arXiv subjects

Xiang Liu

Publications and source records attributed to Xiang Liu.

At least 145 records · Page 8Linked to original sources

arXiv2025

Hypernetworks for Model-Heterogeneous Personalized Federated Learning

Recent advances in personalized federated learning have focused on addressing client model heterogeneity. However, most existing methods still require external data, rely on model decoupling, or adopt partial learning strategies, which can limit their practicality and scalability. In this paper, we revisit hypernetwork-based methods and leverage their strong generalization capabilities to design a simple yet effective framework for heterogeneous personalized federated learning. Specifically, we propose MH-pFedHN, which leverages a server-side hypernetwork that takes client-specific embedding vectors as input and outputs personalized parameters tailored to each client's heterogeneous model. To promote knowledge sharing and reduce computation, we introduce a multi-head structure within the hypernetwork, allowing clients with similar model sizes to share heads. Furthermore, we further propose MH-pFedHNGD, which integrates an optional lightweight global model to improve generalization. Our framework does not rely on external datasets and does not require disclosure of client model architectures, thereby offering enhanced privacy and flexibility. Extensive experiments on multiple benchmarks and model settings demonstrate that our approach achieves competitive accuracy, strong generalization, and serves as a robust baseline for future research in model-heterogeneous personalized federated learning.

cs.LG↗

arXiv2025

Investigating hybrid mesons with $0^{+-}$ and $2^{+-}$ exotic quantum numbers

Hybrid mesons occupy a unique position in the hadronic spectrum, offering a valuable insight into the non-perturbative nature of the glue field. Hybrid states with exotic $J^{PC}$ quantum numbers, such as the $1^{-+}$, $0^{+-}$ and $2^{+-}$, cannot mix with conventional $q\bar{q}$ mesons, making them critical for establishing the full spectrum of hybrid mesons. In this work, we investigate the masses and the strong decays of $0^{+-}$ and $2^{+-}$ hybrid mesons, which could be regarded as the first orbitally excited states of the $π_1(1600)$ and $η_1(1855)$ mesons. We develop a constituent gluon model and apply it to study the mass spectra and decay modes of these exotic states. Our results suggest that experiment may searched for these $0^{+-}$ and $2^{+-}$ hybrid mesons in the mass region between 2.0 GeV and 2.4 GeV. Additionally, we identify some key decay channels which should be targeted in future experiments.

hep-ph↗

arXiv2025

Building State Machine Replication Using Practical Network Synchrony

Distributed systems, such as state machine replication, are critical infrastructures for modern applications. Practical distributed protocols make minimum assumptions about the underlying network: They typically assume a partially synchronous or fully asynchronous network model. In this work, we argue that modern data center systems can be designed to provide strong synchrony properties in the common case, where servers move in synchronous lock-step rounds. We prove this hypothesis by engineering a practical design that uses a combination of kernel-bypass network, multithreaded architecture, and loosened round length, achieving a tight round bound under 2us. Leveraging our engineered networks with strong synchrony, we co-design a new replication protocol, Chora. Chora exploits the network synchrony property to efficiently pipeline multiple replication instances, while allowing all replicas to propose in parallel without extra coordination. Through experiments, we show that Chora achieves 255% and 109% improvement in throughput over state-of-the-art single-leader and multi-leader protocols, respectively.

cs.DC↗

arXiv2025

City-VLM: Towards Multidomain Perception Scene Understanding via Multimodal Incomplete Learning

