arXiv subjects

Liang Dong

Publications and source records attributed to Liang Dong.

At least 37 records · Page 2Linked to original sources

arXiv2026

ELUTQ: Optimizing Quantization Accuracy under LUT-Based Computation for Edge LLMs

Weight quantization effectively reduces memory consumption and enable the deployment of Large Language Models on edge devices, yet existing hardware-friendly methods often rely on uniform quantization, which suffers from poor weight-distribution fitting and high dequantization overhead under low-bit settings. In this paper, we propose ELUTQ, an efficient quantization framework featuring a novel quantization format termed Hierarchical Linear Quantization (HLQ). HLQ is designed to better capture the statistical characteristics of weights and eliminate dequantization overhead using Bit-serial LUT-based GEMM operations. HLQ significantly improves model accuracy under low-bit settings and achieves performance comparable to QAT methods without any retraining of the weights. Moreover, an optimized quantization pipeline is integrated into ELUTQ, enabling it to complete the quantization of LLaMA 3.1-70B using only 64 GB of CPU memory and 48 GB of VRAM, reducing the hardware requirements for large-scale model quantization. To enable efficient deployment on edge devices, ELUTQ designs high-performance kernels to support end-to-end inference. Our 2-bit LLaMA3.1-8B achieves 1.5x speedup over AWQ on RTX 3090. Code is available at https://github.com/Nkniexin/ELUTQ.

cs.LG↗

arXiv2026

MemFine: Memory-Aware Fine-Grained Scheduling for MoE Training

The training of large-scale Mixture of Experts (MoE) models faces a critical memory bottleneck due to severe load imbalance caused by dynamic token routing. This imbalance leads to memory overflow on GPUs with limited capacity, constraining model scalability. Existing load balancing methods, which cap expert capacity, compromise model accuracy and fail on memory-constrained hardware. To address this, we propose MemFine, a memory-aware fine-grained scheduling framework for MoE training. MemFine decomposes the token distribution and expert computation into manageable chunks and employs a chunked recomputation strategy, dynamically optimized through a theoretical memory model to balance memory efficiency and throughput. Experiments demonstrate that MemFine reduces activation memory by 48.03% and improves throughput by 4.42% compared to full recomputation-based baselines, enabling stable large-scale MoE training on memory-limited GPUs.

cs.DC↗

arXiv2025

HY-Motion 1.0: Scaling Flow Matching Models for Text-To-Motion Generation

We present HY-Motion 1.0, a series of state-of-the-art, large-scale, motion generation models capable of generating 3D human motions from textual descriptions. HY-Motion 1.0 represents the first successful attempt to scale up Diffusion Transformer (DiT)-based flow matching models to the billion-parameter scale within the motion generation domain, delivering instruction-following capabilities that significantly outperform current open-source benchmarks. Uniquely, we introduce a comprehensive, full-stage training paradigm -- including large-scale pretraining on over 3,000 hours of motion data, high-quality fine-tuning on 400 hours of curated data, and reinforcement learning from both human feedback and reward models -- to ensure precise alignment with the text instruction and high motion quality. This framework is supported by our meticulous data processing pipeline, which performs rigorous motion cleaning and captioning. Consequently, our model achieves the most extensive coverage, spanning over 200 motion categories across 6 major classes. We release HY-Motion 1.0 to the open-source community to foster future research and accelerate the transition of 3D human motion generation models towards commercial maturity.

