量子科技

Objective: Survival prediction in oncology and cardiology requires models that can capture nonlinear prognostic structure while remaining interpretable, calibrated, and clinically safe. This study develops and evaluates a diagnostic-first hybrid quantum-classical framework for right-censored survival analysis. Methods: We introduce KTA-Survival (Kernel-Target Alignment for survival), a pre-training feasibility diagnostic that adapts kernel-target alignment to censored outcomes by comparing a quantum fidelity kernel with a concordance-based survival target kernel. We then propose QResid-Boost (Quantum Residual Boosting), a Cox-LASSO-anchored residual framework in which a variational quantum circuit is trained on martingale residuals through a Quantum-Skip-Residual architecture. A sigmoid-bounded scalar gate, α, constrains the quantum contribution and allows the model to reduce to the classical baseline when the residual signal is uninformative. The framework was evaluated on GBSG2 (German Breast Cancer Study Group 2; n = 686), FLChain (serum free light chain; n = 1500), WHAS500 (Worcester Heart Attack Study; n = 500), and a synthetic Weibull positive-control dataset containing high-frequency periodic interactions. Results: On the GBSG2 hold-out partition, Random Survival Forest achieved the highest concordance (C = 0.7188), followed by the Stacking ensemble (C = 0.7128), Cox-LASSO (C = 0.7019), and QResid-Boost (C = 0.7016). The leading classical and hybrid models did not differ significantly by paired bootstrap testing, whereas all outperformed the pure quantum variants. In the synthetic positive-control cohort, QResid-Boost improved over Cox-LASSO by ΔC = +0.0397, demonstrating that the quantum residual can add value when nonlinear periodic structure remains after the linear baseline. KTA-Survival yielded positive ΔKTA values across the evaluated datasets and correctly identified the regime in which the quantum residual produced its largest measurable gain. Conclusions: The proposed diagnostic-first framework reframes quantum survival modelling as a gated enrichment strategy rather than an unconstrained replacement for classical risk models. In low-dimensional clinical cohorts where linear structure already explains most prognostic signal, the framework behaves conservatively; when residual nonlinear structure is present, it can provide measurable improvement without uncontrolled model drift.

African swine fever virus (ASFV) and porcine epidemic diarrhea virus (PEDV) differ in viral biology and cellular tropism, yet both pathogens suppress macrophage-mediated immune responses in pigs. To identify a conserved macrophage suppression module shared by ASFV and PEDV and evaluate quantum computing as an independent framework for biological network validation. Integrated analysis of publicly available GEO datasets (GSE231435 for ASFV and GSE306895) identified 471 shared downregulated genes. A network- and multi-omics-informed 20-gene core was selected and encoded as a 20-qubit modularity-based Quadratic Unconstrained Binary Optimization (QUBO) problem. Community detection was benchmarked using the Quantum Approximate Optimization Algorithm (QAOA) on both the IBM Quantum Aer simulator and the 156-qubit IBM Fez (Heron r2) quantum processor and compared with brute-force enumeration and simulated annealing. A conserved macrophage suppression module shared by ASFV and PEDV was identified. For the STRING protein-protein interaction network, QAOA at circuit depth p = 3 reproduced the brute-force optimum with an approximation ratio of 1.000. In contrast, performance progressively declined in the denser co-expression network with increasing circuit depth, consistent with noise accumulation under current Noisy Intermediate-Scale Quantum (NISQ) conditions. Multi-run consensus analysis identified stable hub genes, including MMP9 and SLA-DOA, as well as genes exhibiting variable community assignments. These findings reveal a conserved macrophage suppression module shared between ASFV and PEDV and demonstrate that quantum computing can serve as an independent validation framework for biologically meaningful host-response networks. Network topology emerged as a key determinant of QAOA performance on real NISQ hardware.

