arXiv subjects

Jingyuan Zhang

Publications and source records attributed to Jingyuan Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2025

Capybara-OMNI: An Efficient Paradigm for Building Omni-Modal Language Models

With the development of Multimodal Large Language Models (MLLMs), numerous outstanding accomplishments have emerged within the open-source community. Due to the complexity of creating and training multimodal data pairs, it is still a computational and time-consuming process to build powerful MLLMs. In this work, we introduce Capybara-OMNI, an MLLM that trains in a lightweight and efficient manner and supports understanding text, image, video, and audio modalities. We present in detail the framework design, the data construction, and the training recipe, to develop an MLLM step-by-step to obtain competitive performance. We also provide exclusive benchmarks utilized in our experiments to show how to properly verify understanding capabilities across different modalities. Results show that by following our guidance, we can efficiently build an MLLM that achieves competitive performance among models of the same scale on various multimodal benchmarks. Additionally, to enhance the multimodal instruction following and conversational capabilities of the model, we further discuss how to train the chat version upon an MLLM understanding model, which is more in line with user habits for tasks like real-time interaction with humans. We publicly disclose the Capybara-OMNI model, along with its chat-based version. The disclosure includes both the model weights, a portion of the training data, and the inference codes, which are made available on GitHub.

cs.CL↗

arXiv2025

Data Metabolism: An Efficient Data Design Schema For Vision Language Model

Data curation plays a crucial role in training powerful Visual Language Models (VLMs). In this work, we introduce the concept of Data Metabolism and present our data-centric framework to build VLMs throughout the development lifecycle. Starting from a standard model architecture, we discuss and provide insights into two crucial development steps: data curation and iteration, forming a closed-loop system that continuously improves model performance. We show a detailed codebook on how to process existing massive datasets and build user-specific data flywheel. As a demonstration, we release a VLM, named Capybara-VL, which excels in typical multimodal tasks (e.g. , visual question answering, scientific reasoning, and text-rich tasks). Despite its relatively compact size, Capybara-VL surpasses several open-source models that are up to 10 times larger in size. Moreover, it achieves results that are on par with those of several leading proprietary models, demonstrating its remarkable competitiveness. These results highlight the power of our data-centric framework and the potential of training smaller and more efficient VLMs.

cs.CL↗

arXiv2025

AIBrix: Towards Scalable, Cost-Effective Large Language Model Inference Infrastructure

We introduce AIBrix, a cloud-native, open-source framework designed to optimize and simplify large-scale LLM deployment in cloud environments. Unlike traditional cloud-native stacks, AIBrix follows a co-design philosophy, ensuring every layer of the infrastructure is purpose-built for seamless integration with inference engines like vLLM. AIBrix introduces several key innovations to reduce inference costs and enhance performance including high-density LoRA management for dynamic adapter scheduling, LLM-specific autoscalers, and prefix-aware, load-aware routing. To further improve efficiency, AIBrix incorporates a distributed KV cache, boosting token reuse across nodes, leading to a 50% increase in throughput and a 70% reduction in inference latency. AIBrix also supports unified AI runtime which streamlines model management while maintaining vendor-agnostic engine compatibility. For large-scale multi-node inference, AIBrix employs hybrid orchestration -- leveraging Kubernetes for coarse-grained scheduling and Ray for fine-grained execution -- to balance efficiency and flexibility. Additionally, an SLO-driven GPU optimizer dynamically adjusts resource allocations, optimizing heterogeneous serving to maximize cost efficiency while maintaining service guarantees. Finally, AIBrix enhances system reliability with AI accelerator diagnostic tools, enabling automated failure detection and mock-up testing to improve fault resilience. AIBrix is available at https://github.com/vllm-project/aibrix.

cs.DC↗

arXiv2025

Benchmarking Multimodal RAG through a Chart-based Document Question-Answering Generation Framework