cs.CV↗

arXiv2025

MoFa: A Unified Performance Modeling Framework for LLM Pretraining

The exponential growth in LLM scales, with parameters soaring from billions to trillions, has necessitated distributed pretraining across large clusters comprising thousands to tens of thousands of devices. While hybrid parallelization strategies enable such pretraining, the vast combinatorial strategy space introduces significant optimization challenges. Traditional manual tuning methods incur prohibitive trial-and-error costs, and existing performance modeling approaches exhibit critical limitations: they fail to comprehensively account for prevalent optimization features and ignore the substantial overhead imposed by essential fault tolerance mechanisms like checkpoint recovery in long-duration pretraining. To address these gaps, we propose MoFa, a novel pretraining performance modeling framework that unifies multi-dimensional optimization features and fault tolerance. MoFa incorporates an enhanced cost model to accurately capture the effects of key optimizations and integrates a fault tolerance model based on historical cluster reliability data. Besides, a MoFa-based tuning system is developed to explore optimal pretraining performance and potential bottlenecks in various scenarios. Extensive modeling evaluations demonstrate that MoFa can achieve high prediction accuracy across various scenarios. In addition, through comprehensive tuning experiments, our framework systematically reveals the key factors influencing pretraining performance under different configurations, which provides solid a priori guidance for LLM pretraining system design and deployment.

cs.DC↗

arXiv2025

A central limit theorem for unbalanced step-reinforced random walks

In this paper, we study a class of unbalanced step-reinforced random walks that unifies the elephant random walk, the positively step-reinforced random walk, and the negatively step-reinforced random walk. By establishing a connection with bond percolation on random recursive trees, these processes can be represented as randomly weighted sums of independent and identically distributed random variables. We first derive normal and stable central limit theorems for such randomly weighted sums, and then apply these results to obtain a unified central limit theorem for unbalanced step-reinforced random walks.

math.PR↗

arXiv2025

Hunyuan3D Studio: End-to-End AI Pipeline for Game-Ready 3D Asset Generation

The creation of high-quality 3D assets, a cornerstone of modern game development, has long been characterized by labor-intensive and specialized workflows. This paper presents Hunyuan3D Studio, an end-to-end AI-powered content creation platform designed to revolutionize the game production pipeline by automating and streamlining the generation of game-ready 3D assets. At its core, Hunyuan3D Studio integrates a suite of advanced neural modules (such as Part-level 3D Generation, Polygon Generation, Semantic UV, etc.) into a cohesive and user-friendly system. This unified framework allows for the rapid transformation of a single concept image or textual description into a fully-realized, production-quality 3D model complete with optimized geometry and high-fidelity PBR textures. We demonstrate that assets generated by Hunyuan3D Studio are not only visually compelling but also adhere to the stringent technical requirements of contemporary game engines, significantly reducing iteration time and lowering the barrier to entry for 3D content creation. By providing a seamless bridge from creative intent to technical asset, Hunyuan3D Studio represents a significant leap forward for AI-assisted workflows in game development and interactive media.

cs.CV↗

arXiv2025

K-Syn: K-space Data Synthesis in Ultra Low-data Regimes

Owing to the inherently dynamic and complex characteristics of cardiac magnetic resonance (CMR) imaging, high-quality and diverse k-space data are rarely available in practice, which in turn hampers robust reconstruction of dynamic cardiac MRI. To address this challenge, we perform feature-level learning directly in the frequency domain and employ a temporal-fusion strategy as the generative guidance to synthesize k-space data. Specifically, leveraging the global representation capacity of the Fourier transform, the frequency domain can be considered a natural global feature space. Therefore, unlike traditional methods that use pixel-level convolution for feature learning and modeling in the image domain, this letter focuses on feature-level modeling in the frequency domain, enabling stable and rich generation even with ultra low-data regimes. Moreover, leveraging the advantages of feature-level modeling in the frequency domain, we integrate k-space data across time frames with multiple fusion strategies to steer and further optimize the generative trajectory. Experimental results demonstrate that the proposed method possesses strong generative ability in low-data regimes, indicating practical potential to alleviate data scarcity in dynamic MRI reconstruction.

