arXiv subjects

Dong Zhu

Publications and source records attributed to Dong Zhu.

12 recordsLinked to original sources

arXiv2026

ARB4WM: An Adversarial Robustness Benchmark for World Models in Continuous Control

World models are widely used in robotic and agentic engineering control systems due to their ability to learn latent dynamics for planning and decision-making. As these systems are increasingly deployed in safety-critical settings, understanding their robustness under adversarial conditions has become essential. However, existing evaluations lack a unified benchmark for testing adversarial threats across the policy, value, and latent-dynamics levels of world-model agents. To fill this gap, we present ARB4WM, a unified evaluation framework for pre-deployment robustness and risk assessment of world-model agents under visual perturbations. ARB4WM defines five white-box loss objectives across these three levels and studies their effects when combined with single-step or multi-step perturbation strategies and temporal attack modes, including full-frame, half-sequence, and sparse-frame exposure. Specifically, we evaluate four Dreamer-style agents across 20 tasks from MetaWorld and the DeepMind Control Suite under different loss objectives, perturbation strategies, and temporal attack modes. Results show that attacks targeting value estimation, latent representations, and RSSM dynamics can be as damaging as direct policy disruption, and that early or frequent perturbations are especially harmful, while input-level defenses provide limited recovery under adaptive attacks. These findings suggest that safety, risk, and reliability assessment for world models should cover multiple component-oriented attack objectives and temporal exposure protocols rather than relying solely on action-space robustness. Source code is available at https://github.com/zaoanguai/ARB4WM.

cs.AI↗

arXiv2026

A Conditional Denoising Diffusion Probabilistic Model for RFI Mitigation in Synthetic Aperture Interferometric Radiometer

In Earth remote sensing, spatial-frequency domain visibility samples are inversely transformed into spatial-domain brightness temperature (BT) images through the signal processing pipeline of synthetic aperture interferometric radiometers (SAIR). However, L-band radio-frequency interference (RFI) contaminates the measured visibilities and severely degrades BT image quality, thereby impairing geophysical parameter retrieval. To address this issue, we propose VFDM, a Visibility-Function Diffusion Model based on Denoising Diffusion Probabilistic Models (DDPM), to mitigate RFI in the spatial-frequency domain while preserving fine-scale structures consistent with natural scene statistics. Furthermore, we construct a comprehensive dataset comprising more than ten thousand pairs of RFI-free natural scene visibility sample sets and their corresponding simulated contaminated counterparts, categorized by varying RFI intensities, numbers, and distributions. Finally, comprehensive experiments on both simulated and real-world data demonstrate the effectiveness and robustness of the proposed VFDM-based approach.

eess.SP↗

arXiv2025

Magnetic-field dependent vortex dynamics and critical currents in superconducting microwires with regular large-area perforation by pinholes

We report on results of simulations and experiments of vortex states in superconducting micro-wires with periodic rectangular pinhole structures. The simulations have been performed by means of numerically solving the time-dependent Ginzburg-Landau (TDGL) equations. With increasing bias current and for different values of the external magnetic field applied normal to the structure plane, we observe at first a vortex free Meissner state, followed by a resistive vortex-flow mixed state and a state with a more complex vortex pattern. The resulting dependence of the critical current Ic on magnetic field exhibits two plateaus with distinctly different vortex dynamics. Corresponding experimentally measured magnetic-field dependences of Ic of WSi microwires with periodic pinhole structures and varying hole spacing confirmed the predictions of these simulations, showing two ranges of magnetic field with almost field-independent critical currents. The experimentally determined critical currents are larger for a smaller pinhole spacing, in agreement with the results of the TDGL simulations. The good agreement of the simulations with the experimental results provides a convenient strategy for the optimization of single-photon detectors with or without artificial and natural defects.

cond-mat.supr-con↗

arXiv2025

PNE-SGAN: Probabilistic NDT-Enhanced Semantic Graph Attention Network for LiDAR Loop Closure Detection

