arXiv subjects

Cheng Zhang

Publications and source records attributed to Cheng Zhang.

At least 163 records · Page 9Linked to original sources

arXiv2025

Beyond-Voice: Towards Continuous 3D Hand Pose Tracking on Commercial Home Assistant Devices

The surging popularity of home assistants and their voice user interface (VUI) have made them an ideal central control hub for smart home devices. However, current form factors heavily rely on VUI, which poses accessibility and usability issues; some latest ones are equipped with additional cameras and displays, which are costly and raise privacy concerns. These concerns jointly motivate Beyond-Voice, a novel high-fidelity acoustic sensing system that allows commodity home assistant devices to track and reconstruct hand poses continuously. It transforms the home assistant into an active sonar system using its existing onboard microphones and speakers. We feed a high-resolution range profile to the deep learning model that can analyze the motions of multiple body parts and predict the 3D positions of 21 finger joints, bringing the granularity for acoustic hand tracking to the next level. It operates across different environments and users without the need for personalized training data. A user study with 11 participants in 3 different environments shows that Beyond-Voice can track joints with an average mean absolute error of 16.47mm without any training data provided by the testing subject.

cs.SD↗

arXiv2025

InfraDiffusion: zero-shot depth map restoration with diffusion models and prompted segmentation from sparse infrastructure point clouds

Point clouds are widely used for infrastructure monitoring by providing geometric information, where segmentation is required for downstream tasks such as defect detection. Existing research has automated semantic segmentation of structural components, while brick-level segmentation (identifying defects such as spalling and mortar loss) has been primarily conducted from RGB images. However, acquiring high-resolution images is impractical in low-light environments like masonry tunnels. Point clouds, though robust to dim lighting, are typically unstructured, sparse, and noisy, limiting fine-grained segmentation. We present InfraDiffusion, a zero-shot framework that projects masonry point clouds into depth maps using virtual cameras and restores them by adapting the Denoising Diffusion Null-space Model (DDNM). Without task-specific training, InfraDiffusion enhances visual clarity and geometric consistency of depth maps. Experiments on masonry bridge and tunnel point cloud datasets show significant improvements in brick-level segmentation using the Segment Anything Model (SAM), underscoring its potential for automated inspection of masonry assets. Our code and data is available at https://github.com/Jingyixiong/InfraDiffusion-official-implement.

cs.CV↗

arXiv2025

On the Dbar method and direct linearization approach of the lattice KdV type equations

The purpose of this paper is to bridge the gap between the Dbar method and the direct linearization approach for the lattice Korteweg-de Vries (KdV) type equations. We develop the Dbar method to study some discrete integrable equations in the Adler-Bobenko-Suris list. A Dbar problem is considered to define the eigenfunctions of the Lax pair of the lattice potential KdV equation. We show how an extra parameter is introduced in this approach so that the lattice potential modified KdV equation and lattice Schwarzian KdV equation are derived. We also explain how the so-called spectral Wronskians make sense in constructing the H3$(δ)$, Q1$(δ)$ and Q3$(δ)$ equations. Explicit formulae of multi-soliton solutions are given for the derived equations, from which one can see the connections between the direct linearization variables ($S^{(i,j)}$ and $V(p)$) and the eigenfunctions and their expansions respectively at infinity and a finite point.

nlin.SI↗

arXiv2025

Strichartz estimates for orthonormal systems on compact manifolds

We establish new Strichartz estimates for orthonormal systems on compact Riemannian manifolds in the case of wave, Klein-Gordon and fractional Schrödinger equations. Our results generalize the classical (single-function) Strichartz estimates on compact manifolds by Kapitanski, Burq-Gérard-Tzvetkov, Dinh, and extend the Euclidean orthonormal version by Frank-Lewin-Lieb-Seiringer, Frank-Sabin, Bez-Lee-Nakamura. On the flat torus, our new results for the Schrödinger equation cover prior work of Nakamura, which exploits the dispersive estimate of Kenig-Ponce-Vega. We achieve sharp results on compact manifolds by combining the frequency localized dispersive estimates for small time intervals with the duality principle due to Frank-Sabin. We construct examples to show these results can be saturated on the sphere, and we can improve them on the flat torus by using Bourgain-Demeter's decoupling theorem to obtain new decoupling inequalities for certain non-smooth hypersurfaces. As an application, we obtain the well-posedness of infinite systems of dispersive equations with Hartree-type nonlinearity.

