arXiv subjects

Zeyu Wang

Publications and source records attributed to Zeyu Wang.

At least 127 records · Page 7Linked to original sources

arXiv2025

SpecOffload: Unlocking Latent GPU Capacity for LLM Inference on Resource-Constrained Devices

Efficient LLM inference on resource-constrained devices presents significant challenges in compute and memory utilization. Due to limited GPU memory, existing systems offload model weights to CPU memory, incurring substantial I/O overhead between the CPU and GPU. This leads to two major inefficiencies: (1) GPU cores are underutilized, often remaining idle while waiting for data to be loaded; and (2) GPU memory has low impact on performance, as reducing its capacity has minimal effect on overall throughput.In this paper, we propose SpecOffload, a high-throughput inference engine that embeds speculative decoding into offloading. Our key idea is to unlock latent GPU resources for storing and executing a draft model used for speculative decoding, thus accelerating inference at near-zero additional cost. To support this, we carefully orchestrate the interleaved execution of target and draft models in speculative decoding within the offloading pipeline, and propose a planner to manage tensor placement and select optimal parameters. Compared to the best baseline, SpecOffload improves GPU core utilization by 4.49x and boosts inference throughput by 2.54x. Our code is available at https://github.com/MobiSense/SpecOffload-public .

cs.LG↗

arXiv2025

AdaServe: Accelerating Multi-SLO LLM Serving with SLO-Customized Speculative Decoding

Modern large language model (LLM) applications exhibit diverse service-level objectives (SLOs), from low-latency requirements in interactive coding assistants to more relaxed constraints in data wrangling tasks. Existing LLM serving systems, which rely on uniform batching and scheduling strategies, often fail to meet these heterogeneous SLOs concurrently. We present AdaServe, the first LLM serving system designed to support efficient multi-SLO serving through SLO-customized speculative decoding. AdaServe formulates multi-SLO serving as a constrained optimization problem and introduces a hardware-aware algorithm that constructs a speculation tree tailored to each request's latency target. It features a speculate-select-verify pipeline that enables fine-grained control over decoding speed while maximizing system throughput. AdaServe further adapts to workload variation by dynamically adjusting speculation parameters. Evaluations across diverse workloads show that AdaServe reduces SLO violations by up to 4.3$\times$ and improves goodput by up to 1.9$\times$ compared to the best performing baselines, highlighting its effectiveness in multi-SLO serving.

cs.CL↗

arXiv2025

AR Secretary Agent: Real-time Memory Augmentation via LLM-powered Augmented Reality Glasses

Interacting with a significant number of individuals on a daily basis is commonplace for many professionals, which can lead to challenges in recalling specific details: Who is this person? What did we talk about last time? The advant of augmented reality (AR) glasses, equipped with visual and auditory data capture capabilities, presents a solution. In our work, we implemented an AR Secretary Agent with advanced Large Language Models (LLMs) and Computer Vision technologies. This system could discreetly provide real-time information to the wearer, identifying who they are conversing with and summarizing previous discussions. To verify AR Secretary, we conducted a user study with 13 participants and showed that our technique can efficiently help users to memorize events by up to 20\% memory enhancement on our study.

cs.HC↗

arXiv2025

AI Hiring with LLMs: A Context-Aware and Explainable Multi-Agent Framework for Resume Screening

Resume screening is a critical yet time-intensive process in talent acquisition, requiring recruiters to analyze vast volume of job applications while remaining objective, accurate, and fair. With the advancements in Large Language Models (LLMs), their reasoning capabilities and extensive knowledge bases demonstrate new opportunities to streamline and automate recruitment workflows. In this work, we propose a multi-agent framework for resume screening using LLMs to systematically process and evaluate resumes. The framework consists of four core agents, including a resume extractor, an evaluator, a summarizer, and a score formatter. To enhance the contextual relevance of candidate assessments, we integrate Retrieval-Augmented Generation (RAG) within the resume evaluator, allowing incorporation of external knowledge sources, such as industry-specific expertise, professional certifications, university rankings, and company-specific hiring criteria. This dynamic adaptation enables personalized recruitment, bridging the gap between AI automation and talent acquisition. We assess the effectiveness of our approach by comparing AI-generated scores with ratings provided by HR professionals on a dataset of anonymized online resumes. The findings highlight the potential of multi-agent RAG-LLM systems in automating resume screening, enabling more efficient and scalable hiring workflows.

