arXiv subjects

Xin Zhang

Publications and source records attributed to Xin Zhang.

At least 433 records · Page 24Linked to original sources

arXiv2025

From Flat to Round: Redefining Brain Decoding with Surface-Based fMRI and Cortex Structure

Reconstructing visual stimuli from human brain activity (e.g., fMRI) bridges neuroscience and computer vision by decoding neural representations. However, existing methods often overlook critical brain structure-function relationships, flattening spatial information and neglecting individual anatomical variations. To address these issues, we propose (1) a novel sphere tokenizer that explicitly models fMRI signals as spatially coherent 2D spherical data on the cortical surface; (2) integration of structural MRI (sMRI) data, enabling personalized encoding of individual anatomical variations; and (3) a positive-sample mixup strategy for efficiently leveraging multiple fMRI scans associated with the same visual stimulus. Collectively, these innovations enhance reconstruction accuracy, biological interpretability, and generalizability across individuals. Experiments demonstrate superior reconstruction performance compared to SOTA methods, highlighting the effectiveness and interpretability of our biologically informed approach.

cs.CV↗

arXiv2025

Constraining parity and Lorentz violations in gravity with future ground- and space-based gravitational wave detectors

The future ground- and space-based gravitational wave (GW) detectors offer unprecedented opportunities to test general relativity (GR) with greater precision. In this work, we investigate the capability of future ground-based GW detectors, the Einstein Telescope (ET) and the Cosmic Explorer (CE), and space-based GW detectors, LISA, Taiji, and TianQin, for constraining parity and Lorentz violations in gravity. We inject several typical GW signals from compact binary systems into GW detectors and perform Bayesian inferences with the modified waveforms with parity and Lorentz-violating effects. These effects are modeled in the amplitude and phase corrections to the GW waveforms with their frequency-dependence described by factors $β_ν$, $β_μ$, $β_{\bar ν}$, and $β_{\bar μ}$. Our results show that the combined observations of ET and CE will impose significantly tighter bounds on the energy scale of parity and Lorentz violations ($M_{\rm PV}$ and $M_{\rm LV}$) compared to those given by LIGO-Virgo-KAGRA (LVK) detectors. For cases with positive values of $β_ν$, $β_μ$, $β_{\bar ν}$, and $β_{\bar μ}$, the constraints on $M_{\rm PV}$ and $M_{\rm LV}$ from ground-based detectors are tighter than those from the space-based detectors. For the $β_μ = -1$ case, space-based GW detectors provide constraints on $M_{\rm PV}$ that are better than current LVK observations and comparable to those from ET and CE. Additionally, space-based detectors exhibit superior sensitivity in constraining $M_{\rm LV}$ for $β_{\bar μ} = -2$ case, which is approximately three orders of magnitude tighter than those from ground-based GW detectors. This scenario also enables bounds on the graviton mass at $m_g \lesssim 10^{-35}\; {\rm GeV}$. These findings highlight the promising role of future GW observatories in probing fundamental physics beyond GR.

gr-qc↗

arXiv2025

Prospects for probing dark matter particles and primordial black holes with the Hongmeng mission using the 21 cm global spectrum at cosmic dawn

Probing dark matter particles and primordial black holes remains a pivotal challenge in modern cosmology. Exotic energy injections from dark matter annihilation, decay, and PBH Hawking evaporation can alter the thermal and ionization histories of the early universe, leaving distinctive imprints on the 21 cm global spectrum. We assess the potential of the upcoming space project, the Hongmeng mission, to probe dark matter particles and PBHs using the 21 cm global spectrum. Under ideal conditions with 1000 hours of integration time and negligible foreground residuals, the Hongmeng project can reach sensitivities to dark matter annihilation cross sections and decay lifetimes to $\langle σv \rangle \sim 10^{-28}\,\mathrm{cm^3\,s^{-1}}$ and $τ\sim 10^{28}\,\mathrm{s}$, respectively, for dark matter particles with a mass of $10\,\mathrm{GeV}$. It can also probe PBHs with masses of $10^{16}\,\mathrm{g}$ and abundances as low as $f_{\mathrm{PBH}} \simeq 10^{-6}$. These results indicate that the Hongmeng mission can improve current constraints on dark matter annihilation, decay, and PBH Hawking radiation by nearly two orders of magnitude. Moreover, the Hongmeng mission surpasses current limits on sub-GeV dark matter probing and enables the probing of Hawking radiation from PBHs with masses above $10^{17}\,\mathrm{g}$, which remain undetectable through conventional cosmological means. Overall, the upcoming Hongmeng project holds great promise for advancing the investigation of both dark matter and PBHs, potentially deepening our understanding of the nature of dark matter.

