arXiv subjects

Xin Wang

Publications and source records attributed to Xin Wang.

At least 649 records · Page 36Linked to original sources

arXiv2025

Unveiling Multiple Physical Processes on a Cluster Galaxy at z=0.3 Using JWST

We present a study of a previously identified candidate jellyfish galaxy in the Abell 2744 cluster, F0083, which showed faint signs of a tidal interaction in archival imaging. We used publicly available PSF-matched deep photometric data from the Hubble and James Webb Space Telescopes to infer the spatially resolved star formation history of this galaxy. F0083 shows clear signs of ram-pressure stripping (RPS), with a recently enhanced star formation rate (SFR) orientated towards the south-west quadrant of the stellar disc. The stellar mass surface density is heavily asymmetric, with a variation of nearly 1 dex between the western spiral arm and the postulated tidal feature. This feature appears to contain a high proportion of older stars, ruling out RPS as the cause of this `unwinding'. We identified two potential interaction candidates, 28171 and 26055, with masses $\log_{10}(M_*/\rm{M}_{\odot})=8.56\pm0.06$ and $\log_{10}(M_*/\rm{M}_{\odot})=9.24\pm0.09$, respectively, and projected separations of 31 kpc and 46 kpc. The star formation history (SFH) of the tidal feature in F0083 indicates a steep change in SFR at lookback times $t_L\lesssim1\,$Gyr, consistent with a burst in the SFH of 26055. The most probable formation scenario of F0083 thus indicates a significant tidal interaction, followed by RPS as the combined system approaches pericentre passage. Our results demonstrate that by using photometric data we are able to distinguish between these consecutive processes, and represent the first observational analysis of the contributions of each process at this redshift.

astro-ph.GA↗

arXiv2025

Quantum Recurrent Embedding Neural Network

Quantum neural networks have emerged as promising quantum machine learning models, leveraging the properties of quantum systems and classical optimization to solve complex problems in physics and beyond. However, previous studies have demonstrated inevitable trainability issues that severely limit their capabilities in the large-scale regime. In this work, we propose a quantum recurrent embedding neural network (QRENN) inspired by fast-track information pathways in ResNet and general quantum circuit architectures in quantum information theory. By employing dynamical Lie algebras, we provide a rigorous proof of the trainability of QRENN circuits, demonstrating that this deep quantum neural network can avoid barren plateaus. Notably, the general QRENN architecture resists classical simulation as it encompasses powerful quantum circuits such as QSP, QSVT, and DQC1, which are widely believed to be classically intractable. Building on this theoretical foundation, we apply our QRENN to accurately classify quantum Hamiltonians and detect symmetry-protected topological phases, demonstrating its applicability in quantum supervised learning. Our results highlight the power of recurrent data embedding in quantum neural networks and the potential for scalable quantum supervised learning in predicting physical properties and solving complex problems.

quant-ph↗

arXiv2025

MetaEformer: Unveiling and Leveraging Meta-patterns for Complex and Dynamic Systems Load Forecasting

Time series forecasting is a critical and practical problem in many real-world applications, especially for industrial scenarios, where load forecasting underpins the intelligent operation of modern systems like clouds, power grids and traffic networks.However, the inherent complexity and dynamics of these systems present significant challenges. Despite advances in methods such as pattern recognition and anti-non-stationarity have led to performance gains, current methods fail to consistently ensure effectiveness across various system scenarios due to the intertwined issues of complex patterns, concept-drift, and few-shot problems. To address these challenges simultaneously, we introduce a novel scheme centered on fundamental waveform, a.k.a., meta-pattern. Specifically, we develop a unique Meta-pattern Pooling mechanism to purify and maintain meta-patterns, capturing the nuanced nature of system loads. Complementing this, the proposed Echo mechanism adaptively leverages the meta-patterns, enabling a flexible and precise pattern reconstruction. Our Meta-pattern Echo transformer (MetaEformer) seamlessly incorporates these mechanisms with the transformer-based predictor, offering end-to-end efficiency and interpretability of core processes. Demonstrating superior performance across eight benchmarks under three system scenarios, MetaEformer marks a significant advantage in accuracy, with a 37% relative improvement on fifteen state-of-the-art baselines.

