arXiv subjects

Jun Zhang

Publications and source records attributed to Jun Zhang.

At least 379 records · Page 21Linked to original sources

arXiv2025

HMI: Hierarchical Knowledge Management for Efficient Multi-Tenant Inference in Pretrained Language Models

The significant computational demands of pretrained language models (PLMs), which often require dedicated hardware, present a substantial challenge in serving them efficiently, especially in multi-tenant environments. To address this, we introduce HMI, a Hierarchical knowledge management-based Multi-tenant Inference system, designed to manage tenants with distinct PLMs resource-efficiently. Our approach is three-fold: Firstly, we categorize PLM knowledge into general, domain-specific, and task-specific. Leveraging insights on knowledge acquisition across different model layers, we construct hierarchical PLMs (hPLMs) by extracting and storing knowledge at different levels, significantly reducing GPU memory usage per tenant. Secondly, we establish hierarchical knowledge management for hPLMs generated by various tenants in HMI. We manage domain-specific knowledge with acceptable storage increases by constructing and updating domain-specific knowledge trees based on frequency. We manage task-specific knowledge within limited GPU memory through parameter swapping. Finally, we propose system optimizations to enhance resource utilization and inference throughput. These include fine-grained pipelining via hierarchical knowledge prefetching to overlap CPU and I/O operations with GPU computations, and optimizing parallel implementations with batched matrix multiplications. Our experimental results demonstrate that the proposed HMI can efficiently serve up to 10,000 hPLMs (hBERTs and hGPTs) on a single GPU, with only a negligible compromise in accuracy.

cs.LG↗

arXiv2025

Advancing CMA-ES with Learning-Based Cooperative Coevolution for Scalable Optimization

Recent research in Cooperative Coevolution~(CC) have achieved promising progress in solving large-scale global optimization problems. However, existing CC paradigms have a primary limitation in that they require deep expertise for selecting or designing effective variable decomposition strategies. Inspired by advancements in Meta-Black-Box Optimization, this paper introduces LCC, a pioneering learning-based cooperative coevolution framework that dynamically schedules decomposition strategies during optimization processes. The decomposition strategy selector is parameterized through a neural network, which processes a meticulously crafted set of optimization status features to determine the optimal strategy for each optimization step. The network is trained via the Proximal Policy Optimization method in a reinforcement learning manner across a collection of representative problems, aiming to maximize the expected optimization performance. Extensive experimental results demonstrate that LCC not only offers certain advantages over state-of-the-art baselines in terms of optimization effectiveness and resource consumption, but it also exhibits promising transferability towards unseen problems.

cs.LG↗

arXiv2025

Study of neutron-deficient nucleus 224Np and its α decay by particle-number-conserving method in the framework of deformed shell model

The particle-number-conserving (PNC) method in the framework of the deformed shell model (DSM) is employed to study the properties of the newly discovered short-lived neutron-deficient nucleus 224Np and its α-decay. The calculated energy of α-particle lies within 300 keV of the experimental data. This is the first application of the PNC method to the region of neutron-deficient nuclei. This work provides the first attempt to combine the microscopic PNC theory with empirical formulas to study the nuclear α decay. The configurations of ground states are assigned as π5/2-[523]{\otimes}ν5/2+[633] for 224Np, π1/2-[530]{\otimes}ν3/2+[642] for 220Pa, π3/2+[651]{\otimes}ν1/2-[501] for 216Ac, and π7/2-[514]{\otimes}ν3/2-[501] for 212Fr. The absence of the Z = 92 subshell closure in 224Np is explained by analyzing the proton single-particle levels. Low-lying excited statesare predicted for nuclei along the α-decay chain by the PNC method. Based on the PNC predicted α-decay energy and the assigned configurations, the α-decay half-lives are calculated by the empirical formulas, in which the angular momentum taken away by the α particle is taken into account. The angular momentum have an important effect on the α-decay half-life. The errors of the empirical formulas calculation are in two orders of magnitude with the experimental data.

nucl-th↗

arXiv2025

Pulmonary electrical impedance tomography based on deep recurrent neural networks