We study the quantum dynamics in Krylov space of a spatially flat, homogeneous, and isotropic universe sourced with a massless scalar field within Wheeler-DeWitt (WDW) quantum cosmology and loop quantum cosmology (LQC) frameworks. The availability of a physical Hilbert space and physical Hamiltonian and the presence of an internal clock enable us to construct the Krylov basis analytically by applying the Lanczos algorithm. We then evaluate both the Krylov state and operator complexity for WDW quantum cosmology and LQC on this basis. In regimes where the wave function of the universe is sharply peaked, our results indicate that the Krylov complexity grows quadratically with the scalar field clock for the state and operator complexities in both the WDW quantum cosmology and LQC. We further show that the operator complexity is exactly twice the state complexity in these regimes. We discuss the interpretation of the global behavior of these systems by calculating the Krylov entropy for both quantum cosmological frameworks. We observe that in LQC, the Krylov complexity and entropy remain finite at the bounce, whereas in the WDW quantum cosmology, they diverge at the big bang/crunch singularity. Our work provides the first example of computing Krylov complexity for a system with a totally constrained Hamiltonian and no external time, a framework to calculate a purely quantum-mechanical entropy in quantum cosmology, and, to our knowledge, the first direct bridge between Krylov complexity and canonical quantum cosmology, as a first step toward understanding how polymerized quantum geometry modifies complexity and entropy.

Quantum computing offers new possibilities for solving combinatorial optimization problems with rapidly growing search spaces. Among emerging hardware platforms, photonic quantum computers based on boson sampling provide a promising approach for sampling-based optimization methods. In this work, we investigate the application of the Binary Bosonic Solver, a hybrid quantum-classical algorithm designed for photonic quantum processors, to the binary portfolio optimization problem derived from the classical mean-variance framework. In addition to evaluating the feasibility of solving such problems on photonic quantum hardware, we analyze the behavior of the Binary Bosonic Solver algorithm under different architectural and optimization parameters, including interferometer loop configurations and gradient estimation methods. Benchmark instances are generated using historical financial market data, and experiments are performed both on a photonic quantum computer simulator and on the ORCA PT-1 photonic quantum processor installed at Poznan Supercomputing and Networking Center, with results compared to those obtained using a classical optimization algorithm. The results demonstrate that portfolio optimization can be successfully executed on current photonic quantum hardware and that the Binary Bosonic Solver algorithm consistently produces feasible and high-quality solutions, highlighting the practical potential of photonic quantum computing for combinatorial optimization problems.

Brain disorder diagnosis and prediction remain challenging because neuroimaging, electrophysiological, behavioral, and multimodal data are high-dimensional, noisy, heterogeneous, and limited by small clinical cohorts. This systematic review synthesised applications of quantum artificial intelligence (QAI) for brain disorder diagnosis, prediction, detection, and monitoring. Following PRISMA guidelines, studies published from 2016 to 13 January 2026 were retrieved from Scopus, Web of Science, and IEEE Xplore. After screening, 36 studies met the eligibility criteria and were qualitatively analysed according to disorder category, data modality, QAI method, implementation setting, validation strategy, and performance. At the broader disease-group level, neurodegenerative disorders were the most frequently investigated, followed by mental health and psychiatric disorders. At the individual level, Parkinson's disease and schizophrenia were the leading applications, followed by depression, anxiety, Alzheimer's disease, and stress-related tasks. MRI-based modalities were the most frequently used data source, followed by multimodal data and EEG. Methodologically, primary QAI approaches were dominated by quantum neural and QDL architectures, followed by quantum-inspired optimization or feature-selection methods and quantum-kernel/conventional QML classifiers. Qiskit/IBM Quantum and PennyLane were the most frequently reported quantum software frameworks. However, most studies relied on simulators, classical quantum-inspired implementations, or unclear implementation settings, with limited real-hardware evaluation. QAI shows emerging potential for brain disorder analysis, particularly through hybrid quantum-classical learning, quantum neural architectures, quantum-kernel methods, and quantum-inspired optimization. Nevertheless, current evidence remains preliminary and requires larger datasets, subject-level and external validation, fair classical benchmarking, noise-resilient circuits, real quantum hardware evaluation, explainability, and clinical validation.

Quantum Secure Multi-party Summation (QSMS) is a fundamental primitive of Quantum Secure Multi-party Computation (QSMC), enabling multiple participants to jointly compute the sum of their private inputs without disclosing individual data. However, most existing QSMS protocols require all participants to possess full quantum capabilities and often rely on pre-shared keys, auxiliary mask transmission, or multiple trusted third parties, resulting in high communication overhead and limited practicality. To address these limitations, we propose an efficient Semi-Quantum Secure Multi-party Summation (SQSMS) protocol based on d-dimensional n-particle entangled states. By exploiting the global correlation properties of high-dimensional entangled states, the proposed protocol generates correlated random masks directly from quantum measurement outcomes. These masks cancel automatically during the aggregation process, eliminating the need for additional mask distribution and transmission. Compared with existing QSMS schemes, the proposed protocol reduces communication overhead, improves quantum efficiency, and avoids reliance on pre-shared keys or multiple trusted third parties. Moreover, only simple measurement operations are required from classical participants, making the protocol more practical for semi-quantum environments. We further provide formal correctness and security analyses of the proposed protocol and conduct quantum circuit simulations using the IBM Qiskit platform to demonstrate its feasibility. Moreover, based on the proposed summation protocol, we design several extended application protocols, including anonymous voting, anonymous auction, and anonymous ranking, which further illustrate the scalability and practical applicability of the proposed scheme.

