the program's test actually reads restores it. The prefix sweep localises a fault confined to the two-qubit gate at exactly that gate, and reports the earliest divergence when the
Shadow tomography is a scalable technique to characterise the quantum state of a quantum computer or quantum simulator. The protocol is based on the transformation of the outcomes of random measurements into the so-called classical shadows, which can later be transformed into samples of expectation values of the observables of interest. By construction, classical shadows are intrinsically sensitive to readout noise. In fact, the complicated structure of the readout noise due to crosstalk appears to be detrimental to its scalability. We show that classical shadows accept much more flexible constructions beyond the standard ones, which can eventually be made more conformable with readout noise. With this construction, we show that readout errors in classical shadows can be efficiently mitigated by randomly flipping the qubit before, and the classical outcome bit after the measurement, referred to as XXX-twirling. That a single XXX-gate is sufficient for mitigating readout noise for classical shadows is in contrast to Clifford-twirling, where the implementation of random Clifford gates is required.
Characterizing quantum processes is essential for hardware benchmarking, error diagnosis, and algorithm verification. While recent work [PRX QUANTUM \textbf{4}, 040337 (2023)] extended classical shadows from quantum state to quantum process, enabling efficient single-channel E\mathcal{E}E property prediction, its applicability to composite processes f(E1,E2,⋯ ,Ek)f(\mathcal{E}_1, \mathcal{E}_2,\cdots, \mathcal{E}_k)f(E1​,E2​,⋯,Ek​) remains unexplored. We introduce shadow engineering, a framework encoding the classical shadows of processes into sparse transfer matrices to predict f(E1,E2,⋯ ,Ek)f(\mathcal{E}_1, \mathcal{E}_2,\cdots, \mathcal{E}_k)f(E1​,E2​,⋯,Ek​) properties with proven polynomial sample complexity, matching single-channel efficiency while exponentially lower than quantum process tomography. Crucially, this approach repurposes existing Em\mathcal{E}_mEm​-shadow data without physical execution of f(E1,E2,⋯ ,Ek)f(\mathcal{E}_1, \mathcal{E}_2,\cdots, \mathcal{E}_k)f(E1​,E2​,⋯,Ek​), enabling flexible quantum process characterization with minimal hardware overhead. We demonstrate the framework's effectiveness and practicality on a superconducting quantum processor for typical applications such as error mitigation and Hamiltonian dynamical simulation. This framework unlocks new capabilities for predicting complex quantum behaviors without physical re-execution, with immediate applications in near-term device calibration and quantum simulation.
Near-term quantum machine learning (QML) models operate in environments wherein noise is unavoidable, arising from both imperfect classical data acquisition and the limitations of noisy intermediate-scale quantum (NISQ) hardware. Although most existing studies have focused primarily on quantum circuit noise in isolation, the combined influence of corrupted classical inputs and quantum hardware noise has received comparatively little attention. In this work, we present a systematic experimental study of the robustness of a variational quantum classifier under realistic multi-level noise conditions. Using the Titanic dataset as a benchmark, a range of dataset-level noise models-including speckle noise, impulse noise, quantization noise, and feature dropout are applied to classical features prior to quantum encoding using a ZZ feature map. In parallel, hardware-inspired quantum noise channels such as depolarizing noise, amplitude damping, phase damping, Pauli errors, and readout errors are incorporated at the circuit level using the Qiskit Aer simulator. The experimental results indicate that noise in classical input data can significantly intensify the effects of quantum decoherence, resulting in less stable training and noticeably lower classification accuracy. Together, these observations emphasize the importance of designing and evaluating quantum machine learning pipelines with noise in mind, and highlight the need to consider classical and quantum noise simultaneously when assessing QML performance in the NISQ era
Coherent errors accumulate linearly rather than diffusively, making them invisible to Pauli-twirl-based characterization. This poses a central challenge for quantum error correction, yet its characterization typically requires multi-qubit process tomography whose cost grows exponentially with system size. We show that measurement-free error correction projects arbitrary few-qubit coherent noise onto a fixed single-logical-qubit effective channel, enabling tomography at constant cost. Exploiting the polarization effects of bit-flip and phase-flip codes, we construct a concatenated effective-channel fingerprint and prove, by a parity-based contradiction argument, that the observation map is globally injective over a prescribed coherent-noise family; hence the underlying noise parameters are uniquely recovered without any sparsity or low-rank assumptions. We further corroborate the theoretical guarantee by a synthetic numerical test: over 500 randomly sampled points in the compact parameter domain, the 64×1164\times1164×11 Jacobian of the concatenated fingerprint maintains full column rank with a strictly positive smallest singular value in every case, confirming that all 11 parameter directions remain first-order identifiable throughout the region. The recovered coherent noise model can be fed into decoding algorithms; relative to independent noise priors, we expect this to reduce the logical error rate for surface-code decoders. The protocol offers a scalable route to coherent-noise identification whenever the noise admits a moderate-dimensional parameterization.
Extracting information efficiently from quantum systems is crucial for quantum information processing. Classical shadows enable predicting many properties of arbitrary quantum states using few measurements. While random single-qubit measurements are experimentally friendly and suitable for learning low-weight Pauli observables, they perform poorly for nonlocal observables. Introducing a shallow random quantum circuit before measurements improves sample efficiency for high-weight Pauli observables and low-rank properties. However, in practice, these circuits can be noisy and bias the measurement results. Here, we propose the robust shallow shadows, which employs Bayesian inference to learn and mitigate noise in postprocessing. We analyze noise effects on sample complexity and the optimal circuit depth. We provide theoretical guarantees for the success of error mitigation under a wide class of noise processes. Experimental validation on a superconducting quantum processor confirms the advantage of our method, even in the presence of realistic noise, over single-qubit measurements for predicting diverse state properties, such as fidelity and entanglement entropy. Our protocol thus offers a scalable, robust, and sample-efficient method for quantum state characterization on near-term quantum devices. Fast and reliable characterisation of quantum systems is a key part of quantum technologies development. Here, the authors propose and demonstrate a way to embed noise characterisation in classical shadow estimation of nonlocal properties, enabling an efficient way to extract information from noisy quantum systems.