math.AP↗

arXiv2025

MobiAgent: A Systematic Framework for Customizable Mobile Agents

With the rapid advancement of Vision-Language Models (VLMs), GUI-based mobile agents have emerged as a key development direction for intelligent mobile systems. However, existing agent models continue to face significant challenges in real-world task execution, particularly in terms of accuracy and efficiency. To address these limitations, we propose MobiAgent, a comprehensive mobile agent system comprising three core components: the MobiMind-series agent models, the AgentRR acceleration framework, and the MobiFlow benchmarking suite. Furthermore, recognizing that the capabilities of current mobile agents are still limited by the availability of high-quality data, we have developed an AI-assisted agile data collection pipeline that significantly reduces the cost of manual annotation. Compared to both general-purpose LLMs and specialized GUI agent models, MobiAgent achieves state-of-the-art performance in real-world mobile scenarios.

cs.MA↗

arXiv2025

HRS: Hybrid Representation Framework with Scheduling Awareness for Time Series Forecasting in Crowdsourced Cloud-Edge Platforms

With the rapid proliferation of streaming services, network load exhibits highly time-varying and bursty behavior, posing serious challenges for maintaining Quality of Service (QoS) in Crowdsourced Cloud-Edge Platforms (CCPs). While CCPs leverage Predict-then-Schedule architecture to improve QoS and profitability, accurate load forecasting remains challenging under traffic surges. Existing methods either minimize mean absolute error, resulting in underprovisioning and potential Service Level Agreement (SLA) violations during peak periods, or adopt conservative overprovisioning strategies, which mitigate SLA risks at the expense of increased resource expenditure. To address this dilemma, we propose HRS, a hybrid representation framework with scheduling awareness that integrates numerical and image-based representations to better capture extreme load dynamics. We further introduce a Scheduling-Aware Loss (SAL) that captures the asymmetric impact of prediction errors, guiding predictions that better support scheduling decisions. Extensive experiments on four real-world datasets demonstrate that HRS consistently outperforms ten baselines and achieves state-of-the-art performance, reducing SLA violation rates by 63.1% and total profit loss by 32.3%.

cs.LG↗

arXiv2025

Grab-n-Go: On-the-Go Microgesture Recognition with Objects in Hand

As computing devices become increasingly integrated into daily life, there is a growing need for intuitive, always-available interaction methods, even when users' hands are occupied. In this paper, we introduce Grab-n-Go, the first wearable device that leverages active acoustic sensing to recognize subtle hand microgestures while holding various objects. Unlike prior systems that focus solely on free-hand gestures or basic hand-object activity recognition, Grab-n-Go simultaneously captures information about hand microgestures, grasping poses, and object geometries using a single wristband, enabling the recognition of fine-grained hand movements occurring within activities involving occupied hands. A deep learning framework processes these complex signals to identify 30 distinct microgestures, with 6 microgestures for each of the 5 grasping poses. In a user study with 10 participants and 25 everyday objects, Grab-n-Go achieved an average recognition accuracy of 92.0%. A follow-up study further validated Grab-n-Go's robustness against 10 more challenging, deformable objects. These results underscore the potential of Grab-n-Go to provide seamless, unobtrusive interactions without requiring modifications to existing objects. The complete dataset, comprising data from 18 participants performing 30 microgestures with 35 distinct objects, is publicly available at https://github.com/cjlisalee/Grab-n-Go_Data with the DOI: https://doi.org/10.7298/7kbd-vv75.