Quantum neural architecture search (QNAS) has emerged as a promising approach for automatically designing parameterized quantum circuits (PQCs) for near-term quantum machine learning tasks. However, quantum evolutionary algorithm (QEA)-based QNAS methods often suffer from slow distribution concentration and insufficient update adaptivity in high-dimensional discrete search spaces, which limits both search efficiency and final model performance. To address these issues, this paper proposes Structural Probability Adaptive Quantum Neural Architecture Search (SPA-QNAS), a search-dynamics-enhanced QNAS method built upon the EQNAS benchmark framework. SPA-QNAS introduces two complementary mechanisms into the QPV-driven evolutionary search loop: Structural Probability Enhancement (SPE) and Adaptive Evolutionary Control (AEC). SPE reinforces elite structural decisions to accelerate the concentration of the structural sampling distribution toward high-fitness regions, while AEC adaptively regulates the rotation updates of non-elite individuals according to fitness feedback, thereby improving update stability and suppressing ineffective disturbances. Under the same search space, circuit template, and quantum resource budget as EQNAS, SPA-QNAS is evaluated on the MNIST and Warship benchmark datasets. Experimental results across multiple independent runs demonstrate that SPA-QNAS achieves higher classification accuracy and more stable performance compared with EQNAS. In representative experiments, SPA-QNAS achieves classification accuracies of 99.42% on MNIST and 85.33% on Warship under the same search space and quantum resource budget as EQNAS. These results indicate that improving QPV-based evolutionary update dynamics is an effective way to enhance the stability and robustness of QNAS under fixed quantum resource constraints.

Continuous-variable quantum systems are foundational to quantum computation, communication, and sensing. While traditional representations using wave functions or density matrices are often impractical, the tomographic picture of quantum mechanics provides an accessible alternative by associating quantum states with classical probability distribution functions called tomograms. Despite its advantages, including compatibility with classical statistical methods, the tomographic method remains underutilized due to a lack of robust estimation techniques. This Letter addresses this gap by introducing a nonparametric kernel quantum state estimation (KQSE) framework for reconstructing quantum states and their trace characteristics from noisy data, without prior knowledge of the state. In contrast to existing methods, KQSE yields estimates of the density matrix in various bases, as well as trace quantities such as purity, higher moments, overlap, and trace distance, with a near-optimal convergence rate of O[over ˜](T^{-1}), where T is the total number of measurements. KQSE is robust for multimodal, non-Gaussian states, making it particularly well suited for characterizing states essential for quantum science.

This review examines recent quantum chemical methodologies applied to investigating antibiotic structure and mechanisms of action. The discussion is organised into three sections: (1) enzymatic hydrolysis of the β-lactam ring, (2) interactions of antibiotics with ribosomal subunits and enzyme active sites, and (3) complex formation with metal ions. Each section evaluates how quantum chemical approaches, particularly density functional theory (DFT) and hybrid QM/MM techniques, model molecular processes relevant to antibiotic function, including transition states, electron density analyses, and metal coordination effects on antibacterial activity. Selected studies demonstrate the utility of these methodologies in interpreting experimental data and predicting physicochemical and biological properties of novel compounds. Distinct from previous literature, this review provides a comparative and up-to-date synthesis of quantum chemical methods related to enzymatic mechanisms and metal-based antibiotic systems, emphasising experimental validation strategies and practical guidelines for method selection in antibiotic research. It also identifies areas where quantum chemical modelling can integrate with experimental pharmacology and structural biology to support the rational design of next-generation antimicrobial agents. The review concludes by advocating an interdisciplinary framework combining quantum chemistry, biochemistry, and pharmacology to address antibiotic resistance. The review focuses primarily on antibiotics targeting bacterial cell wall and protein synthesis, particularly β-lactam antibiotics, ribosome-targeting agents, and their interactions with metal ions. Computational methods discussed are mainly limited to DFT, ab initio, and hybrid QM/MM approaches. It does not cover membrane-disrupting antibiotics, antiviral or antifungal agents, machine learning-based prediction methods, or purely molecular dynamics approaches outside a quantum mechanical context.