cs.LG↗

arXiv2025

Quantum Fidelity Estimation in the Resource Theory of Nonstabilizerness

Quantum fidelity estimation is essential for benchmarking quantum states and processes on noisy quantum devices. While stabilizer operations form the foundation of fault-tolerant quantum computing, non-stabilizer resources further enable universal quantum computation through state injection. In this work, we propose several efficient fidelity estimation protocols for both quantum states and channels within the resource theory of nonstabilizerness, focusing on qudit systems with odd prime dimensions. Our protocols require measuring only a constant number of phase-space point operator expectation values, with operators selected randomly according to an importance weighting scheme tailored to the target state. Notably, we demonstrate that mathematically defined nonstabilizerness measures--such as Wigner rank and mana--quantify the sample complexity of the proposed protocols, thereby endowing them with a clear operational interpretation in the fidelity estimation task. This connection reveals a fundamental trade-off: while fidelity estimation for general quantum states and channels requires resources that scale exponentially with their nonstabilizerness, the task remains tractable for states and channels that admit efficient classical simulation.

quant-ph↗

arXiv2025

From Sharpness to Better Generalization for Speech Deepfake Detection

Generalization remains a critical challenge in speech deepfake detection (SDD). While various approaches aim to improve robustness, generalization is typically assessed through performance metrics like equal error rate without a theoretical framework to explain model performance. This work investigates sharpness as a theoretical proxy for generalization in SDD. We analyze how sharpness responds to domain shifts and find it increases in unseen conditions, indicating higher model sensitivity. Based on this, we apply Sharpness-Aware Minimization (SAM) to reduce sharpness explicitly, leading to better and more stable performance across diverse unseen test sets. Furthermore, correlation analysis confirms a statistically significant relationship between sharpness and generalization in most test settings. These findings suggest that sharpness can serve as a theoretical indicator for generalization in SDD and that sharpness-aware training offers a promising strategy for improving robustness.

eess.AS↗

arXiv2025

Dynamic Mixture of Curriculum LoRA Experts for Continual Multimodal Instruction Tuning

Continual multimodal instruction tuning is crucial for adapting Multimodal Large Language Models (MLLMs) to evolving tasks. However, most existing methods adopt a fixed architecture, struggling with adapting to new tasks due to static model capacity. We propose to evolve the architecture under parameter budgets for dynamic task adaptation, which remains unexplored and imposes two challenges: 1) task architecture conflict, where different tasks require varying layer-wise adaptations, and 2) modality imbalance, where different tasks rely unevenly on modalities, leading to unbalanced updates. To address these challenges, we propose a novel Dynamic Mixture of Curriculum LoRA Experts (D-MoLE) method, which automatically evolves MLLM's architecture with controlled parameter budgets to continually adapt to new tasks while retaining previously learned knowledge. Specifically, we propose a dynamic layer-wise expert allocator, which automatically allocates LoRA experts across layers to resolve architecture conflicts, and routes instructions layer-wisely to facilitate knowledge sharing among experts. Then, we propose a gradient-based inter-modal continual curriculum, which adjusts the update ratio of each module in MLLM based on the difficulty of each modality within the task to alleviate the modality imbalance problem. Extensive experiments show that D-MoLE significantly outperforms state-of-the-art baselines, achieving a 15% average improvement over the best baseline. To the best of our knowledge, this is the first study of continual learning for MLLMs from an architectural perspective.

cs.CV↗

arXiv2025

DAVID-XR1: Detecting AI-Generated Videos with Explainable Reasoning

As AI-generated video becomes increasingly pervasive across media platforms, the ability to reliably distinguish synthetic content from authentic footage has become both urgent and essential. Existing approaches have primarily treated this challenge as a binary classification task, offering limited insight into where or why a model identifies a video as AI-generated. However, the core challenge extends beyond simply detecting subtle artifacts; it requires providing fine-grained, persuasive evidence that can convince auditors and end-users alike. To address this critical gap, we introduce DAVID-X, the first dataset to pair AI-generated videos with detailed defect-level, temporal-spatial annotations and written rationales. Leveraging these rich annotations, we present DAVID-XR1, a video-language model designed to deliver an interpretable chain of visual reasoning-including defect categorization, temporal-spatial localization, and natural language explanations. This approach fundamentally transforms AI-generated video detection from an opaque black-box decision into a transparent and verifiable diagnostic process. We demonstrate that a general-purpose backbone, fine-tuned on our compact dataset and enhanced with chain-of-thought distillation, achieves strong generalization across a variety of generators and generation modes. Our results highlight the promise of explainable detection methods for trustworthy identification of AI-generated video content.