Electrical impedance tomography (EIT) is a non-invasive functional imaging technology. In order to enhance the quality of lung EIT images, novel algorithms, namely LSTM-LSTM, LSTM-BiLSTM, BiLSTM-LSTM, and BiLSTM-BiLSTM, leveraging LSTM or BiLSTM networks, were developed. Simulation results demonstrate that the optimized deep recurrent neural network significantly enhanced the quality of the reconstructed images. Specifically, the correlation coefficients of the LSTM-LSTM and the LSTM-BiLSTM algorithms exhibited maximum increases of 27.5% and 25.4% over the LSTM algorithm, respectively. Moreover, in comparison to the BiLSTM algorithm, the correlation coefficients of the BiLSTM-LSTM and BiLSTM-BiLSTM algorithms increased by 11.7% and 13.4%, respectively. Overall, the quality of EIT images showed notable enhancement. This research offers a valuable approach for enhancing EIT image quality and presents a novel application of LSTM networks in EIT technology.

physics.bio-ph↗

arXiv2025

AI for CSI Prediction in 5G-Advanced and Beyond

Artificial intelligence (AI) is pivotal in advancing fifth-generation (5G)-Advanced and sixth-generation systems, capturing substantial research interest. Both the 3rd Generation Partnership Project (3GPP) and leading corporations champion AI's standardization in wireless communication. This piece delves into AI's role in channel state information (CSI) prediction, a sub-use case acknowledged in 5G-Advanced by the 3GPP. We offer an exhaustive survey of AI-driven CSI prediction, highlighting crucial elements like accuracy, generalization, and complexity. Further, we touch on the practical side of model management, encompassing training, monitoring, and data gathering. Moreover, we explore prospects for CSI prediction in future wireless communication systems, entailing integrated design with feedback, multitasking synergy, and predictions in rapid scenarios. This article seeks to be a touchstone for subsequent research in this burgeoning domain.

eess.SP↗

arXiv2025

On the Feasibility of Using MultiModal LLMs to Execute AR Social Engineering Attacks

Augmented Reality (AR) and Multimodal Large Language Models (LLMs) are rapidly evolving, providing unprecedented capabilities for human-computer interaction. However, their integration introduces a new attack surface for social engineering. In this paper, we systematically investigate the feasibility of orchestrating AR-driven Social Engineering attacks using Multimodal LLM for the first time, via our proposed SEAR framework, which operates through three key phases: (1) AR-based social context synthesis, which fuses Multimodal inputs (visual, auditory and environmental cues); (2) role-based Multimodal RAG (Retrieval-Augmented Generation), which dynamically retrieves and integrates contextual data while preserving character differentiation; and (3) ReInteract social engineering agents, which execute adaptive multiphase attack strategies through inference interaction loops. To verify SEAR, we conducted an IRB-approved study with 60 participants in three experimental configurations (unassisted, AR+LLM, and full SEAR pipeline) compiling a new dataset of 180 annotated conversations in simulated social scenarios. Our results show that SEAR is highly effective at eliciting high-risk behaviors (e.g., 93.3% of participants susceptible to email phishing). The framework was particularly effective in building trust, with 85% of targets willing to accept an attacker's call after an interaction. Also, we identified notable limitations such as ``occasionally artificial'' due to perceived authenticity gaps. This work provides proof-of-concept for AR-LLM driven social engineering attacks and insights for developing defensive countermeasures against next-generation augmented reality threats.

cs.CR↗

arXiv2025

Imagining the Far East: Exploring Perceived Biases in AI-Generated Images of East Asian Women

Image-generating AI, which allows users to create images from text, is increasingly used to produce visual content. Despite its advancements, cultural biases in AI-generated images have raised significant concerns. While much research has focused on issues within Western contexts, our study examines the perceived biases regarding the portrayal of East Asian women. In this exploratory study, we invited East Asian users to audit three popular models (DALL-E, Midjourney, Stable Diffusion) and identified 18 specific perceived biases, categorized into four patterns: Westernization, overuse or misuse of cultural symbols, sexualization & feminization, and racial stereotypes. This work highlights the potential challenges posed by AI models in portraying Eastern individuals.