An efficient method of solving the time-dependent Schrödinger equation for pure states is described: At each time step, a restricted subspace of the total Hilbert space is systematically and naturally constructed via the image of repeated applications of the Hamiltonian operator and the time evolution is computed exactly within the restricted subspace. The subspace is dynamically recomputed such that it co-evolves with the state vector. The method is built from the ground up as a parallel algorithm for graphics processing units and suited to Hamiltonians that are sparse in a given basis. We benchmark the method by comparing its results for a 1D Holstein model with previously published multiset-matrix-product state results and then apply the method to compute optical spectra and non-equilibrium dynamics of one-, two-, and three-dimensional model chromophore nanoaggregates.

A reactive agent operating in a complex environment must classify its perceived state and select an action under uncertainty. This uncertainty may arise from sensor noise, ambiguous perceptual configurations, or the limited separability of the action regions induced by the agent's policy. We propose a hybrid classical-quantum architecture for a reactive agent in which the perceived state, represented as a classical sensor vector, is mapped onto a quantum feature space. In this space, learned conceptualizations or rule-defined perceptual regions are represented as reference states, and similarities between the current perception and such references are used to support action selection. The architecture is evaluated on a public wall-following robot dataset. Two implementations are considered: (i) a quantum-kernel classifier based on ZZ feature maps and (ii) an illustrative quantum circuit that explicitly encodes sensor conditions into qubits and performs measurement-based action selection. The experimental evaluation is intended as an offline proxy for reactive decision-making, not as a demonstration of a complete closed-loop robotic controller or of quantum advantage. The results show that the proposed framework can represent perceptual ambiguity and connect quantum-state measurement to the selection of discrete reactive actions.

Blockchain-based sensor networks rely on public-key cryptography for transaction verification, auditability, and data integrity. However, widely used public-key mechanisms are quantum-vulnerable in the presence of Cryptographically Relevant Quantum Computers (CRQCs), requiring sensor-to-blockchain transactions to address both post-quantum security and exposure control. This paper proposes a post-quantum sensor-to-blockchain transaction framework that minimizes CRQC-aware exposure while preserving low-cost auditability. It defines a transaction workflow that represents sensor data through hash-based commitments instead of storing raw measurements on-chain. The workflow combines Module-Lattice-Based Digital Signature Algorithm (ML-DSA)-based authentication, threshold-based authorization, Module-Lattice-Based Key Encapsulation Mechanism (ML-KEM)-protected relay communication, and an event-based smart contract (EBSC) for compact audit recording. A Quantum Exposure Score (QES) is introduced as a transaction-level metric to quantify CRQC-induced exposure across cryptographic, relay, key-lifecycle, migration-readiness, and authorization dimensions. The framework is evaluated using differential pulse voltammetry (DPV) electrochemical sensor data, Constrained Application Protocol (CoAP) communication, and a Ganache-based blockchain, with scalability runs of up to 10,000 sensor transactions and ablation baselines. Compared with full on-chain storage, EBSC reduces gas consumption by approximately 80%, while QES decreases from 100 in the classical open scenario to 4 in the full framework. These results demonstrate that the proposed design provides a practical path for post-quantum secure sensor-to-blockchain transactions.

Optical nanofibers provide a versatile platform for exploring strong light-matter interactions at the single-quantum level, a central goal of quantum optics. Their subwavelength diameter supports tightly confined guided modes with strong evanescent fields, enabling efficient coupling between photons and nearby atoms. By incorporating fiber Bragg gratings, nanofibers can form fully fiber-integrated high-finesse cavities, combining strong atom-photon coupling with efficient and low-loss optical input-output interfacing in the setting of cavity quantum electrodynamics (QED). In parallel, nanofiber systems naturally realize waveguide QED, where atoms interact strongly with propagating photons in a one-dimensional geometry. In this review, we introduce the underlying physical principles, summarize key experimental techniques and representative phenomena, and discuss prospects for applications in quantum information processing, quantum networks, and related quantum technologies.