cs.CL↗

arXiv2025

CaMDN: Enhancing Cache Efficiency for Multi-tenant DNNs on Integrated NPUs

With the rapid development of DNN applications, multi-tenant execution, where multiple DNNs are co-located on a single SoC, is becoming a prevailing trend. Although many methods are proposed in prior works to improve multi-tenant performance, the impact of shared cache is not well studied. This paper proposes CaMDN, an architecture-scheduling co-design to enhance cache efficiency for multi-tenant DNNs on integrated NPUs. Specifically, a lightweight architecture is proposed to support model-exclusive, NPU-controlled regions inside shared cache to eliminate unexpected cache contention. Moreover, a cache scheduling method is proposed to improve shared cache utilization. In particular, it includes a cache-aware mapping method for adaptability to the varying available cache capacity and a dynamic allocation algorithm to adjust the usage among co-located DNNs at runtime. Compared to prior works, CaMDN reduces the memory access by 33.4% on average and achieves a model speedup of up to 2.56$\times$ (1.88$\times$ on average).

cs.AR↗

arXiv2025

A Dataset and Toolkit for Multiparameter Cardiovascular Physiology Sensing on Rings

Smart rings offer a convenient way to continuously and unobtrusively monitor cardiovascular physiological signals. However, a gap remains between the ring hardware and reliable methods for estimating cardiovascular parameters, partly due to the lack of publicly available datasets and standardized analysis tools. In this work, we present $τ$-Ring, the first open-source ring-based dataset designed for cardiovascular physiological sensing. The dataset comprises photoplethysmography signals (infrared and red channels) and 3-axis accelerometer data collected from two rings (reflective and transmissive optical paths), with 28.21 hours of raw data from 34 subjects across seven activities. $τ$-Ring encompasses both stationary and motion scenarios, as well as stimulus-evoked abnormal physiological states, annotated with four ground-truth labels: heart rate, respiratory rate, oxygen saturation, and blood pressure. Using our proposed RingTool toolkit, we evaluated three widely-used physics-based methods and four cutting-edge deep learning approaches. Our results show superior performance compared to commercial rings, achieving best MAE values of 5.18 BPM for heart rate, 2.98 BPM for respiratory rate, 3.22\% for oxygen saturation, and 13.33/7.56 mmHg for systolic/diastolic blood pressure estimation. The open-sourced dataset and toolkit aim to foster further research and community-driven advances in ring-based cardiovascular health sensing.

eess.IV↗

arXiv2025

Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion

Text-to-image generation via Stable Diffusion models (SDM) have demonstrated remarkable capabilities. However, their computational intensity, particularly in the iterative denoising process, hinders real-time deployment in latency-sensitive applications. While Recent studies have explored post-training quantization (PTQ) and quantization-aware training (QAT) methods to compress Diffusion models, existing methods often overlook the consistency between results generated by quantized models and those from floating-point models. This consistency is paramount for professional applications where both efficiency and output reliability are essential. To ensure that quantized SDM generates high-quality and consistent images, we propose an efficient quantization framework for SDM. Our framework introduces a Serial-to-Parallel pipeline that simultaneously maintains training-inference consistency and ensures optimization stability. Building upon this foundation, we further develop several techniques including multi-timestep activation quantization, time information precalculation, inter-layer distillation, and selective freezing, to achieve high-fidelity generation in comparison to floating-point models while maintaining quantization efficiency. Through comprehensive evaluation across multiple Stable Diffusion variants (v1-4, v2-1, XL 1.0, and v3), our method demonstrates superior performance over state-of-the-art approaches with shorter training times. Under W4A8 quantization settings, we achieve significant improvements in both distribution similarity and visual fidelity, while preserving a high image quality.

