arXiv subjects

Yong Zhang

Publications and source records attributed to Yong Zhang.

At least 127 records · Page 7Linked to original sources

arXiv2025

EmoCAST: Emotional Talking Portrait via Emotive Text Description

Emotional talking head synthesis aims to generate talking portrait videos with vivid expressions. Existing methods still exhibit limitations in control flexibility, motion naturalness, and expression quality. Moreover, currently available datasets are mainly collected in lab settings, further exacerbating these shortcomings and hindering real-world deployment. To address these challenges, we propose EmoCAST, a diffusion-based talking head framework for precise, text-driven emotional synthesis. Its contributions are threefold: (1) architectural modules that enable effective text control; (2) an emotional talking-head dataset that expands the framework's ability; and (3) training strategies that further improve performance. Specifically, for appearance modeling, emotional prompts are integrated through a text-guided emotive attention module, enhancing spatial knowledge to improve emotion understanding. To strengthen audio-emotion alignment, we introduce an emotive audio attention module to capture the interplay between controlled emotion and driving audio, generating emotion-aware features to guide precise facial motion synthesis. Additionally, we construct a large-scale, in-the-wild emotional talking head dataset with emotive text descriptions to optimize the framework's performance. Based on this dataset, we propose an emotion-aware sampling strategy and a progressive functional training strategy that improve the model's ability to capture nuanced expressive features and achieve accurate lip-sync. Overall, EmoCAST achieves state-of-the-art performance in generating realistic, emotionally expressive, and audio-synchronized talking-head videos. Project Page: https://github.com/GVCLab/EmoCAST

cs.CV↗

arXiv2025

DecoKAN: Interpretable Decomposition for Forecasting Cryptocurrency Market Dynamics

Accurate and interpretable forecasting of multivariate time series is crucial for understanding the complex dynamics of cryptocurrency markets in digital asset systems. Advanced deep learning methodologies, particularly Transformer-based and MLP-based architectures, have achieved competitive predictive performance in cryptocurrency forecasting tasks. However, cryptocurrency data is inherently composed of long-term socio-economic trends and local high-frequency speculative oscillations. Existing deep learning-based 'black-box' models fail to effectively decouple these composite dynamics or provide the interpretability needed for trustworthy financial decision-making. To overcome these limitations, we propose DecoKAN, an interpretable forecasting framework that integrates multi-level Discrete Wavelet Transform (DWT) for decoupling and hierarchical signal decomposition with Kolmogorov-Arnold Network (KAN) mixers for transparent and interpretable nonlinear modeling. The DWT component decomposes complex cryptocurrency time series into distinct frequency components, enabling frequency-specific analysis, while KAN mixers provide intrinsically interpretable spline-based mappings within each decomposed subseries. Furthermore, interpretability is enhanced through a symbolic analysis pipeline involving sparsification, pruning, and symbolization, which produces concise analytical expressions offering symbolic representations of the learned patterns. Extensive experiments demonstrate that DecoKAN achieves the lowest average Mean Squared Error on all tested real-world cryptocurrency datasets (BTC, ETH, XMR), consistently outperforming a comprehensive suite of competitive state-of-the-art baselines. These results validate DecoKAN's potential to bridge the gap between predictive accuracy and model transparency, advancing trustworthy decision support within complex cryptocurrency markets.