Quantum networks and quantum repeaters represent the promising avenues for building large-scale quantum information systems, serving as foundational infrastructure for distributed quantum computing, long-distance quantum communication, and networked quantum sensing. A critical step in realizing a functional quantum network is the efficient and high-fidelity establishment of heralded entanglement between remote quantum nodes. A multimode entangling scheme offers a powerful strategy to accelerate remote entanglement distribution, particularly over long optical fibers. Here, we experimentally demonstrate multimode-enhanced heralded entanglement between two trapped-ion quantum network nodes. By harnessing ten temporal photonic modes, we achieve a 4.59-fold speedup in ion-ion entanglement generation and attain an entanglement fidelity of 95.9%±1.5% over 1.2 km of fiber. Employing a dual-type architecture, our system is readily scalable to multiple nodes, thereby establishing a key building block for future large-scale quantum networks.

k-entanglement, including entanglement relative to full separability and genuinely multipartite entanglement, is a fundamental quantum resource in multipartite quantum systems. Its identification and quantification play essential roles in quantum information processing, quantum simulation, and quantum metrology. However, the practical computation of rigorous k-entanglement measures remains highly challenging due to the need for high-dimensional optimization. In this work, we propose a machine-learning-based surrogate framework for approximating the witness-based k-entanglement measure Ew(k,n). The numerical evaluation of the computationally realized quantity E˜w(k,n)(ρ) is reformulated as a supervised regression problem, where the input is the density matrix ρ and the labels are obtained from finite witness databases. The framework combines multilayer perceptrons (MLPs), convolutional neural networks (CNNs), and light gradient boosting machine (LightGBM) through a stacking ensemble. Numerical experiments are performed for 3- and 4-qubit systems as representative demonstrations of the proposed workflow. The results show that the learned models achieve high predictive accuracy in terms of MAE, MSE, and R2, while providing millisecond-level inference for single-state evaluation. Werner state tests serve as symmetric benchmark checks, and an additional four-qubit noisy circuit-generated state family, obtained from finite-depth circuit preparation followed by local amplitude-damping noise, is used as a structured physical test beyond random density matrices. Compared with the optimization-based evaluation, the trained surrogate model significantly reduces the computational time while maintaining accuracy within the tested system sizes and data distributions. These results show that the proposed framework provides an efficient numerical surrogate for rapid approximation of witness-based k-entanglement measures, while extensions to larger systems and experimental data require further validation.

Pixelless infrared imaging devices based on optoelectronic up-conversion offer a compact and scalable alternative to conventional focal plane arrays; however, their performance is fundamentally limited by lateral carrier diffusion, image smearing, and the trade-off between spatial resolution and conversion efficiency. Existing systems employing quantum well infrared phototransistors (QWIPTs) integrated with light-emitting diodes (LEDs) suffer from degraded modulation transfer function (MTF) at high spatial frequencies and restricted design flexibility. In this article, a quantum cascade detector (QCD)-LED pixelless imaging architecture is proposed and comprehensively modeled as a next-generation alternative. A unified analytical framework is developed to describe carrier concentration, modulation transfer function, image resolution, and image conversion efficiency (ICE) in QCD-LED systems under spatially modulated far-infrared illumination. The models explicitly account for cascade transport, diffusion-drift dynamics, photon recycling, and radiative recombination, enabling direct comparison with conventional QWIPT-LED imagers. Numerical investigations reveal that multi-stage cascade transport significantly suppresses lateral carrier spreading, resulting in a pronounced enhancement in spatial-frequency response. The proposed QCD-LED architecture demonstrates a >32.5% improvement in maximum MTF, a 32.5% increase in conversion efficiency, and a 25% enhancement in response speed, while maintaining comparable or improved image resolution. An optimal performance is achieved for a 10-stage quantum cascade detector with a 2.5 μm period length and a radiative-to-nonradiative lifetime ratio of 0.999, yielding a figure of merit (R × ICE) of 30.99, outperforming previously reported QWIPT-LED systems. Experimental validation confirms excellent agreement with theoretical predictions (R2 = 0.989), particularly at high spatial frequencies where QCD-LED devices exhibit more than 140% improvement in contrast transfer. These results establish quantum cascade detector-based pixelless imagers as a robust platform for high-resolution, high-speed infrared imaging, offering superior spatial fidelity and design flexibility for next-generation optoelectronic imaging systems.