LiDAR loop closure detection (LCD) is crucial for consistent Simultaneous Localization and Mapping (SLAM) but faces challenges in robustness and accuracy. Existing methods, including semantic graph approaches, often suffer from coarse geometric representations and lack temporal robustness against noise, dynamics, and viewpoint changes. We introduce PNE-SGAN, a Probabilistic NDT-Enhanced Semantic Graph Attention Network, to overcome these limitations. PNE-SGAN enhances semantic graphs by using Normal Distributions Transform (NDT) covariance matrices as rich, discriminative geometric node features, processed via a Graph Attention Network (GAT). Crucially, it integrates graph similarity scores into a probabilistic temporal filtering framework (modeled as an HMM/Bayes filter), incorporating uncertain odometry for motion modeling and utilizing forward-backward smoothing to effectively handle ambiguities. Evaluations on challenging KITTI sequences (00 and 08) demonstrate state-of-the-art performance, achieving Average Precision of 96.2\% and 95.1\%, respectively. PNE-SGAN significantly outperforms existing methods, particularly in difficult bidirectional loop scenarios where others falter. By synergizing detailed NDT geometry with principled probabilistic temporal reasoning, PNE-SGAN offers a highly accurate and robust solution for LiDAR LCD, enhancing SLAM reliability in complex, large-scale environments.

cs.CV↗

arXiv2023

Addressing the Accuracy-Cost Tradeoff in Material Property Prediction: A Teacher-Student Strategy

Deep learning has revolutionized the process of new material discovery, with state-of-the-art models now able to predict material properties based solely on chemical compositions, thus eliminating the necessity for material structures. However, this cost-effective method has led to a trade-off in model accuracy. Specifically, the accuracy of Chemical Composition-based Property Prediction Models (CPMs) significantly lags behind that of Structure-based Property Prediction Models (SPMs). To tackle this challenge, we propose an innovative Teacher-Student (T-S) strategy, where a pre-trained SPM serves as the 'teacher' to enhance the accuracy of the CPM. Leveraging the T-S strategy, T-S CrabNet has risen to become the most accurate model among current CPMs. Initially, we demonstrated the universality of this strategy. On the Materials Project (MP) and Jarvis datasets, we validated the effectiveness of the T-S strategy in boosting the accuracy of CPMs with two distinct network structures, namely CrabNet and Roost. This led to CrabNet, under the guidance of the T-S strategy, emerging as the most accurate model among the current CPMs. Moreover, this strategy shows remarkable efficacy in small datasets. When predicting the formation energy on a small MP dataset comprising merely 5% of the samples, the T-S strategy boosted CrabNet's accuracy by 37.1%, exceeding the enhancement effect of the T-S strategy on the whole dataset.

cond-mat.mtrl-sci↗

arXiv2023

Effective suppression of dark counts in superconducting microstructures with grid of pinholes in a magnetic field

In a magnetic field, vortices significantly contribute to the dark counts of single-photon detectors made of superconducting wires, and they are also limiting the critical current of such devices. To address this issue, we prepared superconducting microwires with a pinhole grid from WSi thin films and report on corresponding critical-current and count-rate measurements in an external magnetic field B. When compared to corresponding devices without pinholes, the critical current only weakly depends on the magnetic field at B < 16 mT and it is even larger already at B > 10 mT. Moreover, dark counts are not only suppressed in zero field, but particularly in magnetic fields B < 16 mT, while photon counts are virtually field insensitive in the same range of the magnetic field.

cond-mat.supr-con↗

arXiv2022

An Unsupervised Short- and Long-Term Mask Representation for Multivariate Time Series Anomaly Detection

Anomaly detection of multivariate time series is meaningful for system behavior monitoring. This paper proposes an anomaly detection method based on unsupervised Short- and Long-term Mask Representation learning (SLMR). The main idea is to extract short-term local dependency patterns and long-term global trend patterns of the multivariate time series by using multi-scale residual dilated convolution and Gated Recurrent Unit(GRU) respectively. Furthermore, our approach can comprehend temporal contexts and feature correlations by combining spatial-temporal masked self-supervised representation learning and sequence split. It considers the importance of features is different, and we introduce the attention mechanism to adjust the contribution of each feature. Finally, a forecasting-based model and a reconstruction-based model are integrated to focus on single timestamp prediction and latent representation of time series. Experiments show that the performance of our method outperforms other state-of-the-art models on three real-world datasets. Further analysis shows that our method is good at interpretability.

cs.LG↗

arXiv2021

Physical properties of amorphous molybdenum silicide films for single photon detectors