cs.CV↗

arXiv2025

ConvD: Attention Enhanced Dynamic Convolutional Embeddings for Knowledge Graph Completion

Knowledge graphs often suffer from incompleteness issues, which can be alleviated through information completion. However, current state-of-the-art deep knowledge convolutional embedding models rely on external convolution kernels and conventional convolution processes, which limits the feature interaction capability of the model. This paper introduces a novel dynamic convolutional embedding model, ConvD, which directly reshapes relation embeddings into multiple internal convolution kernels. This approach effectively enhances the feature interactions between relation embeddings and entity embeddings. Simultaneously, we incorporate a priori knowledge-optimized attention mechanism that assigns different contribution weight coefficients to the multiple relation convolution kernels in dynamic convolution, further boosting the expressive power of the model. Extensive experiments on various datasets show that our proposed model consistently outperforms the state-of-the-art baseline methods, with average improvements ranging from 3.28% to 14.69% across all model evaluation metrics, while the number of parameters is reduced by 50.66% to 85.40% compared to other state-of-the-art models.

cs.CL↗

arXiv2025

Virasoro constraints for Hodge integrals

The purpose of this paper is to study Virasoro constraints for Hodge integrals in Gromov-Witten theory of any target varieties. Results consist of the following: Firstly, we propose Virasoro conjecture for Hodge integrals in Gromov-Witten theory of any target varieties; Secondly, we prove Virasoro constraints for Hodge integrals in genus zero of any target varieties; Thirdly, we prove the genus-1 $L_1^{\mathbb{E}}$ constraint with one Hodge character class insertion for any target varieties; Lastly, we obtain certain new vanishing identities in higher genus for Gromov-Witten invariants of any target varieties.

math-ph↗

arXiv2025

Efficient Diffusion Models: A Survey

Diffusion models have emerged as powerful generative models capable of producing high-quality contents such as images, videos, and audio, demonstrating their potential to revolutionize digital content creation. However, these capabilities come at the cost of their significant computational resources and lengthy generation time, underscoring the critical need to develop efficient techniques for practical deployment. In this survey, we provide a systematic and comprehensive review of research on efficient diffusion models. We organize the literature in a taxonomy consisting of three main categories, covering distinct yet interconnected efficient diffusion model topics from algorithm-level, system-level, and framework perspective, respectively. We have also created a GitHub repository where we organize the papers featured in this survey at https://github.com/AIoT-MLSys-Lab/Efficient-Diffusion-Model-Survey. We hope our survey can serve as a valuable resource to help researchers and practitioners gain a systematic understanding of efficient diffusion model research and inspire them to contribute to this important and exciting field.

cs.LG↗

arXiv2025

Modular domain walls and gravitational waves

We discuss modular domain walls and gravitational waves in a class of supersymmetric models where quark and lepton flavour symmetry emerges from modular symmetry. In such models a single modulus field $τ$ is often assumed to be stabilised at or near certain fixed point values such as $τ= {\rm i}$ and $τ= ω$ (the cube root of unity), in its fundamental domain. We show that, in the global supersymmetry limit of certain classes of potentials, the vacua at these fixed points may be degenerate, leading to the formation of modular domain walls in the early Universe. Taking supergravity effects into account, in the background of a fixed dilaton field $S$, the degeneracy may be lifted, leading to a bias term in the potential allowing the domain walls to collapse. We study the resulting gravitational wave spectra arising from the dynamics of such modular domain walls, and assess their observability by current and future experiments, as a window into modular flavour symmetry.