Scene understanding enables intelligent agents to interpret and comprehend their environment. While existing large vision-language models (LVLMs) for scene understanding have primarily focused on indoor household tasks, they face two significant limitations when applied to outdoor large-scale scene understanding. First, outdoor scenarios typically encompass larger-scale environments observed through various sensors from multiple viewpoints (e.g., bird view and terrestrial view), while existing indoor LVLMs mainly analyze single visual modalities within building-scale contexts from humanoid viewpoints. Second, existing LVLMs suffer from missing multidomain perception outdoor data and struggle to effectively integrate 2D and 3D visual information. To address the aforementioned limitations, we build the first multidomain perception outdoor scene understanding dataset, named \textbf{\underline{SVM-City}}, deriving from multi\textbf{\underline{S}}cale scenarios with multi\textbf{\underline{V}}iew and multi\textbf{\underline{M}}odal instruction tuning data. It contains $420$k images and $4, 811$M point clouds with $567$k question-answering pairs from vehicles, low-altitude drones, high-altitude aerial planes, and satellite. To effectively fuse the multimodal data in the absence of one modality, we introduce incomplete multimodal learning to model outdoor scene understanding and design the LVLM named \textbf{\underline{City-VLM}}. Multimodal fusion is realized by constructing a joint probabilistic distribution space rather than implementing directly explicit fusion operations (e.g., concatenation). Experimental results on three typical outdoor scene understanding tasks show City-VLM achieves $18.14 \%$ performance surpassing existing LVLMs in question-answering tasks averagely. Our method demonstrates pragmatic and generalization performance across multiple outdoor scenes.

cs.CV↗

arXiv2025

Topological Data Analysis and Topological Deep Learning Beyond Persistent Homology -- A Review

Topological data analysis (TDA) is a rapidly evolving field in applied mathematics and data science that leverages tools from topology to uncover robust, shape-driven insights in complex datasets. The main workhorse is persistent homology, a technique rooted in algebraic topology. Paired with topological deep learning (TDL) or topological machine learning, persistent homology has achieved tremendous success in a wide variety of applications in science, engineering, medicine, and industry. However, persistent homology has many limitations due to its high-level abstraction, insensitivity to non-topological changes, and reliance on point cloud data. This paper presents a comprehensive review of TDA and TDL beyond persistent homology. It analyzes how persistent topological Laplacians and Dirac operators provide spectral representations to capture both topological invariants and homotopic evolution. Other formulations are presented in terms of sheaf theory, Mayer topology, and interaction topology. For data on differentiable manifolds, techniques rooted in differential topology, such as persistent de Rham cohomology, persistent Hodge Laplacian, and Hodge decomposition, are reviewed. For one-dimensional (1D) curves embedded in 3-space, approaches from geometric topology are discussed, including multiscale Gauss-link integrals, persistent Jones polynomials, and persistent Khovanov homology. This paper further discusses the appropriate selection of topological tools for different input data, such as point clouds, sequential data, data on manifolds, curves embedded in 3-space, and data with additional non-geometric information. A review is also given of various topological representations, software packages, and machine learning vectorizations. Finally, this review ends with concluding remarks.

math.HO↗

arXiv2025

Double parton distributions of the proton from basis light-front quantization

Within the basis light-front quantization framework, we systematically investigate the unpolarized and longitudinally polarized double parton distributions (DPDs) of quarks inside the proton. We utilize the light-front wave functions of the proton derived in the valence sector from a Hamiltonian quantized on the light-front. The interaction terms of the Hamiltonian consist of a one-gluon exchange interaction at fixed coupling and a three-dimensional confinement potential. Our current analysis yields significant correlations of the quarks' longitudinal momenta with their transverse separation. We also demonstrate that our calculations do not support the commonly used $x-\vec{k}_\perp$ factorization of the DPDs in $x$ and $k_\perp$. Our results are qualitatively consistent with those of other phenomenological models.

hep-ph↗

arXiv2025

iLearnRobot: An Interactive Learning-Based Multi-Modal Robot with Continuous Improvement

It is crucial that robots' performance can be improved after deployment, as they are inherently likely to encounter novel scenarios never seen before. This paper presents an innovative solution: an interactive learning-based robot system powered by a Multi-modal Large Language Model(MLLM). A key feature of our system is its ability to learn from natural dialogues with non-expert users. We also propose chain of question to clarify the exact intent of the question before providing an answer and dual-modality retrieval modules to leverage these interaction events to avoid repeating same mistakes, ensuring a seamless user experience before model updates, which is in contrast to current mainstream MLLM-based robotic systems. Our system marks a novel approach in robotics by integrating interactive learning, paving the way for superior adaptability and performance in diverse environments. We demonstrate the effectiveness and improvement of our method through experiments, both quantitively and qualitatively.