astro-ph.CO↗

arXiv2025

External Large Foundation Model: How to Efficiently Serve Trillions of Parameters for Online Ads Recommendation

Ads recommendation is a prominent service of online advertising systems and has been actively studied. Recent studies indicate that scaling-up and advanced design of the recommendation model can bring significant performance improvement. However, with a larger model scale, such prior studies have a significantly increasing gap from industry as they often neglect two fundamental challenges in industrial-scale applications. First, training and inference budgets are restricted for the model to be served, exceeding which may incur latency and impair user experience. Second, large-volume data arrive in a streaming mode with data distributions dynamically shifting, as new users/ads join and existing users/ads leave the system. We propose the External Large Foundation Model (ExFM) framework to address the overlooked challenges. Specifically, we develop external distillation and a data augmentation system (DAS) to control the computational cost of training/inference while maintaining high performance. We design the teacher in a way like a foundation model (FM) that can serve multiple students as vertical models (VMs) to amortize its building cost. We propose Auxiliary Head and Student Adapter to mitigate the data distribution gap between FM and VMs caused by the streaming data issue. Comprehensive experiments on internal industrial-scale applications and public datasets demonstrate significant performance gain by ExFM.

cs.IR↗

arXiv2025

Exploring non-cold dark matter in a scenario of dynamical dark energy with DESI DR2 data

Recent observations of DESI hint that dark matter (DM) may not be cold but have a non-zero equation of state (EoS) parameter, and that dark energy (DE) may not be a cosmological constant. In this work, we explore the possibility of a non-zero DM EoS parameter within the framework of dynamical DE. We perform analysis by using the latest baryon acoustic oscillation (BAO) data from DESI DR2, the cosmic microwave background (CMB) data from Planck, and the type Ia supernova (SN) data from DESY5 and PantheonPlus. When using the combination of CMB, BAO, and SN data, our results indicate a preference for a non-zero DM EoS parameter at the $2.8σ$ and $3.3σ$ level within the content of a constant DE EoS. In contrast, for a time-evolving DE EoS parameterized by $w_0$ and $w_a$, this preference decreases to $0.8σ$ and $1.1σ$. Furthermore, allowing a non-zero DM EoS yields best-fit values of $w_0$ and $w_a$ that exhibit smaller deviations from the $Λ$CDM expectations, and Bayesian evidence analysis shows a comparable preference for this model relative to $Λ$CDM. The overall results of this work indicate that a non-zero DM EoS parameter warrants further exploration and investigation.

astro-ph.CO↗

arXiv2025

Search for exotic gravitational wave signals beyond general relativity using deep learning

The direct detection of gravitational waves by LIGO has confirmed general relativity (GR) and sparked rapid growth in gravitational wave (GW) astronomy. However, subtle post-Newtonian (PN) deviations observed during the analysis of high signal-to-noise ratio events from the observational runs suggest that standard waveform templates, which assume strict adherence to GR, might overlook signals from alternative theories of gravity. Incorporating these exotic signals into traditional search algorithms is computationally infeasible due to the vast template space required. This paper introduces a proof-of-principle deep learning framework for detecting exotic GW signals, leveraging neural networks trained on GR-based templates. Through their generalization ability, neural networks learn intricate features from the data, enabling the detection of signals that deviate from GR. We present the first study evaluating the capability of deep learning to detect beyond-GR signals, including a variety of PN orders. Our model achieves rapid and accurate identification of exotic GW signals across different luminosity distances, with performance comparable to GR-based detections. In particular, applying the model to the GW150914 event demonstrates excellent performance, highlighting the potential of AI-driven methods for detecting previously overlooked signals beyond GR. This work paves the way for new discoveries in gravitational wave astronomy, enabling the detection of signals that might escape traditional search pipelines.

gr-qc↗

arXiv2025

XY-Tokenizer: Mitigating the Semantic-Acoustic Conflict in Low-Bitrate Speech Codecs

