arXiv subjects

Yu Zhang

Publications and source records attributed to Yu Zhang.

At least 667 records · Page 37Linked to original sources

arXiv2025

Thermodynamic modifications to Bardeen black holes surrounded by quintessence based on the new higher order GUP

In this article, the thermodynamic properties of Bardeen black holes surrounded by quintessence are investigated in the framework of a new higher order GUP. The modified Hawking temperature, entropy and heat capacity are derived using a heuristic approach. Meanwhile, the remnant temperature and mass are deduced, and the modified black hole state equation is obtained by utilizing the energy density of matter. Ultimately, we analyze the effects of the GUP controlling deformation parameter $β$ on these thermodynamic properties using graphical illustrations, to more comprehensively understand the thermodynamic behavior of the black hole in the context of higher order GUP.

gr-qc↗

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

Three flavor QCD phase transition with Möbius domain wall fermions

We present an updated study of the $N_f=3$ QCD phase transition using Möbius domain wall fermions. Simulations were performed on $N_t=12$ lattices with aspect ratios ranging from 2 to 4 for various quark masses, at a lattice spacing of $a=0.1361(20)$ fm, corresponding to a temperature of 121(2) MeV. To clarify the nature of the phase transition, a large-volume lattice, $48^3 \times 12\times 16$, was added to analyze the volume dependence of disconnected chiral susceptibility. By examining the chiral condensate, disconnected chiral susceptibility, and Binder cumulant, and incorporating results from $24^3 \times 12 \times 16$ and $36^3 \times 12 \times 16$ lattices reported in earlier studies, we observe that the transition is consistent with a crossover at a quark mass of approximately $m_f^{\mathrm{\overline {MS}}}(2\, \mathrm{GeV}) \sim 4$ MeV at this temperature. Furthermore, we discuss the effects of residual chiral symmetry breaking on the chiral condensate and disconnected chiral susceptibility for different sizes in the 5th direction.

hep-lat↗

arXiv2025

DanModCap: Designing a Danmaku Moderation Tool for Video-Sharing Platforms that Leverages Impact Captions with Large Language Models

Online video platforms have gained increased popularity due to their ability to support information consumption and sharing and the diverse social interactions they afford. Danmaku, a real-time commentary feature that overlays user comments on a video, has been found to improve user engagement, however, the use of Danmaku can lead to toxic behaviors and inappropriate comments. To address these issues, we propose a proactive moderation approach inspired by Impact Captions, a visual technique used in East Asian variety shows. Impact Captions combine textual content and visual elements to construct emotional and cognitive resonance. Within the context of this work, Impact Captions were used to guide viewers towards positive Danmaku-related activities and elicit more pro-social behaviors. Leveraging Impact Captions, we developed DanModCap, an moderation tool that collected and analyzed Danmaku and used it as input to large generative language models to produce Impact Captions. Our evaluation of DanModCap demonstrated that Impact Captions reduced negative antagonistic emotions, increased users' desire to share positive content, and elicited self-control in Danmaku social action to fostering proactive community maintenance behaviors. Our approach highlights the benefits of using LLM-supported content moderation methods for proactive moderation in a large-scale live content contexts.

cs.HC↗

arXiv2025

Study of symmetries in finite temperature $N_f=2$ QCD with Möbius Domain Wall Fermions

We report on the ongoing study of symmetry of $N_f=2$ QCD around the critical temperature. Our simulations of $N_f = 2$ QCD employ the Möbius domain-wall fermion action with residual mass $\sim 1\mbox{MeV}$ or less, maintaining a good chiral symmetry. Using the screening masses from the two point spatial correlators we compare the mass difference between channels connected through various symmetry transformations. Our analysis focuses on restoration of the $SU(2)_L\times SU(2)_R$ as well as anomalously broken axial $U(1)_A$. We also present additional study of a potential $SU(2)_{CS}$ symmetry which may emerge at sufficiently high temperatures.