cs.CV↗

arXiv2025

QuarkMed Medical Foundation Model Technical Report

Recent advancements in large language models have significantly accelerated their adoption in healthcare applications, including AI-powered medical consultations, diagnostic report assistance, and medical search tools. However, medical tasks often demand highly specialized knowledge, professional accuracy, and customization capabilities, necessitating a robust and reliable foundation model. QuarkMed addresses these needs by leveraging curated medical data processing, medical-content Retrieval-Augmented Generation (RAG), and a large-scale, verifiable reinforcement learning pipeline to develop a high-performance medical foundation model. The model achieved 70% accuracy on the Chinese Medical Licensing Examination, demonstrating strong generalization across diverse medical benchmarks. QuarkMed offers a powerful yet versatile personal medical AI solution, already serving over millions of users at ai.quark.cn.

cs.AI↗

arXiv2025

Temporal variability in low-frequency radio interference: Insight from high-cadence monitoring at a candidate radio notification zone in Malaysia

Extensive radio frequency interference (RFI) monitoring is essential in the site selection process before constructing radio astronomy observatories, followed by mitigation strategies to minimize its adverse effects. Malaysia has an enormous prospect for radio astronomy due to its prominent location in the centre of Southeast Asia, but is challenged by its relatively high population density. In this research article, we perform high-cadence, low-frequency RFI monitoring at two sites, each representing an urban and a rural environment. Using modified generalized spectral kurtosis (GSK) as an RFI detection method, we ascertain the suitability of Glami Lemi, a rural area in the centre of Peninsular Malaysia previously assigned as a candidate radio notification zone (RNZ), as a potential site for radio astronomy observations due to its lower RFI contamination in our high-cadence monitoring, especially when compared with urban areas. We identified a number of persistent and transient RFI in our dataset, associate each of them with their potential origins and, if present, characterize their temporal evolution. A few types of RFI mitigation strategies were also tested and discussed. This study lays the groundwork for Malaysia's endeavours in establishing its first research-grade radio telescope, emphasizing the importance of robust RFI detection and mitigation strategies in optimizing observational outcomes.

astro-ph.IM↗

arXiv2025

HunyuanWorld 1.0: Generating Immersive, Explorable, and Interactive 3D Worlds from Words or Pixels

Creating immersive and playable 3D worlds from texts or images remains a fundamental challenge in computer vision and graphics. Existing world generation approaches typically fall into two categories: video-based methods that offer rich diversity but lack 3D consistency and rendering efficiency, and 3D-based methods that provide geometric consistency but struggle with limited training data and memory-inefficient representations. To address these limitations, we present HunyuanWorld 1.0, a novel framework that combines the best of both worlds for generating immersive, explorable, and interactive 3D scenes from text and image conditions. Our approach features three key advantages: 1) 360° immersive experiences via panoramic world proxies; 2) mesh export capabilities for seamless compatibility with existing computer graphics pipelines; 3) disentangled object representations for augmented interactivity. The core of our framework is a semantically layered 3D mesh representation that leverages panoramic images as 360° world proxies for semantic-aware world decomposition and reconstruction, enabling the generation of diverse 3D worlds. Extensive experiments demonstrate that our method achieves state-of-the-art performance in generating coherent, explorable, and interactive 3D worlds while enabling versatile applications in virtual reality, physical simulation, game development, and interactive content creation.

cs.CV↗

arXiv2025

Noise Reduction Method for Radio Astronomy Single Station Observation Based on Wavelet Transform and Mathematical Morphology

The 21 cm radiation of neutral hydrogen provides crucial information for studying the early universe and its evolution. To advance this research, countries have made significant investments in constructing large low-frequency radio telescope arrays, such as the Low Frequency Array (LOFAR) and the Square Kilometre Array Phase 1 Low Frequency (SKA1-low). These instruments are pivotal for radio astronomy research. However, challenges such as ionospheric plasma interference, ambient radio noise, and instrument-related effects have become increasingly prominent, posing major obstacles in cosmology research. To address these issues, this paper proposes an efficient signal processing method that combines wavelet transform and mathematical morphology. The method involves the following steps: Background Subtraction: Background interference in radio observation signals is eliminated. Wavelet Transform: The signal, after removing background noise, undergoes a two-dimensional discrete wavelet transform. Threshold processing is then applied to the wavelet coefficients to effectively remove interference components. Wavelet Inversion: The processed signal is reconstructed using wavelet inversion. Mathematical Morphology: The reconstructed signal is further optimized using mathematical morphology to refine the results. Experimental verification was conducted using solar observation data from the Xinjiang Observatory and the Yunnan Observatory. The results demonstrate that this method successfully removes interference signals while preserving useful signals, thus improving the accuracy of radio astronomy observations and reducing the impact of radio frequency interference (RFI).