cs.HC↗

arXiv2025

Yotta: A Large-Scale Trustless Data Trading Scheme for Blockchain System

Data trading is one of the key focuses of Web 3.0. However, all the current methods that rely on blockchain-based smart contracts for data exchange cannot support large-scale data trading while ensuring data security, which falls short of fulfilling the spirit of Web 3.0. Even worse, there is currently a lack of discussion on the essential properties that large-scale data trading should satisfy. In this work, we are the first to formalize the property requirements for enabling data trading in Web 3.0. Based on these requirements, we are the first to propose Yotta, a complete batch data trading scheme for blockchain, which features a data trading design that leverages our innovative cryptographic workflow with IPFS and zk-SNARK. Our simulation results demonstrate that Yotta outperforms baseline approaches up to 130 times and exhibits excellent scalability to satisfy all the properties.

cs.CR↗

arXiv2025

Integrating Generative Adversarial Networks and Convolutional Neural Networks for Enhanced Traffic Accidents Detection and Analysis

Accident detection using Closed Circuit Television (CCTV) footage is one of the most imperative features for enhancing transport safety and efficient traffic control. To this end, this research addresses the issues of supervised monitoring and data deficiency in accident detection systems by adapting excellent deep learning technologies. The motivation arises from rising statistics in the number of car accidents worldwide; this calls for innovation and the establishment of a smart, efficient and automated way of identifying accidents and calling for help to save lives. Addressing the problem of the scarcity of data, the presented framework joins Generative Adversarial Networks (GANs) for synthesizing data and Convolutional Neural Networks (CNN) for model training. Video frames for accidents and non-accidents are collected from YouTube videos, and we perform resizing, image enhancement and image normalisation pixel range adjustments. Three models are used: CNN, Fine-tuned Convolutional Neural Network (FTCNN) and Vision Transformer (VIT) worked best for detecting accidents from CCTV, obtaining an accuracy rate of 94% and 95%, while the CNN model obtained 88%. Such results show that the proposed framework suits traffic safety applications due to its high real-time accident detection capabilities and broad-scale applicability. This work lays the foundation for intelligent surveillance systems in the future for real-time traffic monitoring, smart city framework, and integration of intelligent surveillance systems into emergency management systems.

cs.CV↗

arXiv2025

Combining Self-attention and Dilation Convolutional for Semantic Segmentation of Coal Maceral Groups

The segmentation of coal maceral groups can be described as a semantic segmentation process of coal maceral group images, which is of great significance for studying the chemical properties of coal. Generally, existing semantic segmentation models of coal maceral groups use the method of stacking parameters to achieve higher accuracy. It leads to increased computational requirements and impacts model training efficiency. At the same time, due to the professionalism and diversity of coal maceral group images sampling, obtaining the number of samples for model training requires a long time and professional personnel operation. To address these issues, We have innovatively developed an IoT-based DA-VIT parallel network model. By utilizing this model, we can continuously broaden the dataset through IoT and achieving sustained improvement in the accuracy of coal maceral groups segmentation. Besides, we decouple the parallel network from the backbone network to ensure the normal using of the backbone network during model data updates. Secondly, DCSA mechanism of DA-VIT is introduced to enhance the local feature information of coal microscopic images. This DCSA can decompose the large kernels of convolutional attention into multiple scales and reduce 81.18% of parameters.Finally, we performed the contrast experiment and ablation experiment between DA-VIT and state-of-the-art methods at lots of evaluation metrics. Experimental results show that DA-VIT-Base achieves 92.14% pixel accuracy and 63.18% mIoU. Params and FLOPs of DA-VIT-Tiny are 4.95M and 8.99G, respectively. All of the evaluation metrics of the proposed DA-VIT are better than other state-of-the-art methods.

cs.CV↗

arXiv2025

Coupled-channel study of $4S$-$3D$ mixing dynamics in $ψ(4220)$ and $ψ(4380)$

Among charmoniumlike $XYZ$ states, the $ψ(4220)$ and $ψ(4380)$ states have emerged as key candidates for exploring the charmonium spectrum. In this work, we propose a $4S$-$3D$ charmonium mixing scheme for the $ψ(4220)$ and $ψ(4380)$, induced by coupled-channel effects. By constructing a coupled-channel model, we identify the dynamical mechanism responsible for the large mixing angle observed in previous studies, which cannot be explained by conventional potential models alone. Our analysis reveals that the $DD_1$ channel significantly influences the lower state ($ψ(4220)$), while the $D^*D_1$ channel primarily affects the higher state ($ψ(4380)$). Furthermore, we investigate the two-body Okubo-Zweig-Iizuka (OZI)-allowed strong decay behaviors of these states, providing insights into their total widths. This study not only supports the $4S$-$3D$ mixing scheme but also offers a deeper understanding of the role of coupled channels in shaping the charmonium spectrum above 4 GeV. Our results align with experimental observations and provide a framework for interpreting future data on charmonium states.