cs.HC↗

arXiv2025

Scaling Laws For Mixed Quantization

Post-training quantization of Large Language Models (LLMs) has proven effective in reducing the memory and computational requirements for inference. In this study, we focus on a straightforward question: When aiming for a target accuracy or perplexity with low-precision quantization, how much high-precision computation needs to be preserved, and how fine-grained this quantization would need to be as we scale LLMs to larger sizes? We first introduce two critical metrics, named the quantization ratio ($Q_r$) and quantization block size ($Q_b$). The former measures the number of parameters quantized to low-precision arithmetic normalized by the total parameter count, whereas the latter defines the number of values within a block that share a scaling factor, akin to the block size concept introduced in the FP4 format in NVIDIA's Blackwell architecture. Through extensive and carefully controlled experiments across different models and quantization methods, we propose a unified scaling law on post-training quantization (PTQ) that can predict loss degeneration for varying $Q_r$ and $Q_b$. For $Q_r$, our scaling law implies that parameter scaling and ratio scaling have a multiplicative relationship. Consequently, larger models are more amenable to a higher quantization ratio $Q_r$, thus supporting an increase in the adoption of mixed quantization for inference. Regarding $Q_b$, our findings indicate that a small block size, similar to that used in Blackwell, is not essential for large models. Employing a small $Q_b$ can instead unnecessarily complicate the design of the hardware circuit.

cs.CL↗

arXiv2025

A simple-to-implement nonlinear preconditioning of Newton's method for solving the steady Navier-Stokes equations

The Newton's method for solving stationary Navier-Stokes equations (NSE) is known to convergent fast, however, may fail due to a bad initial guess. This work presents a simple-to-implement nonlinear preconditioning of Newton's iteration, that remains the quadratic convergence and enlarges the domain of convergence. The proposed AAPicard-Newton method adds the Anderson accelerated Picard step at each iteration of Newton's method for solving NSE, which has been shown globally stable for the relaxation parameter $β_{k+1}\equiv1$ in the Anderson acceleration optimization step, convergent quadratically, and converges faster with a smaller convergence rate for large Reynolds number. Several benchmark numerical tests have been tested and are well-aligned with the theoretical results.

math.NA↗

arXiv2025

EchoForce: Continuous Grip Force Estimation from Skin Deformation Using Active Acoustic Sensing on a Wristband

Grip force is commonly used as an overall health indicator in older adults and is valuable for tracking progress in physical training and rehabilitation. Existing methods for wearable grip force measurement are cumbersome and user-dependent, making them insufficient for practical, continuous grip force measurement. We introduce EchoForce, a novel wristband using acoustic sensing for low-cost, non-contact measurement of grip force. EchoForce captures acoustic signals reflected from subtle skin deformations by flexor muscles on the forearm. In a user study with 11 participants, EchoForce achieved a fine-tuned user-dependent mean error rate of 9.08% and a user-independent mean error rate of 12.3% using a foundation model. Our system remained accurate between sessions, hand orientations, and users, overcoming a significant limitation of past force sensing systems. EchoForce makes continuous grip force measurement practical, providing an effective tool for health monitoring and novel interaction techniques.

cs.HC↗

arXiv2025

ARTreeFormer: A Faster Attention-based Autoregressive Model for Phylogenetic Inference

Probabilistic modeling over the combinatorially large space of tree topologies remains a central challenge in phylogenetic inference. Previous approaches often necessitate pre-sampled tree topologies, limiting their modeling capability to a subset of the entire tree space. A recent advancement is ARTree, a deep autoregressive model that offers unrestricted distributions for tree topologies. However, its reliance on repetitive tree traversals and inefficient local message passing for computing topological node representations may hamper the scalability to large datasets. This paper proposes ARTreeFormer, a novel approach that harnesses fixed-point iteration and attention mechanisms to accelerate ARTree. By introducing a fixed-point iteration algorithm for computing the topological node embeddings, ARTreeFormer allows fast vectorized computation, especially on CUDA devices. This, together with an attention-based global message passing scheme, significantly improves the computation speed of ARTree while maintaining great approximation performance. We demonstrate the effectiveness and efficiency of our method on a benchmark of challenging real data phylogenetic inference problems.