astro-ph.IM↗

arXiv2025

Enhancing Vehicular Platooning with Wireless Federated Learning: A Resource-Aware Control Framework

This paper aims to enhance the performance of Vehicular Platooning (VP) systems integrated with Wireless Federated Learning (WFL). In highly dynamic environments, vehicular platoons experience frequent communication changes and resource constraints, which significantly affect information exchange and learning model synchronization. To address these challenges, we first formulate WFL in VP as a joint optimization problem that simultaneously considers Age of Information (AoI) and Federated Learning Model Drift (FLMD) to ensure timely and accurate control. Through theoretical analysis, we examine the impact of FLMD on convergence performance and develop a two-stage Resource-Aware Control framework (RACE). The first stage employs a Lagrangian dual decomposition method for resource configuration, while the second stage implements a multi-agent deep reinforcement learning approach for vehicle selection. The approach integrates Multi-Head Self-Attention and Long Short-Term Memory networks to capture spatiotemporal correlations in communication states. Experimental results demonstrate that, compared to baseline methods, the proposed framework improves AoI optimization by up to 45%, accelerates learning convergence, and adapts more effectively to dynamic VP environments on the AI4MARS dataset.

cs.NI↗

arXiv2025

Kling-Foley: Multimodal Diffusion Transformer for High-Quality Video-to-Audio Generation

We propose Kling-Foley, a large-scale multimodal Video-to-Audio generation model that synthesizes high-quality audio synchronized with video content. In Kling-Foley, we introduce multimodal diffusion transformers to model the interactions between video, audio, and text modalities, and combine it with a visual semantic representation module and an audio-visual synchronization module to enhance alignment capabilities. Specifically, these modules align video conditions with latent audio elements at the frame level, thereby improving semantic alignment and audio-visual synchronization. Together with text conditions, this integrated approach enables precise generation of video-matching sound effects. In addition, we propose a universal latent audio codec that can achieve high-quality modeling in various scenarios such as sound effects, speech, singing, and music. We employ a stereo rendering method that imbues synthesized audio with a spatial presence. At the same time, in order to make up for the incomplete types and annotations of the open-source benchmark, we also open-source an industrial-level benchmark Kling-Audio-Eval. Our experiments show that Kling-Foley trained with the flow matching objective achieves new audio-visual SOTA performance among public models in terms of distribution matching, semantic alignment, temporal alignment and audio quality.

eess.AS↗

arXiv2025

Hunyuan3D 2.1: From Images to High-Fidelity 3D Assets with Production-Ready PBR Material

3D AI-generated content (AIGC) is a passionate field that has significantly accelerated the creation of 3D models in gaming, film, and design. Despite the development of several groundbreaking models that have revolutionized 3D generation, the field remains largely accessible only to researchers, developers, and designers due to the complexities involved in collecting, processing, and training 3D models. To address these challenges, we introduce Hunyuan3D 2.1 as a case study in this tutorial. This tutorial offers a comprehensive, step-by-step guide on processing 3D data, training a 3D generative model, and evaluating its performance using Hunyuan3D 2.1, an advanced system for producing high-resolution, textured 3D assets. The system comprises two core components: the Hunyuan3D-DiT for shape generation and the Hunyuan3D-Paint for texture synthesis. We will explore the entire workflow, including data preparation, model architecture, training strategies, evaluation metrics, and deployment. By the conclusion of this tutorial, you will have the knowledge to finetune or develop a robust 3D generative model suitable for applications in gaming, virtual reality, and industrial design.

