arXiv subjects

Zhe Liu

Publications and source records attributed to Zhe Liu.

At least 145 records · Page 8Linked to original sources

arXiv2025

MReg: A Novel Regression Model with MoE-based Video Feature Mining for Mitral Regurgitation Diagnosis

Color Doppler echocardiography is a crucial tool for diagnosing mitral regurgitation (MR). Recent studies have explored intelligent methods for MR diagnosis to minimize user dependence and improve accuracy. However, these approaches often fail to align with clinical workflow and may lead to suboptimal accuracy and interpretability. In this study, we introduce an automated MR diagnosis model (MReg) developed on the 4-chamber cardiac color Doppler echocardiography video (A4C-CDV). It follows comprehensive feature mining strategies to detect MR and assess its severity, considering clinical realities. Our contribution is threefold. First, we formulate the MR diagnosis as a regression task to capture the continuity and ordinal relationships between categories. Second, we design a feature selection and amplification mechanism to imitate the sonographer's diagnostic logic for accurate MR grading. Third, inspired by the Mixture-of-Experts concept, we introduce a feature summary module to extract the category-level features, enhancing the representational capacity for more accurate grading. We trained and evaluated our proposed MReg on a large in-house A4C-CDV dataset comprising 1868 cases with three graded regurgitation labels. Compared to other weakly supervised video anomaly detection and supervised classification methods, MReg demonstrated superior performance in MR diagnosis. Our code is available at: https://github.com/cskdstz/MReg.

cs.CV↗

arXiv2025

Hierarchical Corpus-View-Category Refinement for Carotid Plaque Risk Grading in Ultrasound

Accurate carotid plaque grading (CPG) is vital to assess the risk of cardiovascular and cerebrovascular diseases. Due to the small size and high intra-class variability of plaque, CPG is commonly evaluated using a combination of transverse and longitudinal ultrasound views in clinical practice. However, most existing deep learning-based multi-view classification methods focus on feature fusion across different views, neglecting the importance of representation learning and the difference in class features. To address these issues, we propose a novel Corpus-View-Category Refinement Framework (CVC-RF) that processes information from Corpus-, View-, and Category-levels, enhancing model performance. Our contribution is four-fold. First, to the best of our knowledge, we are the foremost deep learning-based method for CPG according to the latest Carotid Plaque-RADS guidelines. Second, we propose a novel center-memory contrastive loss, which enhances the network's global modeling capability by comparing with representative cluster centers and diverse negative samples at the Corpus level. Third, we design a cascaded down-sampling attention module to fuse multi-scale information and achieve implicit feature interaction at the View level. Finally, a parameter-free mixture-of-experts weighting strategy is introduced to leverage class clustering knowledge to weight different experts, enabling feature decoupling at the Category level. Experimental results indicate that CVC-RF effectively models global features via multi-level refinement, achieving state-of-the-art performance in the challenging CPG task.

cs.CV↗

arXiv2025

Physics-informed generative real-time lens-free imaging

Advancements in high-throughput biomedical applications require real-time, large field-of-view (FOV) imaging. While current 2D lens-free imaging (LFI) systems improve FOV, they are often hindered by time-consuming multi-position measurements, extensive data pre-processing, and strict optical parameterization, limiting their application to static, thin samples. To overcome these limitations, we introduce GenLFI, combining a generative unsupervised physics-informed neural network (PINN) with a large FOV LFI setup for straightforward holographic image reconstruction, without multi-measurement. GenLFI enables real-time 2D imaging for 3D samples, such as droplet-based microfluidics and 3D cell models, in dynamic complex optical fields. Unlike previous methods, our approach decouples the reconstruction algorithm from optical setup parameters, enabling a large FOV limited only by hardware. We demonstrate a real-time FOV exceeding 550 mm$^2$, over 20 times larger than current real-time LFI systems. This framework unlocks the potential of LFI systems, providing a robust tool for advancing automated high-throughput biomedical applications.