We systematically investigated the physical properties of amorphous Mo$_{\rm x}$Si$_{1-x}$ films deposited by the magnetron co-sputtering technique. The critical temperature $T_C$ of Mo$_{\rm x}$ Si$_{1-x}$ films increases gradually with the stoichiometry x, and the highest $T_C$=7.9 K was found in Mo$_{\rm 0.83}$ Si$_{0.17}$. Beyond $x$=0.83, preformed Cooper pairs and superconducting domains persist in the films, despite the superconducting state with perfect zero-resistivity is absent. The thick films of Mo$_{\rm 0.83}$ Si$_{0.17}$ show surprising degradation in which the onset of zero-resistivity is suppressed below 2 K. The thin Mo$_{\rm 0.83}$ Si$_{0.17}$ films, however, reveal robust superconductivity even with thickness d$\leq$1 nm. We also characterized wide microwires based on the 2 nm thin Mo$_{\rm 0.8}$ Si$_{0.2}$ films with widths 40 and 60 $μ$m, which show single-photon sensitivity at 780 nm and 1550 nm wavelength

cond-mat.supr-con↗

arXiv2020

Strong suppression of the resistivity near the transition to superconductivity in narrow micro-bridges in external magnetic fields

We have investigated a series of superconducting bridges based on homogeneous amorphous WSi and MoSi films, with bridge widths w ranging from 2 um to 1000 um and film thicknesses d ~ 4-6 nm and 100 nm. Upon decreasing the bridge widths below the respective Pearl lengths, we observe in all cases distinct changes in the characteristics of the resistive transitions to superconductivity. For each of the films, the resistivity curves R(B,T) separate at a well-defined and field-dependent temperature T*(B) with decreasing the temperature, resulting in a dramatic suppression of the resistivity and a sharpening of the transitions with decreasing bridge width w. The associated excess conductivity in all the bridges scales as 1/w, which may suggest the presence of a highly conducting region that is dominating the electric transport in narrow bridges. We argue that this effect can only be observed in materials with sufficiently weak vortex pinning.

cond-mat.supr-con↗

arXiv2017

Coherent phonon Rabi oscillations with a high frequency carbon nanotube phonon cavity

Phonon-cavity electromechanics allows the manipulation of mechanical oscillations similar to photon-cavity systems. Many advances on this subject have been achieved in various materials. In addition, the coherent phonon transfer (phonon Rabi oscillations) between the phonon cavity mode and another oscillation mode has attracted many interest in nano-science. Here we demonstrate coherent phonon transfer in a carbon nanotube phonon-cavity system with two mechanical modes exhibiting strong dynamical coupling. The gate-tunable phonon oscillation modes are manipulated and detected by extending the red-detuned pump idea of photonic cavity electromechanics. The first- and second-order coherent phonon transfers are observed with Rabi frequencies 591 kHz and 125 kHz, respectively. The frequency quality factor product fQ_m~2=10^12 Hz achieved here is larger thank k_B T_base/h, which may enable the future realization of Rabi oscillations in the quantum regime.

cond-mat.mes-hall↗

arXiv2016

Parametric strong mode-coupling in carbon nanotube mechanical resonators

Carbon nanotubes (CNTs) have attracted much attention for use in nanomechanical devices because of their exceptional properties, such as large resonant frequencies, low mass, and high quality factors. Here, we report the first experimental realization of parametric strong coupling between two mechanical modes on a single CNT nanomechanical resonator, by applying an extra microwave pump. This parametric pump method can be used to couple mechanical modes with arbitrary frequency differences. The properties of the mechanical resonator are detected by single-electron tunneling at low temperature, which is found to be strongly coupled to both modes. The coupling strength between the two modes can be tuned by the pump power, setting the coupling regime from weak to strong. This tunability may be useful in further phonon manipulations in carbon nanotubes.

cond-mat.mes-hall↗

arXiv2016

Strongly-coupled nanotube electromechanical resonators

Coupling an electromechanical resonator with carbon-nanotube quantum dots is a significant method to control both the electronic charge and the spin quantum states. By exploiting a novel micro-transfer technique, we fabricate two strongly-coupled and electrically-tunable mechanical resonators on a single carbon nanotube for the first time. The frequency of the two resonators can be individually tuned by the bottom gates, and strong coupling is observed between the electron charge and phonon modes of each resonator. Furthermore, the conductance of either resonator can be nonlocally modulated by the phonon modes in the other resonator. Strong coupling is observed between the phonon modes of the two resonators, which provides an effective long distance electron-electron interaction. The generation of phonon-mediated-spin entanglement is also theoretically analyzed for the two resonators. This strongly-coupled nanotube electromechanical resonator array provides an experimental platform for future studies of the coherent electron-phonon interaction, the phonon mediated long-distance electron interaction, and entanglement state generation.

cond-mat.mes-hall↗