cs.HC↗

arXiv2025

Enabling Auditory Large Language Models for Automatic Speech Quality Evaluation

Speech quality assessment typically requires evaluating audio from multiple aspects, such as mean opinion score (MOS) and speaker similarity (SIM) \etc., which can be challenging to cover using one small model designed for a single task. In this paper, we propose leveraging recently introduced auditory large language models (LLMs) for automatic speech quality assessment. By employing task-specific prompts, auditory LLMs are finetuned to predict MOS, SIM and A/B testing results, which are commonly used for evaluating text-to-speech systems. Additionally, the finetuned auditory LLM is able to generate natural language descriptions assessing aspects like noisiness, distortion, discontinuity, and overall quality, providing more interpretable outputs. Extensive experiments have been performed on the NISQA, BVCC, SOMOS and VoxSim speech quality datasets, using open-source auditory LLMs such as SALMONN, Qwen-Audio, and Qwen2-Audio. For the natural language descriptions task, a commercial model Google Gemini 1.5 Pro is also evaluated. The results demonstrate that auditory LLMs achieve competitive performance compared to state-of-the-art task-specific small models in predicting MOS and SIM, while also delivering promising results in A/B testing and natural language descriptions. Our data processing scripts and finetuned model checkpoints can be found at https://github.com/bytedance/SALMONN.

eess.AS↗

arXiv2025

Principal Component Stochastic Subspace Identification for Output-Only Modal Analysis

Stochastic Subspace Identification (SSI) is widely used in modal analysis of engineering structures, known for its numerical stability and high accuracy in modal parameter identification. SSI methods are generally classified into two types: Data-Driven (SSI-Data) and Covariance-Driven (SSI-Cov), which have been considered to originate from different theoretical foundations and computational principles. In contrast, this study demonstrates that SSI-Cov and SSI-Data converge to the same solution under the condition of infinite observations, by establishing a unified framework incorporating instrumental variable analysis. Further, a novel modal identification approach, Principal Component Stochastic Subspace Identification (PCSSI), is proposed based on this framework. This method employs Principal Component Analysis (PCA) to extract key components of the signal subspace and project the observed data onto this space, enhancing modal identification stability while significantly reducing computational complexity. Through 5000 Monte Carlo numerical simulations, the statistical analysis shows that PCSSI consistently outperforms traditional SSI methods in terms of numerical stability and noise reduction, demonstrating clear advantages over both SSI-Cov and SSI-Data. Its effectiveness is further validated using experimental data from a scaled bridge model. Compared to conventional SSI approaches, PCSSI demonstrates superior robustness under complex engineering conditions, especially when dealing with limited data or high noise levels, underscoring its strong potential for practical applications.

physics.comp-ph↗

arXiv2025

GI-GS: Global Illumination Decomposition on Gaussian Splatting for Inverse Rendering

We present GI-GS, a novel inverse rendering framework that leverages 3D Gaussian Splatting (3DGS) and deferred shading to achieve photo-realistic novel view synthesis and relighting. In inverse rendering, accurately modeling the shading processes of objects is essential for achieving high-fidelity results. Therefore, it is critical to incorporate global illumination to account for indirect lighting that reaches an object after multiple bounces across the scene. Previous 3DGS-based methods have attempted to model indirect lighting by characterizing indirect illumination as learnable lighting volumes or additional attributes of each Gaussian, while using baked occlusion to represent shadow effects. These methods, however, fail to accurately model the complex physical interactions between light and objects, making it impossible to construct realistic indirect illumination during relighting. To address this limitation, we propose to calculate indirect lighting using efficient path tracing with deferred shading. In our framework, we first render a G-buffer to capture the detailed geometry and material properties of the scene. Then, we perform physically-based rendering (PBR) only for direct lighting. With the G-buffer and previous rendering results, the indirect lighting can be calculated through a lightweight path tracing. Our method effectively models indirect lighting under any given lighting conditions, thereby achieving better novel view synthesis and competitive relighting. Quantitative and qualitative results show that our GI-GS outperforms existing baselines in both rendering quality and efficiency.

cs.CV↗

arXiv2025

Dynamics-Aware Gaussian Splatting Streaming Towards Fast On-the-Fly 4D Reconstruction