hep-lat↗

arXiv2025

EliGen: Entity-Level Controlled Image Generation with Regional Attention

Recent advancements in diffusion models have significantly advanced text-to-image generation, yet global text prompts alone remain insufficient for achieving fine-grained control over individual entities within an image. To address this limitation, we present EliGen, a novel framework for Entity-level controlled image Generation. Firstly, we put forward regional attention, a mechanism for diffusion transformers that requires no additional parameters, seamlessly integrating entity prompts and arbitrary-shaped spatial masks. By contributing a high-quality dataset with fine-grained spatial and semantic entity-level annotations, we train EliGen to achieve robust and accurate entity-level manipulation, surpassing existing methods in both spatial precision and image quality. Additionally, we propose an inpainting fusion pipeline, extending its capabilities to multi-entity image inpainting tasks. We further demonstrate its flexibility by integrating it with other open-source models such as IP-Adapter, In-Context LoRA and MLLM, unlocking new creative possibilities. The source code, model, and dataset are published at https://github.com/modelscope/DiffSynth-Studio.git.

cs.CV↗

arXiv2025

Beyond Prior Limits: Addressing Distribution Misalignment in Particle Filtering

Particle filtering is a Bayesian inference method and a fundamental tool in state estimation for dynamic systems, but its effectiveness is often limited by the constraints of the initial prior distribution, a phenomenon we define as the Prior Boundary Phenomenon. This challenge arises when target states lie outside the prior's support, rendering traditional particle filtering methods inadequate for accurate estimation. Although techniques like unbounded priors and larger particle sets have been proposed, they remain computationally prohibitive and lack adaptability in dynamic scenarios. To systematically overcome these limitations, we propose the Diffusion-Enhanced Particle Filtering Framework, which introduces three key innovations: adaptive diffusion through exploratory particles, entropy-driven regularisation to prevent weight collapse, and kernel-based perturbations for dynamic support expansion. These mechanisms collectively enable particle filtering to explore beyond prior boundaries, ensuring robust state estimation for out-of-boundary targets. Theoretical analysis and extensive experiments validate framework's effectiveness, indicating significant improvements in success rates and estimation accuracy across high-dimensional and non-convex scenarios.

stat.ML↗

arXiv2025

Long-lived sterile neutrinos from an axionlike particle at Belle II

Axionlike particles (ALPs) can be produced in meson decays via tree-level quark-flavor-violating couplings, and decay to a pair of sterile neutrinos if also coupled to them. Such light sterile neutrinos are necessarily long-lived and may give rise to striking signatures of displaced vertices (DVs) at terrestrial experiments. In this work, we study the prospect of the ongoing Belle II experiment for detecting the sterile neutrinos from the ALP, leveraging the $\mathcal{O}(10^{10})$ $B$-mesons projected to be produced at the experiment with an integrated luminosity of 50 ab$^{-1}$. We propose search strategies for one or two DVs, and perform Monte-Carlo simulations in order to estimate the sensitivity reach of Belle II to active-sterile-neutrino mixing $|V_{eN}|^2$ as functions of the sterile-neutrino mass. Signatures include a selected list of sterile-neutrino final states, for which we estimate an $\mathcal{O}(1)$ or lower background level. Our results show that the proposed search strategies can probe values of $|V_{eN}|^2$ up to about two orders of magnitude beyond the existing bounds, for ALP mass over 4 GeV and sterile-neutrino mass above the $D$-meson threshold. Compared to the one-DV search, the two-DV one, despite its weaker sensitivities as a result of double exponential suppression of the sterile-neutrino decay positions, possesses the advantage of possible full reconstruction of the signal event allowing for pinning down the masses of the sterile neutrino and the ALP, if a discovery is made.

hep-ph↗

arXiv2025

Searching for heavy neutral leptons coupled to axion-like particles at the LHC far detectors and SHiP