physics.optics↗

arXiv2025

RationalVLA: A Rational Vision-Language-Action Model with Dual System

A fundamental requirement for real-world robotic deployment is the ability to understand and respond to natural language instructions. Existing language-conditioned manipulation tasks typically assume that instructions are perfectly aligned with the environment. This assumption limits robustness and generalization in realistic scenarios where instructions may be ambiguous, irrelevant, or infeasible. To address this problem, we introduce RAtional MAnipulation (RAMA), a new benchmark that challenges models with both unseen executable instructions and defective ones that should be rejected. In RAMA, we construct a dataset with over 14,000 samples, including diverse defective instructions spanning six dimensions: visual, physical, semantic, motion, safety, and out-of-context. We further propose the Rational Vision-Language-Action model (RationalVLA). It is a dual system for robotic arms that integrates the high-level vision-language model with the low-level manipulation policy by introducing learnable latent space embeddings. This design enables RationalVLA to reason over instructions, reject infeasible commands, and execute manipulation effectively. Experiments demonstrate that RationalVLA outperforms state-of-the-art baselines on RAMA by a 14.5% higher success rate and 0.94 average task length, while maintaining competitive performance on standard manipulation tasks. Real-world trials further validate its effectiveness and robustness in practical applications. Our project page is https://irpn-eai.github.io/RationalVLA.

cs.RO↗

arXiv2025

AdInject: Real-World Black-Box Attacks on Web Agents via Advertising Delivery

Vision-Language Model (VLM) based Web Agents represent a significant step towards automating complex tasks by simulating human-like interaction with websites. However, their deployment in uncontrolled web environments introduces significant security vulnerabilities. Existing research on adversarial environmental injection attacks often relies on unrealistic assumptions, such as direct HTML manipulation, knowledge of user intent, or access to agent model parameters, limiting their practical applicability. In this paper, we propose AdInject, a novel and real-world black-box attack method that leverages the internet advertising delivery to inject malicious content into the Web Agent's environment. AdInject operates under a significantly more realistic threat model than prior work, assuming a black-box agent, static malicious content constraints, and no specific knowledge of user intent. AdInject includes strategies for designing malicious ad content aimed at misleading agents into clicking, and a VLM-based ad content optimization technique that infers potential user intents from the target website's context and integrates these intents into the ad content to make it appear more relevant or critical to the agent's task, thus enhancing attack effectiveness. Experimental evaluations demonstrate the effectiveness of AdInject, attack success rates exceeding 60% in most scenarios and approaching 100% in certain cases. This strongly demonstrates that prevalent advertising delivery constitutes a potent and real-world vector for environment injection attacks against Web Agents. This work highlights a critical vulnerability in Web Agent security arising from real-world environment manipulation channels, underscoring the urgent need for developing robust defense mechanisms against such threats. Our code is available at https://github.com/NicerWang/AdInject.

cs.CR↗

arXiv2025

Phonon-limited carrier transport in the Weyl semimetal TaAs

Topological Weyl semimetals represent a novel class of quantum materials that exhibit remarkable properties arising from their unique electronic structure. In this work, we employ state-of-the-art ab initio methods to investigate the role of the electron-phonon interactions on the charge transport properties of TaAs. Our calculations of the temperature-dependent electrical conductivity with the iterative Boltzmann transport equation show excellent agreement with experimental measurements above 100 K. Extending the analysis to doped systems, we demonstrate that even small shifts in the Fermi level can lead to substantial changes in conductivity, driven by the complex topology of the Fermi surface. In particular, modifications in Fermi surface nesting emerge as a key factor influencing scattering processes and carrier lifetimes. These findings offer critical insights into the microscopic mechanisms that govern transport in TaAs and highlight the sensitivity of Weyl semimetals to doping and carrier dynamics.

cond-mat.mtrl-sci↗

arXiv2025

Adversarial Attacks of Vision Tasks in the Past 10 Years: A Survey

With the advent of Large Vision-Language Models (LVLMs), new attack vectors, such as cognitive bias, prompt injection, and jailbreaking, have emerged. Understanding these attacks promotes system robustness improvement and neural networks demystification. However, existing surveys often target attack taxonomy and lack in-depth analysis like 1) unified insights into adversariality, transferability, and generalization; 2) detailed evaluations framework; 3) motivation-driven attack categorizations; and 4) an integrated perspective on both traditional and LVLM attacks. This article addresses these gaps by offering a thorough summary of traditional and LVLM adversarial attacks, emphasizing their connections and distinctions, and providing actionable insights for future research.