The recent development of 3D Gaussian Splatting (3DGS) has led to great interest in 4D dynamic spatial reconstruction. Existing approaches mainly rely on full-length multi-view videos, while there has been limited exploration of online reconstruction methods that enable on-the-fly training and per-timestep streaming. Current 3DGS-based streaming methods treat the Gaussian primitives uniformly and constantly renew the densified Gaussians, thereby overlooking the difference between dynamic and static features as well as neglecting the temporal continuity in the scene. To address these limitations, we propose a novel three-stage pipeline for iterative streamable 4D dynamic spatial reconstruction. Our pipeline comprises a selective inheritance stage to preserve temporal continuity, a dynamics-aware shift stage to distinguish dynamic and static primitives and optimize their movements, and an error-guided densification stage to accommodate emerging objects. Our method achieves state-of-the-art performance in online 4D reconstruction, demonstrating the fastest on-the-fly training, superior representation quality, and real-time rendering capability. Project page: https://www.liuzhening.top/DASS

cs.CV↗

arXiv2025

ICG-MVSNet: Learning Intra-view and Cross-view Relationships for Guidance in Multi-View Stereo

Multi-view Stereo (MVS) aims to estimate depth and reconstruct 3D point clouds from a series of overlapping images. Recent learning-based MVS frameworks overlook the geometric information embedded in features and correlations, leading to weak cost matching. In this paper, we propose ICG-MVSNet, which explicitly integrates intra-view and cross-view relationships for depth estimation. Specifically, we develop an intra-view feature fusion module that leverages the feature coordinate correlations within a single image to enhance robust cost matching. Additionally, we introduce a lightweight cross-view aggregation module that efficiently utilizes the contextual information from volume correlations to guide regularization. Our method is evaluated on the DTU dataset and Tanks and Temples benchmark, consistently achieving competitive performance against state-of-the-art works, while requiring lower computational resources.

cs.CV↗

arXiv2025

Spectrally-large scale geometry via set-heaviness

We show that there exist infinite-dimensional quasi-flats in the compactly supported Hamiltonian diffeomorphism group of the Liouville domain, with respect to the spectral norm, if and only if the symplectic cohomology of this Liouville domain does not vanish. In particular, there exist infinite-dimensional quasi-flats in the compactly supported Hamiltonian diffeomorphism group of the unit co-disk bundle of any closed manifold. A similar conclusion holds for the ${\rm Ham}$-orbit space of an admissible Lagrangian in any Liouville domain. Moreover, we show that if a closed symplectic manifold contains an incompressible Lagrangian with a certain topological condition, then its Hamiltonian diffeomorphism group admits infinite-dimensional flats. Proofs of all these results rely on the existence of a family of heavy hypersurfaces.

math.SG↗

arXiv2025

Mutual Information-Empowered Task-Oriented Communication: Principles, Applications and Challenges

Mutual information (MI)-based guidelines have recently proven to be effective for designing task-oriented communication systems, where the ultimate goal is to extract and transmit task-relevant information for downstream task. This paper provides a comprehensive overview of MI-empowered task-oriented communication, highlighting how MI-based methods can serve as a unifying design framework in various task-oriented communication scenarios. We begin with the roadmap of MI for designing task-oriented communication systems, and then introduce the roles and applications of MI to guide feature encoding, transmission optimization, and efficient training with two case studies. We further elaborate the limitations and challenges of MI-based methods. Finally, we identify several open issues in MI-based task-oriented communication to inspire future research.

cs.IT↗

arXiv2025

Neural Representation for Wireless Radiation Field Reconstruction: A 3D Gaussian Splatting Approach