cs.CV↗

arXiv2025

HeadEvolver: Text to Head Avatars via Expressive and Attribute-Preserving Mesh Deformation

Current text-to-avatar methods often rely on implicit representations (e.g., NeRF, SDF, and DMTet), leading to 3D content that artists cannot easily edit and animate in graphics software. This paper introduces a novel framework for generating stylized head avatars from text guidance, which leverages locally learnable mesh deformation and 2D diffusion priors to achieve high-quality digital assets for attribute-preserving manipulation. Given a template mesh, our method represents mesh deformation with per-face Jacobians and adaptively modulates local deformation using a learnable vector field. This vector field enables anisotropic scaling while preserving the rotation of vertices, which can better express identity and geometric details. We employ landmark- and contour-based regularization terms to balance the expressiveness and plausibility of generated avatars from multiple views without relying on any specific shape prior. Our framework can generate realistic shapes and textures that can be further edited via text, while supporting seamless editing using the preserved attributes from the template mesh, such as 3DMM parameters, blendshapes, and UV coordinates. Extensive experiments demonstrate that our framework can generate diverse and expressive head avatars with high-quality meshes that artists can easily manipulate in graphics software, facilitating downstream applications such as efficient asset creation and animation with preserved attributes.

cs.GR↗

arXiv2025

On the relative Langlands duality for $\operatorname{Sp}_{2n} \backslash \operatorname{GL}_{2n+1}$ (with an appendix by Zeyu Wang)

We verify the relative Langlands duality conjecture proposed by Ben-Zvi, Sakellaridis, Venkatesh for the hyperspherical Hamiltonian variety $T^*(\operatorname{Sp}_{2n}\backslash \operatorname{GL}_{2n+1})$. We provide numerical (over number fields and function fields) and geometric (in the étale setting) evidence that its dual Hamiltonian variety should be $T^*(\operatorname{GL}_n \times \operatorname{GL}_{n+1} \backslash \operatorname{GL}_{2n+1})$ as is predicted by Ben-Zvi, Sakellaridis, Venkatesh.

math.RT↗

arXiv2025

An LLM-enabled Multi-Agent Autonomous Mechatronics Design Framework

Existing LLM-enabled multi-agent frameworks are predominantly limited to digital or simulated environments and confined to narrowly focused knowledge domain, constraining their applicability to complex engineering tasks that require the design of physical embodiment, cross-disciplinary integration, and constraint-aware reasoning. This work proposes a multi-agent autonomous mechatronics design framework, integrating expertise across mechanical design, optimization, electronics, and software engineering to autonomously generate functional prototypes with minimal direct human design input. Operating primarily through a language-driven workflow, the framework incorporates structured human feedback to ensure robust performance under real-world constraints. To validate its capabilities, the framework is applied to a real-world challenge involving autonomous water-quality monitoring and sampling, where traditional methods are labor-intensive and ecologically disruptive. Leveraging the proposed system, a fully functional autonomous vessel was developed with optimized propulsion, cost-effective electronics, and advanced control. The design process was carried out by specialized agents, including a high-level planning agent responsible for problem abstraction and dedicated agents for structural, electronics, control, and software development. This approach demonstrates the potential of LLM-based multi-agent systems to automate real-world engineering workflows and reduce reliance on extensive domain expertise.