hep-ph↗

arXiv2025

ReactDiff: Latent Diffusion for Facial Reaction Generation

Given the audio-visual clip of the speaker, facial reaction generation aims to predict the listener's facial reactions. The challenge lies in capturing the relevance between video and audio while balancing appropriateness, realism, and diversity. While prior works have mostly focused on uni-modal inputs or simplified reaction mappings, recent approaches such as PerFRDiff have explored multi-modal inputs and the one-to-many nature of appropriate reaction mappings. In this work, we propose the Facial Reaction Diffusion (ReactDiff) framework that uniquely integrates a Multi-Modality Transformer with conditional diffusion in the latent space for enhanced reaction generation. Unlike existing methods, ReactDiff leverages intra- and inter-class attention for fine-grained multi-modal interaction, while the latent diffusion process between the encoder and decoder enables diverse yet contextually appropriate outputs. Experimental results demonstrate that ReactDiff significantly outperforms existing approaches, achieving a facial reaction correlation of 0.26 and diversity score of 0.094 while maintaining competitive realism. The code is open-sourced at \href{https://github.com/Hunan-Tiger/ReactDiff}{github}.

cs.CV↗

arXiv2025

A Model-Data Dual-Driven Resource Allocation Scheme for IREE Oriented 6G Networks

The rapid and substantial fluctuations in wireless network capacity and traffic demand, driven by the emergence of 6G technologies, have exacerbated the issue of traffic-capacity mismatch, raising concerns about wireless network energy consumption. To address this challenge, we propose a model-data dual-driven resource allocation (MDDRA) algorithm aimed at maximizing the integrated relative energy efficiency (IREE) metric under dynamic traffic conditions. Unlike conventional model-driven or data-driven schemes, the proposed MDDRA framework employs a model-driven Lyapunov queue to accumulate long-term historical mismatch information and a data-driven Graph Radial bAsis Fourier (GRAF) network to predict the traffic variations under incomplete data, and hence eliminates the reliance on high-precision models and complete spatial-temporal traffic data. We establish the universal approximation property of the proposed GRAF network and provide convergence and complexity analysis for the MDDRA algorithm. Numerical experiments validate the performance gains achieved through the data-driven and model-driven components. By analyzing IREE and EE curves under diverse traffic conditions, we recommend that network operators shall spend more efforts to balance the traffic demand and the network capacity distribution to ensure the network performance, particularly in scenarios with large speed limits and higher driving visibility.

cs.NI↗

arXiv2025

KG-BiLM: Knowledge Graph Embedding via Bidirectional Language Models

Recent advances in knowledge representation learning (KRL) highlight the urgent necessity to unify symbolic knowledge graphs (KGs) with language models (LMs) for richer semantic understanding. However, existing approaches typically prioritize either graph structure or textual semantics, leaving a gap: a unified framework that simultaneously captures global KG connectivity, nuanced linguistic context, and discriminative reasoning semantics. To bridge this gap, we introduce KG-BiLM, a bidirectional LM framework that fuses structural cues from KGs with the semantic expressiveness of generative transformers. KG-BiLM incorporates three key components: (i) Bidirectional Knowledge Attention, which removes the causal mask to enable full interaction among all tokens and entities; (ii) Knowledge-Masked Prediction, which encourages the model to leverage both local semantic contexts and global graph connectivity; and (iii) Contrastive Graph Semantic Aggregation, which preserves KG structure via contrastive alignment of sampled sub-graph representations. Extensive experiments on standard benchmarks demonstrate that KG-BiLM outperforms strong baselines in link prediction, especially on large-scale graphs with complex multi-hop relations - validating its effectiveness in unifying structural information and textual semantics.

cs.CL↗

arXiv2025

Cryogenic in-memory computing using magnetic topological insulators

Machine learning algorithms have been proven effective for essential quantum computation tasks such as quantum error correction and quantum control. Efficient hardware implementation of these algorithms at cryogenic temperatures is essential. Here, we utilize magnetic topological insulators as memristors (termed magnetic topological memristors) and introduce a cryogenic in-memory computing scheme based on the coexistence of the chiral edge state and the topological surface state. The memristive switching and reading of the giant anomalous Hall effect exhibit high energy efficiency, high stability, and low stochasticity. We achieve high accuracy in a proof-of-concept classification task using four magnetic topological memristors. Furthermore, our algorithm-level and circuit-level simulations of large-scale neural networks demonstrate software-level accuracy and lower energy consumption for image recognition and quantum state preparation compared with existing magnetic memristor and CMOS technologies. Our results not only showcase a new application of chiral edge states but also may inspire further topological quantum physics-based novel computing schemes.