cs.LG↗

arXiv2025

Active Intelligence in Video Avatars via Closed-loop World Modeling

Current video avatar generation methods excel at identity preservation and motion alignment but lack genuine agency, they cannot autonomously pursue long-term goals through adaptive environmental interaction. We address this by introducing L-IVA (Long-horizon Interactive Visual Avatar), a task and benchmark for evaluating goal-directed planning in stochastic generative environments, and ORCA (Online Reasoning and Cognitive Architecture), the first framework enabling active intelligence in video avatars. ORCA embodies Internal World Model (IWM) capabilities through two key innovations: (1) a closed-loop OTAR cycle (Observe-Think-Act-Reflect) that maintains robust state tracking under generative uncertainty by continuously verifying predicted outcomes against actual generations, and (2) a hierarchical dual-system architecture where System 2 performs strategic reasoning with state prediction while System 1 translates abstract plans into precise, model-specific action captions. By formulating avatar control as a POMDP and implementing continuous belief updating with outcome verification, ORCA enables autonomous multi-step task completion in open-domain scenarios. Extensive experiments demonstrate that ORCA significantly outperforms open-loop and non-reflective baselines in task success rate and behavioral coherence, validating our IWM-inspired design for advancing video avatar intelligence from passive animation to active, goal-oriented behavior.

cs.CV↗

arXiv2025

Infrared SED Modeling of Velocity-Excess Maser Sources: Identifying Incipient Water-Fountain Candidates

We investigated whether "velocity excess" in circumstellar maser lines can diagnose the earliest evolutionary phases of Water Fountains (WFs). Here we define "velocity excess" as maser emission (e.g., H$_2$O 22.235 GHz or OH 1665/1667 MHz) detected at velocities outside the velocity range of the OH 1612 MHz line, which traces the terminal expansion velocity of a spherical circumstellar envelope (CSE). Such excess velocities serve as an indicator of gas motions deviating from spherical expansion and may signify the onset of asymmetric outflows. Based on recent studies (Fan et al.2024; Xie et al.2025), we analyzed 17 sources showing velocity excess and fitted their infrared spectral energy distributions (SEDs) with the one-dimensional radiative transfer code DUSTY. Seven sources are well reproduced, implying outer CSEs that remain nearly spherical despite inner asymmetries. Among these, five exhibit single-peaked, AGB-like SEDs and two show double-peaked, post-AGB-like profiles. IRAS variability indices and NEOWISE-R light curves reveal pulsations (~600-1000 days) in three sources, supporting their AGB classification. Considering the magnitude of the velocity excess, two objects-IRAS 19229+1708 and IRAS 19052+0922-may represent the earliest or incipient WF phase, in which asymmetric outflows are beginning to emerge within otherwise spherical envelopes. These results support a morphological sequence in which bipolar jets and tori arise first in the central regions while the outer CSE remains spherical, and they show that selecting WF candidates via velocity excess effectively identifies objects at the onset of jet formation and early morphological transformation.

astro-ph.SR↗

arXiv2025

MEPIC: Memory Efficient Position Independent Caching for LLM Serving

Modern LLM applications such as deep-research assistants, coding agents, and Retrieval-Augmented Generation (RAG) systems, repeatedly process long prompt histories containing shared document or code chunks, creating significant pressure on the Key Value (KV) cache, which must operate within limited memory while sustaining high throughput and low latency. Prefix caching partially alleviates some of these costs by reusing KV cache for previously processed tokens, but limited by strict prefix matching. Position-independent caching (PIC) enables chunk-level reuse at arbitrary positions, but requires selective recomputation and positional-encoding (PE) adjustments. However, because these operations vary across queries, KV for the same chunk diverges across requests. Moreover, without page alignment, chunk KV layouts diverge in memory, preventing page sharing. These issues result in only modest HBM savings even when many requests reuse the same content. We present MEPIC, a memory-efficient PIC system that enables chunk KV reuse across positions, requests, and batches. MEPIC aligns chunk KV to paged storage, shifts recomputation from token- to block-level so only the first block is request-specific, removes positional encodings via Rotary Position Embedding (RoPE) fusion in the attention kernel, and makes remaining blocks fully shareable. These techniques eliminate most duplicate chunk KV in HBM, reducing usage by up to 2x over state-of-the-art PIC at comparable latency and accuracy, and up to 5x for long prompts, without any model changes.

cs.LG↗

arXiv2025

CrossTrafficLLM: A Human-Centric Framework for Interpretable Traffic Intelligence via Large Language Model