hep-ph↗

arXiv2025

Can Compressed LLMs Truly Act? An Empirical Evaluation of Agentic Capabilities in LLM Compression

Post-training compression reduces the computational and memory costs of large language models (LLMs), enabling resource-efficient deployment. However, existing compression benchmarks only focus on language modeling (e.g., perplexity) and natural language understanding tasks (e.g., GLUE accuracy), ignoring the agentic capabilities - workflow, tool use/function call, long-context understanding and real-world application. We introduce the Agent Compression Benchmark (ACBench), the first comprehensive benchmark for evaluating how compression impacts LLMs' agentic abilities. ACBench spans (1) 12 tasks across 4 capabilities (e.g., WorfBench for workflow generation, Needle-in-Haystack for long-context retrieval), (2) quantization (GPTQ, AWQ) and pruning (Wanda, SparseGPT), and (3) 15 models, including small (Gemma-2B), standard (Qwen2.5 7B-32B), and distilled reasoning LLMs (DeepSeek-R1-Distill). Our experiments reveal compression tradeoffs: 4-bit quantization preserves workflow generation and tool use (1%-3% drop) but degrades real-world application accuracy by 10%-15%. We introduce ERank, Top-k Ranking Correlation and Energy to systematize analysis. ACBench provides actionable insights for optimizing LLM compression in agentic scenarios. The code can be found in https://github.com/pprp/ACBench.

cs.LG↗

arXiv2025

Electromagnetic properties of possible triple-charm molecular hexaquarks

In this study, we investigate the radiative transitions of predicted triple-charm molecular hexaquarks, which play a significant role in understanding their overall spectroscopic properties. As experimentally measurable quantities, the radiative decay widths provide insights into the internal structures of these triple-charm molecular hexaquarks. Additionally, we calculate their corresponding magnetic moments, which, together with the radiative decay widths, offer a comprehensive picture of the electromagnetic properties of these exotic states. This information is valuable for guiding future experimental searches and advancing our understanding of these unique hadronic systems.

hep-ph↗

arXiv2025

Topological Machine Learning for Protein-Nucleic Acid Binding Affinity Changes Upon Mutation

Understanding how protein mutations affect protein-nucleic acid binding is critical for unraveling disease mechanisms and advancing therapies. Current experimental approaches are laborious, and computational methods remain limited in accuracy. To address this challenge, we propose a novel topological machine learning model (TopoML) combining persistent Laplacian (from topological data analysis) with multi-perspective features: physicochemical properties, topological structures, and protein Transformer-derived sequence embeddings. This integrative framework captures robust representations of protein-nucleic acid binding interactions. To validate the proposed method, we employ two datasets, a protein-DNA dataset with 596 single-point amino acid mutations, and a protein-RNA dataset with 710 single-point amino acid mutations. We show that the proposed TopoML model outperforms state-of-the-art methods in predicting mutation-induced binding affinity changes for protein-DNA and protein-RNA complexes.

q-bio.QM↗

arXiv2025

Efficient Reasoning via Chain of Unconscious Thought

Large Reasoning Models (LRMs) achieve promising performance but compromise token efficiency due to verbose reasoning processes. Unconscious Thought Theory (UTT) posits that complex problems can be solved more efficiently through internalized cognitive processes. Inspired by UTT, we propose a new reasoning paradigm, termed Chain of Unconscious Thought (CoUT), to improve the token efficiency of LRMs by guiding them to mimic human unconscious thought and internalize reasoning processes. Concretely, we first prompt the model to internalize the reasoning by thinking in the hidden layer. Then, we design a bag of token-efficient strategies to further help models reduce unnecessary tokens yet preserve the performance. Our work reveals that models may possess beneficial unconscious thought, enabling improved efficiency without sacrificing performance. Extensive experiments demonstrate the effectiveness of CoUT. Remarkably, it surpasses CoT by reducing token usage by 47.62% while maintaining comparable accuracy, as shown in Figure 1. The code of CoUT is available at this link: https://github.com/Rohan-GRH/CoUT

cs.CL↗

arXiv2025

SLearnLLM: A Self-Learning Framework for Efficient Domain-Specific Adaptation of Large Language Models