cond-mat.mes-hall↗

arXiv2025

ODIN: Star Formation Histories Reveal Formative Starbursts Experienced by Lyman Alpha Emitting Galaxies at Cosmic Noon

In this work, we test the frequent assumption that Lyman Alpha Emitting galaxies (LAEs) are experiencing their first major burst of star formation at the time of observation. To this end, we identify 74 LAEs from the ODIN Survey with rest-UV-through-NIR photometry from UVCANDELS. For each LAE, we perform non-parametric star formation history (SFH) reconstruction using the Dense Basis Gaussian process-based method of spectral energy distribution fitting. We find that a strong majority (67%) of our LAE SFHs align with the frequently assumed archetype of a first major star formation burst, with at most modest star formation rates (SFRs) in the past. However, the rest of our LAE SFHs have significant amounts of star formation in the past, with 28% exhibiting earlier bursts of star formation with the ongoing burst having the highest SFR (dominant bursts), and the final 5% having experienced their highest SFR in the past (non-dominant bursts). Combining the SFHs indicating first and dominant bursts, ~95% of LAEs are experiencing their largest burst yet -- a formative burst. We also find that the fraction of total stellar mass created in the last 200 Myr is ~1.3 times higher in LAEs than in mass-matched Lyman Break Galaxy (LBG) samples, and that a majority of LBGs are experiencing dominant bursts, reaffirming that LAEs differ from other star forming galaxies. Overall, our results suggest that multiple evolutionary paths can produce galaxies with strong observed Ly$α$ emission.

astro-ph.GA↗

arXiv2025

Text-To-Speech Synthesis In The Wild

Traditional Text-to-Speech (TTS) systems rely on studio-quality speech recorded in controlled settings.a Recently, an effort known as noisy-TTS training has emerged, aiming to utilize in-the-wild data. However, the lack of dedicated datasets has been a significant limitation. We introduce the TTS In the Wild (TITW) dataset, which is publicly available, created through a fully automated pipeline applied to the VoxCeleb1 dataset. It comprises two training sets: TITW-Hard, derived from the transcription, segmentation, and selection of raw VoxCeleb1 data, and TITW-Easy, which incorporates additional enhancement and data selection based on DNSMOS. State-of-the-art TTS models achieve over 3.0 UTMOS score with TITW-Easy, while TITW-Hard remains difficult showing UTMOS below 2.8.

eess.AS↗

arXiv2025

Heterodyne detection of low-frequency fields via Rydberg EIT with phase demodulation

Recently, the rapid progress of quantum sensing research reveals that the Rydberg atoms have great potentials in becoming high-precision centimeter-scale antenna of low-frequency fields. In order to facilitate efficient and reliable detection of low-frequency fields via Rydberg atoms, we design, implement and analyze a special but low-cost and scalable method based on heterodyning processes under the condition of electromagnetically induced transparency (EIT) embedded in typical two-photon ground-Rydberg transition. Instead of relying on observing changes in absorption of light by Rydberg atoms, our method focuses on the phase modulation effect on the probe laser induced by the low-frequency fields via the Rydberg EIT mechanism and utilizes a demodulation process to accurately retrieve the signal. The general principles of our method apply to both electric and magnetic fields and it is even possible to realize the combination of both functionalities in the same apparatus. In particular, we experimentally demonstrate the full cycle of operations with respect to both cases. In the measurement of low-frequency electric fields, we discover that the Rydberg dipole-dipole interaction among atoms induce linear superposition of Rydberg states with different angular momentum that generates a first-order response corresponding to the signature of linear Stark effect. As the Rydberg atoms have excellent coupling strengths with electric fields, our results indicate that our method can hopefully reach high-precision performance for practical tasks in the future.

physics.atom-ph↗