While accurate traffic forecasting is vital for Intelligent Transportation Systems (ITS), effectively communicating predicted conditions via natural language for human-centric decision support remains a challenge and is often handled separately. To address this, we propose CrossTrafficLLM, a novel GenAI-driven framework that simultaneously predicts future spatiotemporal traffic states and generates corresponding natural language descriptions, specifically targeting conditional abnormal event summaries. We tackle the core challenge of aligning quantitative traffic data with qualitative textual semantics by leveraging Large Language Models (LLMs) within a unified architecture. This design allows generative textual context to improve prediction accuracy while ensuring generated reports are directly informed by the forecast. Technically, a text-guided adaptive graph convolutional network is employed to effectively merge high-level semantic information with the traffic network structure. Evaluated on the BJTT dataset, CrossTrafficLLM demonstrably surpasses state-of-the-art methods in both traffic forecasting performance and text generation quality. By unifying prediction and description generation, CrossTrafficLLM delivers a more interpretable, and actionable approach to generative traffic intelligence, offering significant advantages for modern ITS applications.

cs.LG↗

arXiv2025

Feeding a Kerr black hole with quantized vortices

By solving a nonlinear Klein-Gordon equation in Kerr geometry, we uncover new phenomena and key characteristics of quantized vortices in quantum fluids near a Kerr black hole. The formation of these vortices induces rotational or turbulent flows, which profoundly alter the fluid properties and revise those dark matter models describing axion condensates, ultralight boson clouds, and other scalar fields in the vicinity of spinning black holes. As macroscopic, quantum, and topological defects, these vortices can stably orbit the black hole over extended periods, establishing their viability as novel probes for investigating black hole physics. For instance, we calculate the angular velocities of orbiting vortices to quantitatively characterize the frame-dragging effect, a classic prediction of general relativity. Additionally, we observe that relatively large vortices are accreted onto the black hole, wrapping around it while undergoing splitting and reconnecting processes. In quantum fluids with high vortex densities, turbulent flows emerge, accompanied by the formation of a vortex boundary layer near the event horizon. Beyond the ergosphere, we find vortex emissions and energetic outbursts, which may provide crucial insights into analogous astrophysical events recently discovered by the XRISM satellite.

gr-qc↗

arXiv2025

UniMo: Unifying 2D Video and 3D Human Motion with an Autoregressive Framework

We propose UniMo, an innovative autoregressive model for joint modeling of 2D human videos and 3D human motions within a unified framework, enabling simultaneous generation and understanding of these two modalities for the first time. Current methods predominantly focus on generating one modality given another as the condition or integrating either of them with other modalities such as text and audio. Unifying 2D videos and 3D motions for simultaneous optimization and generation remains largely unexplored, presenting significant challenges due to their substantial structural and distributional differences. Inspired by the LLM's ability to unify different modalities, our method models videos and 3D motions as a unified tokens sequence, utilizing separate embedding layers to mitigate distribution gaps. Additionally, we devise a sequence modeling strategy that integrates two distinct tasks within a single framework, proving the effectiveness of unified modeling. Moreover, to efficiently align with visual tokens and preserve 3D spatial information, we design a novel 3D motion tokenizer with a temporal expansion strategy, using a single VQ-VAE to produce quantized motion tokens. It features multiple expert decoders that handle body shapes, translation, global orientation, and body poses for reliable 3D motion reconstruction. Extensive experiments demonstrate that our method simultaneously generates corresponding videos and motions while performing accurate motion capture. This work taps into the capacity of LLMs to fuse diverse data types, paving the way for integrating human-centric information into existing models and potentially enabling multimodal, controllable joint modeling of humans, objects, and scenes.