When using supervised fine-tuning (SFT) to adapt large language models (LLMs) to specific domains, a significant challenge arises: should we use the entire SFT dataset for fine-tuning? Common practice often involves fine-tuning directly on the entire dataset due to limited information on the LLM's past training data. However, if the SFT dataset largely overlaps with the model's existing knowledge, the performance gains are minimal, leading to wasted computational resources. Identifying the unknown knowledge within the SFT dataset and using it to fine-tune the model could substantially improve the training efficiency. To address this challenge, we propose a self-learning framework for LLMs inspired by human learning pattern. This framework takes a fine-tuning (SFT) dataset in a specific domain as input. First, the LLMs answer the questions in the SFT dataset. The LLMs then objectively grade the responses and filter out the incorrectly answered QA pairs. Finally, we fine-tune the LLMs based on this filtered QA set. Experimental results in the fields of agriculture and medicine demonstrate that our method substantially reduces training time while achieving comparable improvements to those attained with full dataset fine-tuning. By concentrating on the unknown knowledge within the SFT dataset, our approach enhances the efficiency of fine-tuning LLMs.

cs.CL↗

arXiv2025

Multimodal Online Federated Learning with Modality Missing in Internet of Things

The Internet of Things (IoT) ecosystem generates vast amounts of multimodal data from heterogeneous sources such as sensors, cameras, and microphones. As edge intelligence continues to evolve, IoT devices have progressed from simple data collection units to nodes capable of executing complex computational tasks. This evolution necessitates the adoption of distributed learning strategies to effectively handle multimodal data in an IoT environment. Furthermore, the real-time nature of data collection and limited local storage on edge devices in IoT call for an online learning paradigm. To address these challenges, we introduce the concept of Multimodal Online Federated Learning (MMO-FL), a novel framework designed for dynamic and decentralized multimodal learning in IoT environments. Building on this framework, we further account for the inherent instability of edge devices, which frequently results in missing modalities during the learning process. We conduct a comprehensive theoretical analysis under both complete and missing modality scenarios, providing insights into the performance degradation caused by missing modalities. To mitigate the impact of modality missing, we propose the Prototypical Modality Mitigation (PMM) algorithm, which leverages prototype learning to effectively compensate for missing modalities. Experimental results on two multimodal datasets further demonstrate the superior performance of PMM compared to benchmarks.

cs.LG↗

arXiv2025

Efficient Partitioning Vision Transformer on Edge Devices for Distributed Inference

Deep learning models are increasingly utilized on resource-constrained edge devices for real-time data analytics. Recently, Vision Transformer and their variants have shown exceptional performance in various computer vision tasks. However, their substantial computational requirements and low inference latency create significant challenges for deploying such models on resource-constrained edge devices. To address this issue, we propose a novel framework, ED-ViT, which is designed to efficiently split and execute complex Vision Transformers across multiple edge devices. Our approach involves partitioning Vision Transformer models into several sub-models, while each dedicated to handling a specific subset of data classes. To further reduce computational overhead and inference latency, we introduce a class-wise pruning technique that decreases the size of each sub-model. Through extensive experiments conducted on five datasets using three model architectures and actual implementation on edge devices, we demonstrate that our method significantly cuts down inference latency on edge devices and achieves a reduction in model size by up to 28.9 times and 34.1 times, respectively, while maintaining test accuracy comparable to the original Vision Transformer. Additionally, we compare ED-ViT with two state-of-the-art methods that deploy CNN and SNN models on edge devices, evaluating metrics such as accuracy, inference time, and overall model size. Our comprehensive evaluation underscores the effectiveness of the proposed ED-ViT framework.

cs.CV↗