q-bio.PE↗

arXiv2025

How to Fix Silver for Plasmonics

Silver (Ag) is considered an ideal material for plasmonic applications in the visible wavelength regime due to its superior optical properties, but its use is limited by the poor chemical stability and structural quality of thermally evaporated thin films and resulting nanostructures. In this study, we present a simple approach to enhance the structural and optical quality as well as the chemical stability of Ag thin films by alloying with gold (Au) through thermal co-evaporation. We investigate Ag$_{100-x}$Au$_x$ thin films with Au contents ranging from 5 to 20 at% analyzing their surface morphology, crystallite structure, optical properties, and chemical stability. Our results show that low Au concentrations significantly reduce the roughness of co-evaporated thin films (down to 0.4 nm RMS), and significantly enhance the resistance to oxidation, while maintaining a defined crystallite growth. Importantly, these improvements are achieved without the need for template stripping, metallic wetting layers, or epitaxial substrates, enabling direct deposition on glass. Among the compositions studied, Ag$_{95}$Au$_5$ thin films exhibit the highest chemical stability, lowest optical losses in the visible spectral range, and excellent plasmonic properties even outcompeting pure Ag. As a proof-of-concept, we fabricate high-quality Ag$_{95}$Au$_5$ optical antennas that exhibit long-term durability under ambient conditions. Our approach provides a practical solution to overcome the limitations of Ag for plasmonic device applications.

cond-mat.mtrl-sci↗

arXiv2025

Artificial intelligence in drug discovery: A comprehensive review with a case study on hyperuricemia, gout arthritis, and hyperuricemic nephropathy

This paper systematically reviews recent advances in artificial intelligence (AI), with a particular focus on machine learning (ML), across the entire drug discovery pipeline. Due to the inherent complexity, escalating costs, prolonged timelines, and high failure rates of traditional drug discovery methods, there is a critical need to comprehensively understand how AI/ML can be effectively integrated throughout the full process. Currently available literature reviews often narrowly focus on specific phases or methodologies, neglecting the dependence between key stages such as target identification, hit screening, and lead optimization. To bridge this gap, our review provides a detailed and holistic analysis of AI/ML applications across these core phases, highlighting significant methodological advances and their impacts at each stage. We further illustrate the practical impact of these techniques through an in-depth case study focused on hyperuricemia, gout arthritis, and hyperuricemic nephropathy, highlighting real-world successes in molecular target identification and therapeutic candidate discovery. Additionally, we discuss significant challenges facing AI/ML in drug discovery and outline promising future research directions. Ultimately, this review serves as an essential orientation for researchers aiming to leverage AI/ML to overcome existing bottlenecks and accelerate drug discovery.

cs.AI↗

arXiv2025

Hardware and Software Platform Inference

It is now a common business practice to buy access to large language model (LLM) inference rather than self-host, because of significant upfront hardware infrastructure and energy costs. However, as a buyer, there is no mechanism to verify the authenticity of the advertised service including the serving hardware platform, e.g. that it is actually being served using an NVIDIA H100. Furthermore, there are reports suggesting that model providers may deliver models that differ slightly from the advertised ones, often to make them run on less expensive hardware. That way, a client pays premium for a capable model access on more expensive hardware, yet ends up being served by a (potentially less capable) cheaper model on cheaper hardware. In this paper we introduce hardware and software platform inference (HSPI) -- a method for identifying the underlying GPU architecture and software stack of a (black-box) machine learning model solely based on its input-output behavior. Our method leverages the inherent differences of various GPU architectures and compilers to distinguish between different GPU types and software stacks. By analyzing the numerical patterns in the model's outputs, we propose a classification framework capable of accurately identifying the GPU used for model inference as well as the underlying software configuration. Our findings demonstrate the feasibility of inferring GPU type from black-box models. We evaluate HSPI against models served on different real hardware and find that in a white-box setting we can distinguish between different GPUs with between $83.9\%$ and $100\%$ accuracy. Even in a black-box setting we achieve results that are up to 3x higher than random guess accuracy. Our code is available at https://github.com/ChengZhang-98/HSPI.