cs.CV↗

arXiv2025

New Factorizations of Yang-Mills Amplitudes

We propose a new factorization pattern for tree-level Yang-Mills (YM) amplitudes, where they decompose into a sum of gluings of two lower-point amplitudes by setting specific two-point non-planar Mandelstam variables within a rectangular configuration to zero. This approach manifests the hidden zeros of YM amplitudes recently identified. Furthermore, by setting specific Lorentz products involving polarization vectors to zero, the amplitudes further reduce to a sum of products of three currents. These novel factorizations provide a fresh perspective on the structure of YM amplitudes, potentially enhancing our understanding and calculation of these fundamental quantities.

hep-th↗

arXiv2025

Microservices Are Dying, A New Method for Module Division Based on Universal Interfaces

Although microservices have physically isolated modules, they have failed to prevent the propagation and diffusion of dependencies. To trace the root cause of the inter-module coupling, this paper, starting from the impact assessment approach for module changes, proposes a conceptual method for calculating module independence and utilizes this method to derive the necessary conditions for module independence. Then, a new system design philosophy and software engineering methodology is proposed, aimed at eliminating dependencies between modules. A specific pattern is employed to design a set of universal interfaces, serving as a universal boundary between modules. Subsequently, this method is used to implement a platform architecture named EIGHT, demonstrating that, as long as module independence is guaranteed, even a monolithic application within a single process can dynamically load, unload, or modify any part at runtime. Finally, the paper concludes that this architecture aims to explore a novel path for increasingly complex systems, beyond microservice and monolithic architectures.

cs.SE↗

arXiv2025

Ultrafast Reconfigurable Topological Photonic Processing Accelerator

The rise of artificial intelligence has triggered exponential growth in data volume, demanding rapid and efficient processing. High-speed, energy-efficient, and parallel-scalable computing hardware is thus increasingly critical. We demonstrate a wafer-scale non-volatile topological photonic computing chip using topological modulators. Leveraging the GHz-speed electro-optic response and nonvolatility of ferroelectric lead zirconate titanate (PZT) thin films via topological photonic confinement, Our chip enables thousand-fold faster reconfiguration, zero-static-power operation, and a computational density of 266 trillion operations per second per square millimeter . This density surpasses that of silicon photonic reconfigurable computing chips by two orders of magnitude and thin-film lithium niobate platforms by four orders of magnitude. A 16-channel wavelength-space multiplexed chip delivers 1.92 TOPS throughput with 95.64% digit-recognition accuracy and 94.5% precision for solving time-varying partial differential equations. Additionally, the chip supports functional reconfiguration for high bandwidth density optical I/O. This work establishes ferroelectric topological photonics for efficient high-speed photonic tensor processing.

physics.optics↗

arXiv2025

The ALMA-QUARKS survey: Hot Molecular Cores are a long-standing phenomenon in the evolution of massive protostars

We present an analysis of the QUARKS survey sample, focusing on protoclusters where Hot Molecular Cores (HMCs, traced by CH3CN(12--11)) and UC HII regions (traced by H30α/H40α) coexist. Using the high-resolution, high-sensitivity 1.3 mm data from the QUARKS survey, we identify 125 Hot Molecular Fragments (HMFs), which represent the substructures of HMCs at higher resolution. From line integrated intensity maps of CH3CN(12--11) and H30α, we resolve the spatial distribution of HMFs and UC HII regions. By combining with observations of CO outflows and 1.3 mm continuum, we classify HMFs into four types: HMFs associated with jet-like outflow, with wide-angle outflow, with non-detectable outflow, and shell-like HMFs near UC HII regions. This diversity possibly indicates that the hot core could be polymorphic and long-standing phenomenon in the evolution of massive protostars. The separation between HMFs and H30α/H40αemission suggests that sequential high-mass star formation within young protoclusters is not likely related to feedback mechanisms.