Speech codecs serve as bridges between speech signals and large language models. An ideal codec for speech language models should not only preserve acoustic information but also capture rich semantic information. However, existing speech codecs struggle to balance high-quality audio reconstruction with ease of modeling by language models. In this study, we analyze the limitations of previous codecs in balancing semantic richness and acoustic fidelity. We propose XY-Tokenizer, a novel codec that mitigates the conflict between semantic and acoustic capabilities through multi-stage, multi-task learning. Experimental results demonstrate that XY-Tokenizer achieves performance in both semantic and acoustic tasks comparable to that of state-of-the-art codecs operating at similar bitrates, even though those existing codecs typically excel in only one aspect. Specifically, XY-Tokenizer achieves strong text alignment, surpassing distillation-based semantic modeling methods such as SpeechTokenizer and Mimi, while maintaining a speaker similarity score of 0.83 between reconstructed and original audio. The reconstruction performance of XY-Tokenizer is comparable to that of BigCodec, the current state-of-the-art among acoustic-only codecs, which achieves a speaker similarity score of 0.84 at a similar bitrate. Code and models are available at https://github.com/gyt1145028706/XY-Tokenizer.

cs.SD↗

arXiv2025

On the solution operators arising from the gas-liquid two-phase problem in unbounded domains with finite depth

This paper studies some evolution equations arising from the sharp interface problem of the compressible-incompressible Navier-Stokes equations in unbounded domains in $\mathbb{R}^N (N\geq2)$, where the viscous gases initially occupy the upper half space and the viscous liquids below initially lie in the strip-like domain. In order to establish the maximal $L_p$-$L_q$ regularity estimates of the evolution problem, we construct the R-solver of the resolvent problem associated to the gas-liquid two-phase operator. The crucial part of our proof lies in the analysis of the explicit resolvent operators defined in the unbounded domains with flat boundaries.

math.AP↗

arXiv2025

Open-loop and closed-loop solvabilities for zero-sum stochastic linear quadratic differential games of Markovian regime switching system

This paper investigates zero-sum stochastic linear quadratic (SLQ) differential games with Markovian jumps. Open-loop and closed-loop solvabilities are studied by employing a new ``decomposition method", which decomposes the open-loop and closed-loop solvability problems of zero-sum SLQ differential games into two coupled SLQ control problems. Under the uniform convexity-concavity condition, we construct the open-loop saddle point along with its closed-loop representation based on the solution to a system of constrained coupled differential Riccati equations (CDREs), whose solvability is also established by employing the dimension extension technique and the continuation method. Finally, we provide a concrete example and present its closed-form saddle point based on the theoretical results obtained.

math.OC↗

arXiv2025

No Eigenvalues Outside the Limiting Support of Generally Correlated and Noncentral Sample Covariance Matrices

Spectral properties of random matrices play an important role in statistics, machine learning, communications, and many other areas. Engaging results regarding the convergence of the empirical spectral distribution (ESD) and the ``no-eigenvalue'' property have been obtained for random matrices with different correlation structures. However, the related spectral analysis for generally correlated and noncentral random matrices is still incomplete, and this paper aims to fill this research gap. Specifically, we consider matrices whose columns are independent but with non-zero means and non-identical correlations. Under high-dimensional asymptotics where both the number of rows and columns grow simultaneously to infinity, we first establish the almost sure convergence of the ESD for the concerned random matrices to a deterministic limit, assuming mild conditions. Furthermore, we prove that with probability 1, no eigenvalues will appear in any closed interval outside the support of the limiting distribution for matrices with sufficiently large dimensions. The above results can be applied to different areas such as statistics, wireless communications, and signal processing. In this paper, we apply the derived results to two communication scenarios: 1) We determine the limiting performance of the signal-to-interference-plus-noise ratio for multi-user multiple-input multiple-output (MIMO) systems with linear minimum mean-square error receivers; and 2) We establish the invertibility of zero-forcing precoding matrices in downlink MIMO systems, providing theoretical guarantees.