cs.CV↗

arXiv2025

Pangu Ultra MoE: How to Train Your Big MoE on Ascend NPUs

Sparse large language models (LLMs) with Mixture of Experts (MoE) and close to a trillion parameters are dominating the realm of most capable language models. However, the massive model scale poses significant challenges for the underlying software and hardware systems. In this paper, we aim to uncover a recipe to harness such scale on Ascend NPUs. The key goals are better usage of the computing resources under the dynamic sparse model structures and materializing the expected performance gain on the actual hardware. To select model configurations suitable for Ascend NPUs without repeatedly running the expensive experiments, we leverage simulation to compare the trade-off of various model hyperparameters. This study led to Pangu Ultra MoE, a sparse LLM with 718 billion parameters, and we conducted experiments on the model to verify the simulation results. On the system side, we dig into Expert Parallelism to optimize the communication between NPU devices to reduce the synchronization overhead. We also optimize the memory efficiency within the devices to further reduce the parameter and activation management overhead. In the end, we achieve an MFU of 30.0% when training Pangu Ultra MoE, with performance comparable to that of DeepSeek R1, on 6K Ascend NPUs, and demonstrate that the Ascend system is capable of harnessing all the training stages of the state-of-the-art language models. Extensive experiments indicate that our recipe can lead to efficient training of large-scale sparse language models with MoE. We also study the behaviors of such models for future reference.

cs.CL↗

arXiv2025

Uncertainty-Guided Self-Questioning and Answering for Video-Language Alignment

The development of multi-modal models has been rapidly advancing, with some demonstrating remarkable capabilities. However, annotating video-text pairs remains expensive and insufficient. Take video question answering (VideoQA) tasks as an example, human annotated questions and answers often cover only part of the video, since the corresponding text is often short and monotonous, leading to underutilization of video. To address this, we propose a Bootstrapping Video-Language Alignment framework (BoViLA), a self-training method that augments question samples during training process through LLM-based self-questioning and answering, which help model exploit video information and the internal knowledge of LLMs more thoroughly to improve modality alignment. However, low-quality self-generated questions may instead contaminate the performance, especially in the early stages of training, as we have observed in our experiments. To filter bad self-generated questions, we introduce Evidential Deep Learning (EDL) to estimate uncertainty and assess the quality of self-generated questions by evaluating the modality alignment within the context. To the best of our knowledge, this work is the first to explore LLM-based self-training frameworks for modality alignment. We evaluate BoViLA on five strong VideoQA benchmarks, where it outperforms several state-of-the-art methods and demonstrate its effectiveness and generality. Additionally, we provide extensive analyses of the self-training framework and the EDL-based uncertainty filtering mechanism. The code will be made available.

cs.CV↗

arXiv2025

Electrospray Thruster Plume Dynamics: Insights from Precise PP Coulomb Field Simulation

Electrospray thrusters are one important type of micropropulsion systems being developed for next-generation space missions, yet the primary challenge to their operational lifespan is propellant overspray resulting from wide plume angles driven by Coulomb interactions among charged droplets. While existing models often employ truncated Coulomb field approximations, such simplifications compromise accuracy in predicting divergence dynamics. In this study, a particle-particle (PP) simulation method is used to directly calculate the interactions between droplets in an electrospray plume coupled with background electric field effects for simulation. The model integrates Boris pusher for numerical integration, validated through binary collision tests verification. Parametric analysis systematically evaluates six key variables, droplet charge, droplet mass, emission interval, droplet initial velocity, and electric field components, to quantify their impacts on plume divergence. The shape of the simulated electrospray plume and the velocity of the droplets in it are analyzed. Parametric analysis demonstrate that reducing droplet charge, increasing droplet mass, extending emission time intervals, and elevating initial drift velocity collectively reduce plume half-angle. These results quantitatively establish parameter plume relationships, providing direct guidance for thruster optimization.

physics.plasm-ph↗

arXiv2025

BEV-GS: Feed-forward Gaussian Splatting in Bird's-Eye-View for Road Reconstruction