In hidden-sector models, axion-like particles (ALPs) can couple to heavy neutral leptons (HNLs), leading to rich phenomenologies. We study ALPs produced from $D$- and $B$-meson decays via quark-flavor-violating couplings, and decaying exclusively into a pair of HNLs which mix with active neutrinos. The ALP can be either short- or long-lived, depending on the masses of the ALP and the HNL, as well as the corresponding coupling strength. Such GeV-scale HNLs are necessarily long-lived given the current bounds on their mixing parameters. We assess the sensitivities of the LHC far detectors and SHiP, to the long-lived HNLs in such theoretical scenarios. We find that for currently allowed values of the ALP couplings, most of these experiments can probe the active-sterile-neutrino mixing parameters multiple orders of magnitude beyond the present bounds, covering large parameter region targeted with the type-I seesaw mechanism. In addition, our results show that compared to the case of a promptly decaying ALP, assuming an ALP of longer lifetimes weakens the sensitivities of the considered experiments to the long-lived HNLs.

hep-ph↗

arXiv2025

Searching for long-lived light neutralinos and axionlike particles at the SHiNESS experiment

Recently Soleti et al. [JHEP03(2024)148] proposed a new experiment called SHiNESS at the upcoming European Spallation Source (ESS) facility, making use of the 2-GeV proton beam there impinging on a fixed target, in order to search for hidden sterile neutrinos that could lie in different mass ranges and arise with distinct signatures. Such signatures include excesses in electron-positron pairs that may originate from displaced decays of long-lived particles (LLPs). At the ESS, the dominant sources of such LLPs are decays at rest of $π^+$ mesons and $μ^+$ leptons. We choose to investigate theoretical scenarios of long-lived light neutralinos in the R-parity-violating supersymmetry and long-lived weak-violating electrophilic axionlike particles, as these LLPs can be produced from decays of $π^+$ and $μ^+$. Since the $π^+$'s and $μ^+$'s decay at rest at the ESS, we compute the spectra of the therefrom produced LLPs, and thus estimate the expected sensitivity reach of SHiNESS to the LLPs in these two scenarios. Our calculation shows that in multiple relevant benchmark scenarios, SHiNESS can probe large parameter regions of these models beyond the existing bounds, in just a couple of years of data-collection time.

hep-ph↗

arXiv2025

3DGS$^2$: Near Second-order Converging 3D Gaussian Splatting

3D Gaussian Splatting (3DGS) has emerged as a mainstream solution for novel view synthesis and 3D reconstruction. By explicitly encoding a 3D scene using a collection of Gaussian kernels, 3DGS achieves high-quality rendering with superior efficiency. As a learning-based approach, 3DGS training has been dealt with the standard stochastic gradient descent (SGD) method, which offers at most linear convergence. Consequently, training often requires tens of minutes, even with GPU acceleration. This paper introduces a (near) second-order convergent training algorithm for 3DGS, leveraging its unique properties. Our approach is inspired by two key observations. First, the attributes of a Gaussian kernel contribute independently to the image-space loss, which endorses isolated and local optimization algorithms. We exploit this by splitting the optimization at the level of individual kernel attributes, analytically constructing small-size Newton systems for each parameter group, and efficiently solving these systems on GPU threads. This achieves Newton-like convergence per training image without relying on the global Hessian. Second, kernels exhibit sparse and structured coupling across input images. This property allows us to effectively utilize spatial information to mitigate overshoot during stochastic training. Our method converges an order faster than standard GPU-based 3DGS training, requiring over $10\times$ fewer iterations while maintaining or surpassing the quality of the compared with the SGD-based 3DGS reconstructions.