math.ST↗

arXiv2025

Specific Wasserstein divergence between continuous martingales

Defining a divergence between the laws of continuous martingales is a delicate task, owing to the fact that these laws tend to be singular to each other. An important idea, put forward by N. Gantert, is to instead consider a scaling limit of the relative entropy between such continuous martingales sampled over a finite time grid. This gives rise to the concept of specific relative entropy. In order to develop a general theory of divergences between continuous martingales, it is only natural to replace the role of the relative entropy in this construction by a different notion of discrepancy between finite dimensional probability distributions. In the present work we take a first step in this direction, taking a power $p$ of the Wasserstein distance instead of the relative entropy. We call the newly obtained scaling limit the specific $p$-Wasserstein divergence. In our first main result we prove that the specific $p$-Wasserstein divergence is well-defined, exhibit an explicit expression for it in terms of the quadratic variations of the martingales involved, and compare it with the specific relative entropy and adapted Wasserstein distance on a class of SDEs. Next we consider the specific $p$-Wasserstein divergence optimization over the set of win-martingales. In our second main result we characterize the solution of this optimization problem for all $p>0$ and, somewhat surprisingly, we single out the case $p=1/2$ as the one with the best probabilistic properties. For instance, the optimal martingale in this case is very explicit and can be connected to the solution of a variant of the Schrödinger problem.

math.PR↗

arXiv2025

pUniFind: a unified large pre-trained deep learning model pushing the limit of mass spectra interpretation

Deep learning has advanced mass spectrometry data interpretation, yet most models remain feature extractors rather than unified scoring frameworks. We present pUniFind, the first large-scale multimodal pre-trained model in proteomics that integrates end-to-end peptide-spectrum scoring with open, zero-shot de novo sequencing. Trained on over 100 million open search-derived spectra, pUniFind aligns spectral and peptide modalities via cross modality prediction and outperforms traditional engines across diverse datasets, particularly achieving a 42.6 percent increase in the number of identified peptides in immunopeptidomics. Supporting over 1,300 modifications, pUniFind identifies 60 percent more PSMs than existing de novo methods despite a 300-fold larger search space. A deep learning based quality control module further recovers 38.5 percent additional peptides including 1,891 mapped to the genome but absent from reference proteomes while preserving full fragment ion coverage. These results establish a unified, scalable deep learning framework for proteomic analysis, offering improved sensitivity, modification coverage, and interpretability.

cs.LG↗

arXiv2025

RoboScape: Physics-informed Embodied World Model

World models have become indispensable tools for embodied intelligence, serving as powerful simulators capable of generating realistic robotic videos while addressing critical data scarcity challenges. However, current embodied world models exhibit limited physical awareness, particularly in modeling 3D geometry and motion dynamics, resulting in unrealistic video generation for contact-rich robotic scenarios. In this paper, we present RoboScape, a unified physics-informed world model that jointly learns RGB video generation and physics knowledge within an integrated framework. We introduce two key physics-informed joint training tasks: temporal depth prediction that enhances 3D geometric consistency in video rendering, and keypoint dynamics learning that implicitly encodes physical properties (e.g., object shape and material characteristics) while improving complex motion modeling. Extensive experiments demonstrate that RoboScape generates videos with superior visual fidelity and physical plausibility across diverse robotic scenarios. We further validate its practical utility through downstream applications including robotic policy training with generated data and policy evaluation. Our work provides new insights for building efficient physics-informed world models to advance embodied intelligence research. The code is available at: https://github.com/tsinghua-fib-lab/RoboScape.

cs.CV↗

arXiv2025

Moderating the Generalization of Score-based Generative Model

Score-based Generative Models (SGMs) have demonstrated remarkable generalization abilities, e.g. generating unseen, but natural data. However, the greater the generalization power, the more likely the unintended generalization, and the more dangerous the abuse. Research on moderated generalization in SGMs remains limited. To fill this gap, we first examine the current 'gold standard' in Machine Unlearning (MU), i.e., re-training the model after removing the undesirable training data, and find it does not work in SGMs. Further analysis of score functions reveals that the MU 'gold standard' does not alter the original score function, which explains its ineffectiveness. Based on this insight, we propose the first Moderated Score-based Generative Model (MSGM), which introduces a novel score adjustment strategy that redirects the score function away from undesirable data during the continuous-time stochastic differential equation process. Extensive experimental results demonstrate that MSGM significantly reduces the likelihood of generating undesirable content while preserving high visual quality for normal image generation. Albeit designed for SGMs, MSGM is a general and flexible MU framework that is compatible with diverse diffusion architectures (SGM and DDPM) and training strategies (re-training and fine-tuning), and enables zero-shot transfer of the pre-trained models to downstream tasks, e.g. image inpainting and reconstruction. The code will be shared upon acceptance.