astro-ph.GA↗

arXiv2025

HRM^2Avatar: High-Fidelity Real-Time Mobile Avatars from Monocular Phone Scans

We present HRM$^2$Avatar, a framework for creating high-fidelity avatars from monocular phone scans, which can be rendered and animated in real time on mobile devices. Monocular capture with smartphones provides a low-cost alternative to studio-grade multi-camera rigs, making avatar digitization accessible to non-expert users. Reconstructing high-fidelity avatars from single-view video sequences poses challenges due to limited visual and geometric data. To address these limitations, at the data level, our method leverages two types of data captured with smartphones: static pose sequences for texture reconstruction and dynamic motion sequences for learning pose-dependent deformations and lighting changes. At the representation level, we employ a lightweight yet expressive representation to reconstruct high-fidelity digital humans from sparse monocular data. We extract garment meshes from monocular data to model clothing deformations effectively, and attach illumination-aware Gaussians to the mesh surface, enabling high-fidelity rendering and capturing pose-dependent lighting. This representation efficiently learns high-resolution and dynamic information from monocular data, enabling the creation of detailed avatars. At the rendering level, real-time performance is critical for animating high-fidelity avatars in AR/VR, social gaming, and on-device creation. Our GPU-driven rendering pipeline delivers 120 FPS on mobile devices and 90 FPS on standalone VR devices at 2K resolution, over $2.7\times$ faster than representative mobile-engine baselines. Experiments show that HRM$^2$Avatar delivers superior visual realism and real-time interactivity, outperforming state-of-the-art monocular methods.

cs.GR↗

arXiv2025

Parallel BiLSTM-Transformer networks for forecasting chaotic dynamics

The nonlinear nature of chaotic systems results in extreme sensitivity to initial conditions and highly intricate dynamical behaviors, posing fundamental challenges for accurately predicting their evolution. To overcome the limitation that conventional approaches fail to capture both local features and global dependencies in chaotic time series simultaneously, this study proposes a parallel predictive framework integrating Transformer and Bidirectional Long Short-Term Memory (BiLSTM) networks. The hybrid model employs a dual-branch architecture, where the Transformer branch mainly captures long-range dependencies while the BiLSTM branch focuses on extracting local temporal features. The complementary representations from the two branches are fused in a dedicated feature-fusion layer to enhance predictive accuracy. As illustrating examples, the model's performance is systematically evaluated on two representative tasks in the Lorenz system. The first is autonomous evolution prediction, in which the model recursively extrapolates system trajectories from the time-delay embeddings of the state vector to evaluate long-term tracking accuracy and stability. The second is inference of unmeasured variable, where the model reconstructs the unobserved states from the time-delay embeddings of partial observations to assess its state-completion capability. The results consistently indicate that the proposed hybrid framework outperforms both single-branch architectures across tasks, demonstrating its robustness and effectiveness in chaotic system prediction.

cs.LG↗

arXiv2025

UREM: A High-performance Unified and Resilient Enhancement Method for Multi- and High-Dimensional Indexes

Numerous multi- or high-dimensional indexes with distinct advantages have been proposed on various platforms to meet application requirements. To achieve higher-performance queries, most indexes employ enhancement methods, including structure-oriented and layout-oriented enhancement methods. Existing structure-oriented methods tailored to specific indexes work well under static workloads but lack generality and degrade under dynamic workloads. The layout-oriented methods exhibit good generality and perform well under dynamic workloads, but exhibit suboptimal performance under static workloads. Therefore, it is an open challenge to develop a unified and resilient enhancement method that can improve query performance for different indexes adaptively under different scenarios. In this paper, we propose UREM, which is the first high-performance Unified and Resilient Enhancement Method designed for both multi- and high-dimensional indexes, capable of adapting to different scenarios. Specifically, UREM (1) can be uniformly applied with different indexes on various platforms; (2) enhances the query performance of indexes by layout optimization under static workloads; (3) enables indexes to stabilize performance when queries shift through partial layout reorganization. We evaluate UREM on 20 widely used indexes. Experimental results demonstrate that UREM improves the query performance of multi- and high-dimensional indexes by up to 5.73x and 9.18x under static workloads, and by an average of 5.72x and 9.47x under dynamic workloads. Moreover, some traditional indexes enhanced by UREM even achieve performance comparable to or even surpassing that of recent advanced indexes.