Road surface is the sole contact medium for wheels or robot feet. Reconstructing road surface is crucial for unmanned vehicles and mobile robots. Recent studies on Neural Radiance Fields (NeRF) and Gaussian Splatting (GS) have achieved remarkable results in scene reconstruction. However, they typically rely on multi-view image inputs and require prolonged optimization times. In this paper, we propose BEV-GS, a real-time single-frame road surface reconstruction method based on feed-forward Gaussian splatting. BEV-GS consists of a prediction module and a rendering module. The prediction module introduces separate geometry and texture networks following Bird's-Eye-View paradigm. Geometric and texture parameters are directly estimated from a single frame, avoiding per-scene optimization. In the rendering module, we utilize grid Gaussian for road surface representation and novel view synthesis, which better aligns with road surface characteristics. Our method achieves state-of-the-art performance on the real-world dataset RSRD. The road elevation error reduces to 1.73 cm, and the PSNR of novel view synthesis reaches 28.36 dB. The prediction and rendering FPS is 26, and 2061, respectively, enabling high-accuracy and real-time applications. The code will be available at: \href{https://github.com/cat-wwh/BEV-GS}{\texttt{https://github.com/cat-wwh/BEV-GS}}

cs.GR↗

arXiv2025

Pangu Ultra: Pushing the Limits of Dense Large Language Models on Ascend NPUs

We present Pangu Ultra, a Large Language Model (LLM) with 135 billion parameters and dense Transformer modules trained on Ascend Neural Processing Units (NPUs). Although the field of LLM has been witnessing unprecedented advances in pushing the scale and capability of LLM in recent years, training such a large-scale model still involves significant optimization and system challenges. To stabilize the training process, we propose depth-scaled sandwich normalization, which effectively eliminates loss spikes during the training process of deep models. We pre-train our model on 13.2 trillion diverse and high-quality tokens and further enhance its reasoning capabilities during post-training. To perform such large-scale training efficiently, we utilize 8,192 Ascend NPUs with a series of system optimizations. Evaluations on multiple diverse benchmarks indicate that Pangu Ultra significantly advances the state-of-the-art capabilities of dense LLMs such as Llama 405B and Mistral Large 2, and even achieves competitive results with DeepSeek-R1, whose sparse model structure contains much more parameters. Our exploration demonstrates that Ascend NPUs are capable of efficiently and effectively training dense models with more than 100 billion parameters. Our model and system will be available for our commercial customers.

cs.CL↗

arXiv2025

Nonadiabatic Dynamics with Constrained Nuclear-Electronic Orbital Theory

Incorporating nuclear quantum effects into nonadiabatic dynamics remains a significant challenge. Herein we introduce new nonadiabatic dynamics approaches based on the recently developed constrained nuclear-electronic orbital (CNEO) theory. The CNEO approach integrates nuclear quantum effects, particularly quantum nuclear delocalization effects, into effective potential energy surfaces. When combined with Ehrenfest dynamics and surface hopping, it efficiently captures both nonadiabaticity and quantum nuclear delocalization effects. We apply these new approaches to a one-dimensional proton-coupled electron transfer model and find that they outperform conventional Ehrenfest dynamics and surface hopping in predicting proton transfer dynamics and proton transmission probabilities, especially in the low momentum regime.

physics.chem-ph↗

arXiv2025

MovSAM: A Single-image Moving Object Segmentation Framework Based on Deep Thinking

Moving object segmentation plays a vital role in understanding dynamic visual environments. While existing methods rely on multi-frame image sequences to identify moving objects, single-image MOS is critical for applications like motion intention prediction and handling camera frame drops. However, segmenting moving objects from a single image remains challenging for existing methods due to the absence of temporal cues. To address this gap, we propose MovSAM, the first framework for single-image moving object segmentation. MovSAM leverages a Multimodal Large Language Model (MLLM) enhanced with Chain-of-Thought (CoT) prompting to search the moving object and generate text prompts based on deep thinking for segmentation. These prompts are cross-fused with visual features from the Segment Anything Model (SAM) and a Vision-Language Model (VLM), enabling logic-driven moving object segmentation. The segmentation results then undergo a deep thinking refinement loop, allowing MovSAM to iteratively improve its understanding of the scene context and inter-object relationships with logical reasoning. This innovative approach enables MovSAM to segment moving objects in single images by considering scene understanding. We implement MovSAM in the real world to validate its practical application and effectiveness for autonomous driving scenarios where the multi-frame methods fail. Furthermore, despite the inherent advantage of multi-frame methods in utilizing temporal information, MovSAM achieves state-of-the-art performance across public MOS benchmarks, reaching 92.5\% on J\&F. Our implementation will be available at https://github.com/IRMVLab/MovSAM.