cs.CV↗

arXiv2025

Geodynamics and artificial gravity in space-time crystals under slow perturbation and deformation

We present a theory of geodynamics in a space-time crystal based on an event wave packet constructed from the Floquet-Bloch waves, which not only involve a scalar dispersion function but also a Berry curvature tensor in the phase space manifold of space-time and the reciprocal quasi energy-momentum. In the presence of structural deformation, this theory is naturally extended into a covariant form with the introduction of a lattice connection constructed out of the gradients of the local lattice vectors. The geodesic equation for a particle not only involves the lattice connection but also higher-order corrections from space-time inhomogeneity of Berry curvatures and quasi energy-momentum dispersion gradients. These emergent connections and metric terms in the geodesic equations indicate the potential to experimentally realize artificial gravitational effects, thereby establishing a direct conceptual link between general relativity and quantum theory.

cond-mat.mtrl-sci↗

arXiv2025

Neural Pruning for 3D Scene Reconstruction: Efficient NeRF Acceleration

Neural Radiance Fields (NeRF) have become a popular 3D reconstruction approach in recent years. While they produce high-quality results, they also demand lengthy training times, often spanning days. This paper studies neural pruning as a strategy to address these concerns. We compare pruning approaches, including uniform sampling, importance-based methods, and coreset-based techniques, to reduce the model size and speed up training. Our findings show that coreset-driven pruning can achieve a 50% reduction in model size and a 35% speedup in training, with only a slight decrease in accuracy. These results suggest that pruning can be an effective method for improving the efficiency of NeRF models in resource-limited settings.

cs.CV↗

arXiv2025

GKAN: Explainable Diagnosis of Alzheimer's Disease Using Graph Neural Network with Kolmogorov-Arnold Networks

Alzheimer's Disease (AD) is a progressive neurodegenerative disorder that poses significant diagnostic challenges due to its complex etiology. Graph Convolutional Networks (GCNs) have shown promise in modeling brain connectivity for AD diagnosis, yet their reliance on linear transformations limits their ability to capture intricate nonlinear patterns in neuroimaging data. To address this, we propose GCN-KAN, a novel single-modal framework that integrates Kolmogorov-Arnold Networks (KAN) into GCNs to enhance both diagnostic accuracy and interpretability. Leveraging structural MRI data, our model employs learnable spline-based transformations to better represent brain region interactions. Evaluated on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, GCN-KAN outperforms traditional GCNs by 4-8% in classification accuracy while providing interpretable insights into key brain regions associated with AD. This approach offers a robust and explainable tool for early AD diagnosis.

cs.CV↗

arXiv2025

Subtraction of the $t\bar{t}$ contribution in $tW\bar{b}$ production at the one-loop level

The $tW\bar{b}$ production contributes to the real corrections to the $tW$ cross section. It would interfere with the top quark pair production, causing difficulties in a clear definition of the $tW{\bar b}$ events. The subtraction of the $t\bar{t}$ contributions has been performed in the diagram removal or diagram subtraction schemes for the tree-level processes. However, these schemes rely on the ability to identify the double resonant diagrams and thus can not be extended to loop diagrams. We propose a new scheme to subtract the $t\bar{t}$ contributions by power expansion of the squared amplitude in the resonant region. In order to cancel the infra-red divergences of the loop amplitudes, a widely used method is to introduce the dipole counter-terms, an ingredient in calculations of the full next-to-leading order QCD corrections. In our scheme, these counter-terms are also power-expanded. As a proof of principle, we calculate the one-loop correction to the $d\bar{d}\to \bar{b}Wt$ process, and present the invariant mass distribution of the $W\bar{b}$ system.

hep-ph↗