cs.CV↗

arXiv2025

Searching for long-lived particles from stopped pions and muons at the CiADS-BDE

The CiADS-BDE is a beam-dump experiment recently proposed for searching for light, long-lived particles (LLPs) at China initiative Accelerator Driven System. Primarily thanks to the large numbers of protons on target at the experiment, it has been shown to be sensitive to large, unique regions of the parameter space of dark photon, with a small detector volume of $\mathcal{O}(0.01\text{--}1)$ m$^3$. Here, we explore the search prospect of the CiADS-BDE for a series of new-physic models predicting LLPs that could emanate from decays at rest of charged pions and muons at the facility, namely, heavy neutral leptons, axionlike particles, and light binos in the R-parity-violating supersymmetry. For these benchmark models, we find that the CiADS-BDE can also probe vast parameter regions beyond the existing bounds.

hep-ph↗

arXiv2025

Exploring the Feasibility of Deep Learning Models for Long-term Disease Prediction: A Case Study for Wheat Yellow Rust in England

Wheat yellow rust, caused by the fungus Puccinia striiformis, is a critical disease affecting wheat crops across Britain, leading to significant yield losses and economic consequences. Given the rapid environmental changes and the evolving virulence of pathogens, there is a growing need for innovative approaches to predict and manage such diseases over the long term. This study explores the feasibility of using deep learning models to predict outbreaks of wheat yellow rust in British fields, offering a proactive approach to disease management. We construct a yellow rust dataset with historial weather information and disease indicator acrossing multiple regions in England. We employ two poweful deep learning models, including fully connected neural networks and long short-term memory to develop predictive models capable of recognizing patterns and predicting future disease outbreaks.The models are trained and validated in a randomly sliced datasets. The performance of these models with different predictive time steps are evaluated based on their accuracy, precision, recall, and F1-score. Preliminary results indicate that deep learning models can effectively capture the complex interactions between multiple factors influencing disease dynamics, demonstrating a promising capacity to forecast wheat yellow rust with considerable accuracy. Specifically, the fully-connected neural network achieved 83.65% accuracy in a disease prediction task with 6 month predictive time step setup. These findings highlight the potential of deep learning to transform disease management strategies, enabling earlier and more precise interventions. Our study provides a methodological framework for employing deep learning in agricultural settings but also opens avenues for future research to enhance the robustness and applicability of predictive models in combating crop diseases globally.

cs.LG↗

arXiv2025

CvhSlicer 2.0: Immersive and Interactive Visualization of Chinese Visible Human Data in XR Environments

The study of human anatomy through advanced visualization techniques is crucial for medical research and education. In this work, we introduce CvhSlicer 2.0, an innovative XR system designed for immersive and interactive visualization of the Chinese Visible Human (CVH) dataset. Particularly, our proposed system operates entirely on a commercial XR headset, offering a range of visualization and interaction tools for dynamic 2D and 3D data exploration. By conducting comprehensive evaluations, our CvhSlicer 2.0 demonstrates strong capabilities in visualizing anatomical data, enhancing user engagement and improving educational effectiveness. A demo video is available at https://youtu.be/CfR72S_0N-4

cs.HC↗

arXiv2025

Acoustic Emission Sensor Network Optimization Based on Grid Loop Search and Particle Swarm Source Location

The layout of acoustic emission sensors plays a critical role in non-destructive structural testing. This study proposes a grid-based optimization method focused on multi-source location results, in contrast to traditional sensor layout optimization methods that construct a correlation matrix based on sensor layout and one source location. Based on the seismic source travel-time theory, the proposed method establishes a location objective function based on minimum travel-time differences, which is solved through the particle swarm optimization (PSO) algorithm. Furthermore, based on location accuracy across various configurations, the method systematically evaluates potential optimal sensor locations through grid search. Synthetic tests and laboratory pencil-lead break (PLB) experiments are conducted to compare the effectiveness of PSO, genetic algorithm, and simulated annealing, with the following conclusions: (1) In synthetic tests, the proposed method achieved an average location error of 1.78 mm, outperforming that based on the traditional layout, genetic algorithm (GA), and simulated annealing (SA). (2) For different noise cases, the location accuracy separately improved by 24.89% (σ=0.5μs), 12.59% (σ=2μs), and 15.06% (σ=5μs) compared with the traditional layout. (3) For the PLB experiments, the optimized layout achieved an average location error of 9.37 mm, which improved the location accuracy by 59.15% compared with the Traditional layout.