cs.RO↗

arXiv2025

Adaptive Substructure-Aware Expert Model for Molecular Property Prediction

Molecular property prediction is essential for applications such as drug discovery and toxicity assessment. While Graph Neural Networks (GNNs) have shown promising results by modeling molecules as molecular graphs, their reliance on data-driven learning limits their ability to generalize, particularly in the presence of data imbalance and diverse molecular substructures. Existing methods often overlook the varying contributions of different substructures to molecular properties, treating them uniformly. To address these challenges, we propose ASE-Mol, a novel GNN-based framework that leverages a Mixture-of-Experts (MoE) approach for molecular property prediction. ASE-Mol incorporates BRICS decomposition and significant substructure awareness to dynamically identify positive and negative substructures. By integrating a MoE architecture, it reduces the adverse impact of negative motifs while improving adaptability to positive motifs. Experimental results on eight benchmark datasets demonstrate that ASE-Mol achieves state-of-the-art performance, with significant improvements in both accuracy and interpretability.

cs.LG↗

arXiv2025

Double Visual Defense: Adversarial Pre-training and Instruction Tuning for Improving Vision-Language Model Robustness

This paper investigates the robustness of vision-language models against adversarial visual perturbations and introduces a novel ``double visual defense" to enhance this robustness. Unlike previous approaches that resort to lightweight adversarial fine-tuning of a pre-trained CLIP model, we perform large-scale adversarial vision-language pre-training from scratch using web-scale data. We then strengthen the defense by incorporating adversarial visual instruction tuning. The resulting models from each stage, $Δ$CLIP and $Δ^2$LLaVA, show substantially enhanced zero-shot robustness and set a new state-of-the-art in adversarial defense for vision-language models. For example, the adversarial robustness of $Δ$CLIP surpasses that of the previous best models on ImageNet-1k by ~20%. %For example, $Δ$CLIP surpasses the previous best models on ImageNet-1k by ~20% in terms of adversarial robustness. Similarly, compared to prior art, $Δ^2$LLaVA brings a ~30% robustness improvement to image captioning task and a ~20% robustness improvement to visual question answering task. Furthermore, our models exhibit stronger zero-shot recognition capability, fewer hallucinations, and superior reasoning performance compared to baselines. Our project page is https://doublevisualdefense.github.io/.

cs.CV↗

arXiv2025

Attention in Diffusion Model: A Survey

Attention mechanisms have become a foundational component in diffusion models, significantly influencing their capacity across a wide range of generative and discriminative tasks. This paper presents a comprehensive survey of attention within diffusion models, systematically analysing its roles, design patterns, and operations across different modalities and tasks. We propose a unified taxonomy that categorises attention-related modifications into parts according to the structural components they affect, offering a clear lens through which to understand their functional diversity. In addition to reviewing architectural innovations, we examine how attention mechanisms contribute to performance improvements in diverse applications. We also identify current limitations and underexplored areas, and outline potential directions for future research. Our study provides valuable insights into the evolving landscape of diffusion models, with a particular focus on the integrative and ubiquitous role of attention.

cs.LG↗

arXiv2025

Murmurations of Hecke $L$-Functions of Imaginary Quadratic Fields

We calculate the murmuration density for the family of Hecke $L$-functions of imaginary quadratic fields associated to non-trivial characters. This density exhibits a universality property like Zubrilina's density for the murmurations of holomorphic modular forms. We show all murmuration functions obtained by averaging over the family with a compactly supported smooth weight function has asymptotics compatible with the 1-level density conjecture of Katz and Sarnak. The novelty of the murmurations of this family of $L$-functions is its pronounced almost periodic feature, which allows one to describe this murmuration without averaging over primes, and which is non-existent or previously unnoticed for other families.

math.NT↗

arXiv2025

Dimensionality-Reduction Techniques for Approximate Nearest Neighbor Search: A Survey and Evaluation