cs.LG↗

arXiv2025

Data Efficacy for Language Model Training

Data is fundamental to the training of language models (LM). Recent research has been dedicated to data efficiency, which aims to maximize performance by selecting a minimal or optimal subset of training data. Techniques such as data filtering, sampling, and selection play a crucial role in this area. To complement it, we define Data Efficacy, which focuses on maximizing performance by optimizing the organization of training data and remains relatively underexplored. This work introduces a general paradigm, DELT, for considering data efficacy in LM training, which highlights the significance of training data organization. DELT comprises three components: Data Scoring, Data Selection, and Data Ordering. Among these components, we design Learnability-Quality Scoring (LQS), as a new instance of Data Scoring, which considers both the learnability and quality of each data sample from the gradient consistency perspective. We also devise Folding Ordering (FO), as a novel instance of Data Ordering, which addresses issues such as model forgetting and data distribution bias. Comprehensive experiments validate the data efficacy in LM training, which demonstrates the following: Firstly, various instances of the proposed DELT enhance LM performance to varying degrees without increasing the data scale and model size. Secondly, among these instances, the combination of our proposed LQS for data scoring and Folding for data ordering achieves the most significant improvement. Lastly, data efficacy can be achieved together with data efficiency by applying data selection. Therefore, we believe that data efficacy is a promising foundational area in LM training.

cs.CL↗

arXiv2025

A Novel Large Vision Foundation Model (LVFM)-based Approach for Generating High-Resolution Canopy Height Maps in Plantations for Precision Forestry Management

Accurate, cost-effective monitoring of plantation aboveground biomass (AGB) is crucial for supporting local livelihoods and carbon sequestration initiatives like the China Certified Emission Reduction (CCER) program. High-resolution canopy height maps (CHMs) are essential for this, but standard lidar-based methods are expensive. While deep learning with RGB imagery offers an alternative, accurately extracting canopy height features remains challenging. To address this, we developed a novel model for high-resolution CHM generation using a Large Vision Foundation Model (LVFM). Our model integrates a feature extractor, a self-supervised feature enhancement module to preserve spatial details, and a height estimator. Tested in Beijing's Fangshan District using 1-meter Google Earth imagery, our model outperformed existing methods, including conventional CNNs. It achieved a mean absolute error of 0.09 m, a root mean square error of 0.24 m, and a correlation of 0.78 against lidar-based CHMs. The resulting CHMs enabled over 90% success in individual tree detection, high accuracy in AGB estimation, and effective tracking of plantation growth, demonstrating strong generalization to non-training areas. This approach presents a promising, scalable tool for evaluating carbon sequestration in both plantations and natural forests.

cs.CV↗

arXiv2025

Benchmarking nuclear matrix elements of $0νββ$ decay with high-energy nuclear collisions

Reducing uncertainties in the nuclear matrix elements (NMEs) remains a critical challenge in designing and interpreting experiments aimed at discovering neutrinoless double beta ($0νββ$) decay. Here, we identify a class of observables, distinct from those employed in low-energy nuclear structure applications, that are strongly correlated with the NMEs: momentum correlations among hadrons produced in high-energy nuclear collisions. Focusing on the $^{150}$Nd$\rightarrow$$^{150}$Sm transition, we combine a Bayesian analysis of the structure of $^{150}$Nd with simulations of high-energy $^{150}$Nd+$^{150}$Nd collisions. We reveal prominent correlations between the NMEs and features of the quark-gluon plasma (QGP) formed in these processes, such as spatial gradients and anisotropies, which are accessible via collective flow measurements. Our findings demonstrate collider experiments involving $0νββ$ decay candidates as a platform for benchmarking theoretical predictions of the NMEs.

nucl-th↗

arXiv2025

Walsh Spider Diffusions as Time Changed Multi-parameter Processes

Inspired by allocation strategies in multi-armed bandit model, we propose a pathwise construction of Walsh spider diffusions. For any infinitesimal generator on a star shaped graph, there exists a unique time change associated with a multi-parameter process such that the time change of this multi-parameter process is the desired diffusion. The time change has an interpretation of time allocation of the process on each edge, and it can be derived explicitly from a family of equations.

math.PR↗