physics.geo-ph↗

arXiv2025

Elliptic Harnack inequality and its applications on Finsler metric measure spaces

In this paper, we study the elliptic Harnack inequality and its applications on forward complete Finsler metric measure spaces under the conditions that the weighted Ricci curvature ${\rm Ric}_{\infty}$ has non-positive lower bound and the distortion $τ$ is of linear growth, $|τ|\leq ar+b$, where $a,b$ are some non-negative constants, $r=d(x_0,x)$ is the distance function for some point $x_{0} \in M$. We obtain an elliptic $p$-Harnack inequality for positive harmonic functions from a local uniform Poincaré inequality and a mean value inequality. As applications of the Harnack inequality, we derive the Hölder continuity estimate and a Liouville theorem for positive harmonic functions. Furthermore, we establish a gradient estimate for positive harmonic functions.

math.DG↗

arXiv2025

Parallelizing Linear Transformers with the Delta Rule over Sequence Length

Transformers with linear attention (i.e., linear transformers) and state-space models have recently been suggested as a viable linear-time alternative to transformers with softmax attention. However, these models still underperform transformers especially on tasks that require in-context retrieval. While more expressive variants of linear transformers which replace the additive update in linear transformers with the delta rule (DeltaNet) have been found to be more effective at associative recall, existing algorithms for training such models do not parallelize over sequence length and are thus inefficient to train on modern hardware. This work describes a hardware-efficient algorithm for training linear transformers with the delta rule, which exploits a memory-efficient representation for computing products of Householder matrices. This algorithm allows us to scale up DeltaNet to standard language modeling settings. We train a 1.3B model for 100B tokens and find that it outperforms recent linear-time baselines such as Mamba and GLA in terms of perplexity and zero-shot performance on downstream tasks. We also experiment with two hybrid models which combine DeltaNet layers with (1) sliding-window attention layers every other layer or (2) two global attention layers, and find that these hybrids outperform strong transformer baselines.

cs.LG↗

arXiv2025

Text-Guided Coarse-to-Fine Fusion Network for Robust Remote Sensing Visual Question Answering

Remote Sensing Visual Question Answering (RSVQA) has gained significant research interest. However, current RSVQA methods are limited by the imaging mechanisms of optical sensors, particularly under challenging conditions such as cloud-covered and low-light scenarios. Given the all-time and all-weather imaging capabilities of Synthetic Aperture Radar (SAR), it is crucial to investigate the integration of optical-SAR images to improve RSVQA performance. In this work, we propose a Text-guided Coarse-to-Fine Fusion Network (TGFNet), which leverages the semantic relationships between question text and multi-source images to guide the network toward complementary fusion at the feature level. Specifically, we develop a Text-guided Coarse-to-Fine Attention Refinement (CFAR) module to focus on key areas related to the question in complex remote sensing images. This module progressively directs attention from broad areas to finer details through key region routing, enhancing the model's ability to focus on relevant regions. Furthermore, we propose an Adaptive Multi-Expert Fusion (AMEF) module that dynamically integrates different experts, enabling the adaptive fusion of optical and SAR features. In addition, we create the first large-scale benchmark dataset for evaluating optical-SAR RSVQA methods, comprising 6,008 well-aligned optical-SAR image pairs and 1,036,694 well-labeled question-answer pairs across 16 diverse question types, including complex relational reasoning questions. Extensive experiments on the proposed dataset demonstrate that our TGFNet effectively integrates complementary information between optical and SAR images, significantly improving the model's performance in challenging scenarios. The dataset is available at: https://github.com/mmic-lcl/. Index Terms: Remote Sensing Visual Question Answering, Multi-source Data Fusion, Multimodal, Remote Sensing, OPT-SAR.

cs.CV↗