cs.DB↗

arXiv2025

Censorship Chokepoints: New Battlegrounds for Regional Surveillance, Censorship and Influence on the Internet

Undoubtedly, the Internet has become one of the most important conduits to information for the general public. Nonetheless, Internet access can be and has been limited systematically or blocked completely during political events in numerous countries and regions by various censorship mechanisms. Depending on where the core filtering component is situated, censorship techniques have been classified as client-based, server-based, or network-based. However, as the Internet evolves rapidly, new and sophisticated censorship techniques have emerged, which involve techniques that cut across locations and involve new forms of hurdles to information access. We argue that modern censorship can be better understood through a new lens that we term chokepoints, which identifies bottlenecks in the content production or delivery cycle where efficient new forms of large-scale client-side surveillance and filtering mechanisms have emerged.

cs.CR↗

arXiv2025

Analytic 3D vector non-uniform Fourier crystal optics in arbitrary $\bar{\bar{\varepsilon}}$ dielectric

To find a suitable framework for nonlinear crystal optics(NCO), we have revisited linear crystal optics(LCO). At the methodological level, three widely used plane wave bases are compared in terms of eigenanalysis in reciprocal space and light field propagation in real space. Inspired by complex ray tracing, we expand M.V. Berry \& M.R. Dennis's 2003 uniform plane wave model to non-uniform Fourier crystal optics and ultimately derive the explicit form of its 3$\times$2 transition matrix, bridging the two major branches of crystal optics in reciprocal space, where either ray direction $\hat{k}$ or spatial frequency $\bar{k}_{\mathrmρ}$ serves as the input variable. Using this model, we create the material-matrix tetrahedral compass to conduct a detailed analysis of how the four fundamental characteristics of materials (linear/circular birefringence/dichroism) influence the eigensystems of the vector electric field in two-dimensional spatial frequency $\bar{k}_{\mathrmρ}$ domain and its distribution in three-dimensional $\bar{r}$ space with a crystal-2f configuration. Along this journey, we have uncovered new territories in LCO in both real and reciprocal space, such as infinite singularities arranged in disk-, ring-, and crescent-like shapes, ``L shorelines'' resembling hearts, generalized haunting theorem, double conical refraction, and optical knots it induces. We also present our model's early applications in focal engineering and NCO. As the opening chapter in a trilogy, this work connects crystal optics, Fourier optics, and nonlinear optics, while integrating theoretical, computational, and experimental physics, advancing all six domains.

physics.optics↗

arXiv2025

The Hybrid Multimodal Graph Index (HMGI): A Comprehensive Framework for Integrated Relational and Vector Search

The proliferation of complex, multimodal datasets has exposed a critical gap between the capabilities of specialized vector databases and traditional graph databases. While vector databases excel at semantic similarity search, they lack the capacity for deep relational querying. Conversely, graph databases master complex traversals but are not natively optimized for high-dimensional vector search. This paper introduces the Hybrid Multimodal Graph Index (HMGI), a novel framework designed to bridge this gap by creating a unified system for efficient, hybrid queries on multimodal data. HMGI leverages the native graph database architecture and integrated vector search capabilities, exemplified by platforms like Neo4j, to combine Approximate Nearest Neighbor Search (ANNS) with expressive graph traversal queries. Key innovations of the HMGI framework include modality-aware partitioning of embeddings to optimize index structure and query performance, and a system for adaptive, low-overhead index updates to support dynamic data ingestion, drawing inspiration from the architectural principles of systems like TigerVector. By integrating semantic similarity search directly with relational context, HMGI aims to outperform pure vector databases like Milvus in complex, relationship-heavy query scenarios and achieve sub-linear query times for hybrid tasks.

cs.DB↗