Approximate Nearest Neighbor Search (ANNS) on high-dimensional vectors has become a fundamental and essential component in various machine learning tasks. Recently, with the rapid development of deep learning models and the applications of Large Language Models (LLMs), the dimensionality of the vectors keeps growing in order to accommodate a richer semantic representation. This poses a major challenge to the ANNS solutions since distance calculation cost in ANNS grows linearly with the dimensionality of vectors. To overcome this challenge, dimensionality-reduction techniques can be leveraged to accelerate the distance calculation in the search process. In this paper, we investigate six dimensionality-reduction techniques that have the potential to improve ANNS solutions, including classical algorithms such as PCA and vector quantization, as well as algorithms based on deep learning approaches. We further describe two frameworks to apply these techniques in the ANNS workflow, and theoretically analyze the time and space costs, as well as the beneficial threshold for the pruning ratio of these techniques. The surveyed techniques are evaluated on six public datasets. The analysis of the results reveals the characteristics of the different families of techniques and provides insights into the promising future research directions.

cs.DB↗

arXiv2025

Spiking Point Transformer for Point Cloud Classification

Spiking Neural Networks (SNNs) offer an attractive and energy-efficient alternative to conventional Artificial Neural Networks (ANNs) due to their sparse binary activation. When SNN meets Transformer, it shows great potential in 2D image processing. However, their application for 3D point cloud remains underexplored. To this end, we present Spiking Point Transformer (SPT), the first transformer-based SNN framework for point cloud classification. Specifically, we first design Queue-Driven Sampling Direct Encoding for point cloud to reduce computational costs while retaining the most effective support points at each time step. We introduce the Hybrid Dynamics Integrate-and-Fire Neuron (HD-IF), designed to simulate selective neuron activation and reduce over-reliance on specific artificial neurons. SPT attains state-of-the-art results on three benchmark datasets that span both real-world and synthetic datasets in the SNN domain. Meanwhile, the theoretical energy consumption of SPT is at least 6.4$\times$ less than its ANN counterpart.

cs.LG↗

arXiv2025

DEGAS: Detailed Expressions on Full-Body Gaussian Avatars

Although neural rendering has made significant advances in creating lifelike, animatable full-body and head avatars, incorporating detailed expressions into full-body avatars remains largely unexplored. We present DEGAS, the first 3D Gaussian Splatting (3DGS)-based modeling method for full-body avatars with rich facial expressions. Trained on multiview videos of a given subject, our method learns a conditional variational autoencoder that takes both the body motion and facial expression as driving signals to generate Gaussian maps in the UV layout. To drive the facial expressions, instead of the commonly used 3D Morphable Models (3DMMs) in 3D head avatars, we propose to adopt the expression latent space trained solely on 2D portrait images, bridging the gap between 2D talking faces and 3D avatars. Leveraging the rendering capability of 3DGS and the rich expressiveness of the expression latent space, the learned avatars can be reenacted to reproduce photorealistic rendering images with subtle and accurate facial expressions. Experiments on an existing dataset and our newly proposed dataset of full-body talking avatars demonstrate the efficacy of our method. We also propose an audio-driven extension of our method with the help of 2D talking faces, opening new possibilities for interactive AI agents.

cs.CV↗

arXiv2025

Computing with Smart Rings: A Systematic Literature Review

A smart ring is a wearable electronic device in the form of a ring that incorporates diverse sensors and computing technologies to perform a variety of functions. Designed for use with fingers, smart rings are capable of sensing more subtle and abundant hand movements, thus making them a good platform for interaction. Meanwhile, fingers are abundant with blood vessels and nerve endings and accustomed to wearing rings, providing an ideal site for continuous health monitoring through smart rings, which combine comfort with the ability to capture vital biometric data, making them suitable for all-day wear. We collected in total of 206 smart ring-related publications and conducted a systematic literature review. We provide a taxonomy regarding the sensing and feedback modalities, applications, and phenomena. We review and categorize these literatures into four main areas: (1) interaction - input, (2) interaction - output, (3) passive sensing - in body feature, (4) passive sensing - out body activity. This comprehensive review highlights the current advancements within the field of smart ring and identifies potential areas for future research.

cs.HC↗