cs.LG↗

arXiv2025

$L^p$-Logvinenko-Sereda sets and $L^p$-Carleson measures on compact manifolds

Marzo and Ortega-Cerdà gave geometric characterizations for $L^p$-Logvinenko-Sereda sets on the standard sphere for all $1\le p<\infty$. Later, Ortega-Cerdà and Pridhnani further investigated $L^2$-Logvinenko-Sereda sets and $L^2$-Carleson measures on compact manifolds without boundary. In this paper, we characterize $L^p$-Logvinenko-Sereda sets and $L^p$-Carleson measures on compact manifolds with or without boundary for all $1 \frac{2m}{m-1}$. For the range $p < \frac{2m}{m-1}$, we conjecture that $L^p$-Logvinenko-Sereda sets for eigenfunctions on the standard sphere $S^m$ are characterized by the tubular geometric control condition and we provide some evidence. These results provide new progress on an open problem raised by Ortega-Cerdà and Pridhnani.

math.AP↗

arXiv2025

Refining Salience-Aware Sparse Fine-Tuning Strategies for Language Models

Parameter-Efficient Fine-Tuning (PEFT) has gained prominence through low-rank adaptation methods like LoRA. In this paper, we focus on sparsity-based PEFT (SPEFT), which introduces trainable sparse adaptations to the weight matrices in the model, offering greater flexibility in selecting fine-tuned parameters compared to low-rank methods. We conduct the first systematic evaluation of salience metrics for SPEFT, inspired by zero-cost NAS proxies, and identify simple gradient-based metrics is reliable, and results are on par with the best alternatives, offering both computational efficiency and robust performance. Additionally, we compare static and dynamic masking strategies, finding that static masking, which predetermines non-zero entries before training, delivers efficiency without sacrificing performance, while dynamic masking offers no substantial benefits. Across NLP tasks, a simple gradient-based, static SPEFT consistently outperforms other fine-tuning methods for LLMs, providing a simple yet effective baseline for SPEFT. Our work challenges the notion that complexity is necessary for effective PEFT, while our open-source framework establishes a reproducible benchmark for future research, which is available at [https://github.com/0-ml/speft].

cs.CL↗

arXiv2025

TaxaDiffusion: Progressively Trained Diffusion Model for Fine-Grained Species Generation

We propose TaxaDiffusion, a taxonomy-informed training framework for diffusion models to generate fine-grained animal images with high morphological and identity accuracy. Unlike standard approaches that treat each species as an independent category, TaxaDiffusion incorporates domain knowledge that many species exhibit strong visual similarities, with distinctions often residing in subtle variations of shape, pattern, and color. To exploit these relationships, TaxaDiffusion progressively trains conditioned diffusion models across different taxonomic levels -- starting from broad classifications such as Class and Order, refining through Family and Genus, and ultimately distinguishing at the Species level. This hierarchical learning strategy first captures coarse-grained morphological traits shared by species with common ancestors, facilitating knowledge transfer before refining fine-grained differences for species-level distinction. As a result, TaxaDiffusion enables accurate generation even with limited training samples per species. Extensive experiments on three fine-grained animal datasets demonstrate that outperforms existing approaches, achieving superior fidelity in fine-grained animal image generation. Project page: https://amink8.github.io/TaxaDiffusion/

cs.CV↗

arXiv2025

Maximizing Quantum Enhancement in Axion Dark Matter Experiments

We provide a comprehensive comparison of linear amplifiers and microwave photon-counters in axion dark matter experiments. The study is done assuming a range of realistic operating conditions and detector parameters, over the frequency range between 1--30 GHz. As expected, photon counters are found to be advantageous under low background, at high frequencies ($ν>$ 5 GHz), if they can be implemented with robust wide-frequency tuning or a very low dark count rate. Additional noteworthy observations emerging from this study include: (1) an expanded applicability of off-resonance photon background reduction, including the single-quadrature state squeezing, for scan rate enhancements; (2) a much broader appeal for operating the haloscope resonators in the over-coupling regime, up to $β\sim 10$; (3) the need for a detailed investigation into the cryogenic and electromagnetic conditions inside haloscope cavities to lower the photon temperature for future experiments; (4) the necessity to develop a distributed network of coupling ports in high-volume axion haloscopes to utilize these potential gains in the scan rate.

hep-ex↗