Wireless channel modeling plays a pivotal role in designing, analyzing, and optimizing wireless communication systems. Nevertheless, developing an effective channel modeling approach has been a long-standing challenge. This issue has been escalated due to denser network deployment, larger antenna arrays, and broader bandwidth in next-generation networks. To address this challenge, we put forth WRF-GS, a novel framework for channel modeling based on wireless radiation field (WRF) reconstruction using 3D Gaussian splatting (3D-GS). WRF-GS employs 3D Gaussian primitives and neural networks to capture the interactions between the environment and radio signals, enabling efficient WRF reconstruction and visualization of the propagation characteristics. The reconstructed WRF can then be used to synthesize the spatial spectrum for comprehensive wireless channel characterization. While WRF-GS demonstrates remarkable effectiveness, it faces limitations in capturing high-frequency signal variations caused by complex multipath effects. To overcome these limitations, we propose WRF-GS+, an enhanced framework that integrates electromagnetic wave physics into the neural network design. WRF-GS+ leverages deformable 3D Gaussians to model both static and dynamic components of the WRF, significantly improving its ability to characterize signal variations. In addition, WRF-GS+ enhances the splatting process by simplifying the 3D-GS modeling process and improving computational efficiency. Experimental results demonstrate that both WRF-GS and WRF-GS+ outperform baselines for spatial spectrum synthesis, including ray tracing and other deep-learning approaches. Notably, WRF-GS+ achieves state-of-the-art performance in the received signal strength indication (RSSI) and channel state information (CSI) prediction tasks, surpassing existing methods by more than 0.7 dB and 3.36 dB, respectively.

cs.NI↗

arXiv2025

Contact quasi-states and their applications

We introduce the notions of partial contact quasi-state and contact quasi-measure. Using the contact spectral invariant from the work by Djordjević-Uljarević-Zhang, one can construct partial contact quasi-states and contact quasi-measures on each contact manifold fillable by a Liouville domain with non-vanishing and $\mathbb{Z}$-graded symplectic homology. As an application, we present an alternative proof of the contact big fibre theorem from the recent work by Sun-Uljarević-Varolgunes under a mild topological condition. Our proof follows a more "classical" approach developed by Entov-Polterovich in the symplectic setting.

math.SG↗

arXiv2025

Notes on symplectic squeezing in $T^* \mathbb T^n$ and spectra of Finsler dynamics

In this paper, on the one hand, we prove that for $n \geq 2$ any subbundle of $T^* \mathbb T^n$ with bounded fibers symplectically embeds into a trivial subbundle of $T^* \mathbb T^n$ where the fiber is an irrational cylinder. This not only resolves an open problem in Gong-Xue's recent work (which was stated for the 4-dimension case, that is, $n =2$) and also generalizes to any higher-dimensional situation. The proof is based on some version of Dirichlet's approximation theorem. On the other hand, we generalize a main result in Gong-Xue's work mentioned above, showing that any topologically trivial Liouville diffeomorphism on $T^*M$ (for instance, a diffeomorphism induced by an isometry on $M$) does not change the full marked length spectrum of a Finsler metric $F$ on $M$, up to a lifting of the Finsler metric $F$ to the unit codisk bundle $D^*_FM$. The proof is based on persistence module theory.

math.SG↗

arXiv2025

Chain of Functions: A Programmatic Pipeline for Fine-Grained Chart Reasoning Data

Visual reasoning is crucial for multimodal large language models (MLLMs) to address complex chart queries, yet high-quality rationale data remains scarce. Existing methods leveraged (M)LLMs for data generation, but direct prompting often yields limited precision and diversity. In this paper, we propose \textit{Chain of Functions (CoF)}, a novel programmatic reasoning data generation pipeline that utilizes freely-explored reasoning paths as supervision to ensure data precision and diversity. Specifically, it starts with human-free exploration among the atomic functions (e.g., maximum data and arithmetic operations) to generate diverse function chains, which are then translated into linguistic rationales and questions with only a moderate open-sourced LLM. \textit{CoF} provides multiple benefits: 1) Precision: function-governed generation reduces hallucinations compared to freeform generation; 2) Diversity: enumerating function chains enables varied question taxonomies; 3) Explainability: function chains serve as built-in rationales, allowing fine-grained evaluation beyond overall accuracy; 4) Practicality: eliminating reliance on extremely large models. Employing \textit{CoF}, we construct the \textit{ChartCoF} dataset, with 1.4k complex reasoning Q\&A for fine-grained analysis and 50k Q\&A for reasoning enhancement. The fine-grained evaluation on \textit{ChartCoF} reveals varying performance across question taxonomies for each MLLM, and the experiments also show that finetuning with \textit{ChartCoF} achieves state-of-the-art performance among same-scale MLLMs on widely used benchmarks. Furthermore, the novel paradigm of function-governed rationale generation in \textit{CoF} could inspire broader applications beyond charts.

cs.CV↗