cs.CV↗

arXiv2025

Learning-enhanced electronic skin for tactile sensing on deformable surface based on electrical impedance tomography

Electrical Impedance Tomography (EIT)-based tactile sensors offer cost-effective and scalable solutions for robotic sensing, especially promising for soft robots. However a major issue of EIT-based tactile sensors when applied in highly deformable objects is their performance degradation due to surface deformations. This limitation stems from their inherent sensitivity to strain, which is particularly exacerbated in soft bodies, thus requiring dedicated data interpretation to disentangle the parameter being measured and the signal deriving from shape changes. This has largely limited their practical implementations. This paper presents a machine learning-assisted tactile sensing approach to address this challenge by tracking surface deformations and segregating this contribution in the signal readout during tactile sensing. We first capture the deformations of the target object, followed by tactile reconstruction using a deep learning model specifically designed to process and fuse EIT data and deformation information. Validations using numerical simulations achieved high correlation coefficients (0.9660 - 0.9999), peak signal-to-noise ratios (28.7221 - 55.5264 dB) and low relative image errors (0.0107 - 0.0805). Experimental validations, using a hydrogel-based EIT e-skin under various deformation scenarios, further demonstrated the effectiveness of the proposed approach in real-world settings. The findings could underpin enhanced tactile interaction in soft and highly deformable robotic applications.

cs.RO↗

arXiv2025

FedCAPrivacy: Privacy-Preserving Heterogeneous Federated Learning with Anonymous Adaptive Clustering

Federated learning (FL) is a distributed machine learning paradigm enabling multiple clients to train a model collaboratively without exposing their local data. Among FL schemes, clustering is an effective technique addressing the heterogeneity issue (i.e., differences in data distribution and computational ability affect training performance and effectiveness) via grouping participants with similar computational resources or data distribution into clusters. However, intra-cluster data exchange poses privacy risks, while cluster selection and adaptation introduce challenges that may affect overall performance. To address these challenges, this paper introduces anonymous adaptive clustering, a novel approach that simultaneously enhances privacy protection and boosts training efficiency. Specifically, an oblivious shuffle-based anonymization method is designed to safeguard user identities and prevent the aggregation server from inferring similarities through clustering. Additionally, to improve performance, we introduce an iteration-based adaptive frequency decay strategy, which leverages variability in clustering probabilities to optimize training dynamics. With these techniques, we build the FedCAPrivacy; experiments show that FedCAPrivacy achieves ~7X improvement in terms of performance while maintaining high privacy.

cs.CR↗

arXiv2025

On-Off Systems with Strategic Customers

Motivated by applications such as urban traffic control and make-to-order systems, we study a fluid model of a single-server, on-off system that can accommodate multiple queues. The server visits each queue in order: when a queue is served, it is "on", and when the server is serving another queue or transitioning between queues, it is "off". Customers arrive over time, observe the state of the system, and decide whether to join. We consider two regimes for the formation of the on and off durations. In the exogenous setting, each queue's on and off durations are predetermined. We explicitly characterize the equilibrium outcome in closed form and give a compact linear program to compute the optimal on-off durations that maximizes total reward collected from serving customers. In the endogenous setting, the durations depend on customers' joining decisions under an exhaustive service policy where the server never leaves a non-empty queue. We show that an optimal policy in this case extends service beyond the first clearance for at most one queue. Using this property, we introduce a closed-form procedure that computes an optimal policy in no more than 2n steps for a system with n queues.

cs.GT↗

arXiv2025

Non-Monotonic Attention-based Read/Write Policy Learning for Simultaneous Translation

Simultaneous or streaming machine translation generates translation while reading the input stream. These systems face a quality/latency trade-off, aiming to achieve high translation quality similar to non-streaming models with minimal latency. We propose an approach that efficiently manages this trade-off. By enhancing a pretrained non-streaming model, which was trained with a seq2seq mechanism and represents the upper bound in quality, we convert it into a streaming model by utilizing the alignment between source and target tokens. This alignment is used to learn a read/write decision boundary for reliable translation generation with minimal input. During training, the model learns the decision boundary through a read/write policy module, employing supervised learning on the alignment points (pseudo labels). The read/write policy module, a small binary classification unit, can control the quality/latency trade-off during inference. Experimental results show that our model outperforms several strong baselines and narrows the gap with the non-streaming baseline model.

cs.CL↗