Fourier Neural Operators have become a central tool for learning solution operators of partial differential equations, but their spectral layers remain entirely classical and rely on digital Fourier processing. In this work, we introduce the Continuous-Variable Quantum Fourier Neural Operator (CV-QFNO), a Gaussian photonic formulation of the FNO spectral layer. The proposed architecture maps the essential operations of Fourier-domain operator learning, Fourier transformation, mode selection, and channel mixing, onto native continuous-variable optical primitives. In this way, the CV-QFNO provides a photonic quantum analogue of the truncated spectral mechanism underlying the classical FNO, while avoiding the compilation overhead and spectral mismatch that arise in qubit-based Quantum FNO constructions. We extended the framework to both one- and two-dimensional operator learning and validated it on standard PDE benchmarks, including Burgers' equation, heat equation, Navier-Stokes dynamics, and Darcy flow. The results show that the proposed model preserves the predictive accuracy, resolution generalisation, and spectral inductive bias of Fourier neural operators while using structurally constrained photonic parameterisation. Since all the experiments were performed as classical simulations, the contribution should be understood as an architectural and algorithmic blueprint for photonic neural operators rather than as a demonstration of quantum computational advantage.

Organic memristors are attractive building blocks for neuromorphic computing owing to their intrinsic synaptic functionalities and solution-processability. However, their operational instability remains a major challenge, primarily arising from poorly controlled semiconductor crystallization and stochastic conductive filament formation. Here, we report a high-performance solution-processed organic memristor based on a TIPS-pentacene/PMMA/CdSe-ZnS quantum-dot hybrid system, in which a dual-engineering strategy is employed to simultaneously regulate film crystallization and filament dynamics. Specifically, the PMMA matrix templates the molecular ordering of TIPS-pentacene to improve film uniformity and crystallinity, while CdSe/ZnS quantum dots locally modulate the electric field to direct and confine conductive filament formation. As a result, the device exhibits ultralow and highly uniform switching voltages (0.473 V for set and -0.430 V for reset), suppressed device-to-device variation, long retention exceeding 104 s, and endurance over 1200 switching cycles. In addition, the memristor supports multilevel data storage and successfully emulates key synaptic functions, including long-term potentiation/depression, paired-pulse facilitation, and spike-timing-dependent plasticity. This work provides a materials-level strategy for achieving reliable and low-power organic memristors, offering a viable route toward high-density nonvolatile memory and neuromorphic computing hardware.

We report a quantum energy truncation terahertz coherence tomography technique that enables highly localized extraction of molecular-species-specific THz responses. The method exploits resonant photon loss when the emitter THz energy matches molecular vibration quanta. Reflected THz radiation is recorded and mapped into depth-coded image sequences, from which subsurface structural information and molecular fingerprints are retrieved to construct three-dimensional tomograms. This approach resolves major limitations of conventional continuous-wave (CW) THz imaging-namely, its lack of spectroscopic specificity and dependence on angular scanning-by directly correlating molecular excitation spectra with depth information. As a result, quantum energy truncation terahertz coherence tomography enables rapid, three-dimensional THz imaging with selective sensitivity to discrete molecular quantum energy transitions, providing new opportunities for nondestructive characterization of material refractive indices, molecular composition, and subsurface structures.

As quantum machine learning modules become increasingly integrated into NISQ-era infrastructures, it remains unclear whether intrinsic device noise can be regarded as a passive defense against adversarial examples, or whether it in fact introduces a new attack surface. To answer this question, we propose a noise-aware four-path evaluation protocol that decouples the noise assumed at attack generation from the noise present at inference, and we systematically test it on a 4-qubit variational quantum classifier over four datasets with depolarizing probabilities in the range p∈[0,0.3], using both standard gradient attacks and expectation over transformation (EOT)-based attacks. The results show that for some datasets, higher noise does suppress attacks, whereas for others attacks remain effective even at p=0.3, and in several cases a moderate noise level even maximizes the attack success rate. Moreover, we find that adversarial examples generated under moderate noise often attack the clean model more successfully than those generated in an ideal setting, demonstrating that noise can be actively exploited by an adversary to discover more transferable adversarial directions. Therefore, ambient noise should not be treated as a built-in security guarantee, and future quantum machine learning (QML) robustness evaluations must explicitly model such noise-aware threats.