Multimodal Retrieval-Augmented Generation (MRAG) enhances reasoning capabilities by integrating external knowledge. However, existing benchmarks primarily focus on simple image-text interactions, overlooking complex visual formats like charts that are prevalent in real-world applications. In this work, we introduce a novel task, Chart-based MRAG, to address this limitation. To semi-automatically generate high-quality evaluation samples, we propose CHARt-based document question-answering GEneration (CHARGE), a framework that produces evaluation data through structured keypoint extraction, crossmodal verification, and keypoint-based generation. By combining CHARGE with expert validation, we construct Chart-MRAG Bench, a comprehensive benchmark for chart-based MRAG evaluation, featuring 4,738 question-answering pairs across 8 domains from real-world documents. Our evaluation reveals three critical limitations in current approaches: (1) unified multimodal embedding retrieval methods struggles in chart-based scenarios, (2) even with ground-truth retrieval, state-of-the-art MLLMs achieve only 58.19% Correctness and 73.87% Coverage scores, and (3) MLLMs demonstrate consistent text-over-visual modality bias during Chart-based MRAG reasoning. The CHARGE and Chart-MRAG Bench are released at https://github.com/Nomothings/CHARGE.git.

cs.AI↗

arXiv2025

Simplifying DINO via Coding Rate Regularization

DINO and DINOv2 are two model families being widely used to learn representations from unlabeled imagery data at large scales. Their learned representations often enable state-of-the-art performance for downstream tasks, such as image classification and segmentation. However, they employ many empirically motivated design choices and their training pipelines are highly complex and unstable -- many hyperparameters need to be carefully tuned to ensure that the representations do not collapse -- which poses considerable difficulty to improving them or adapting them to new domains. In this work, we posit that we can remove most such-motivated idiosyncrasies in the pre-training pipelines, and only need to add an explicit coding rate term in the loss function to avoid collapse of the representations. As a result, we obtain highly simplified variants of the DINO and DINOv2 which we call SimDINO and SimDINOv2, respectively. Remarkably, these simplified models are more robust to different design choices, such as network architecture and hyperparameters, and they learn even higher-quality representations, measured by performance on downstream tasks, offering a Pareto improvement over the corresponding DINO and DINOv2 models. This work highlights the potential of using simplifying design principles to improve the empirical practice of deep learning.

cs.CV↗

arXiv2025

Enhancing Security and Privacy in Federated Learning using Low-Dimensional Update Representation and Proximity-Based Defense

Federated Learning (FL) is a promising privacy-preserving machine learning paradigm that allows data owners to collaboratively train models while keeping their data localized. Despite its potential, FL faces challenges related to the trustworthiness of both clients and servers, particularly against curious or malicious adversaries. In this paper, we introduce a novel framework named \underline{F}ederated \underline{L}earning with Low-Dimensional \underline{U}pdate \underline{R}epresentation and \underline{P}roximity-Based defense (FLURP), designed to address privacy preservation and resistance to Byzantine attacks in distributed learning environments. FLURP employs $\mathsf{LinfSample}$ method, enabling clients to compute the $l_{\infty}$ norm across sliding windows of updates, resulting in a Low-Dimensional Update Representation (LUR). Calculating the shared distance matrix among LURs, rather than updates, significantly reduces the overhead of Secure Multi-Party Computation (SMPC) by three orders of magnitude while effectively distinguishing between benign and poisoned updates. Additionally, FLURP integrates a privacy-preserving proximity-based defense mechanism utilizing optimized SMPC protocols to minimize communication rounds. Our experiments demonstrate FLURP's effectiveness in countering Byzantine adversaries with low communication and runtime overhead. FLURP offers a scalable framework for secure and reliable FL in distributed environments, facilitating its application in scenarios requiring robust data management and security.

cs.CR↗

arXiv2025

Design, Construction, and Test of Compact, Distributed-Charge, X-Band Accelerator Systems that Enable Image-Guided, VHEE FLASH Radiotherapy

The design and optimization of laser-Compton x-ray systems based on compact distributed charge accelerator structures can enable micron-scale imaging of disease and the concomitant production of beams of Very High Energy Electrons (VHEEs) capable of producing FLASH-relevant dose rates. The physics of laser-Compton x-ray scattering ensures that the scattered x-rays follow exactly the trajectory of the incident electrons, thus providing a route to image-guided, VHEE FLASH radiotherapy. The keys to a compact architecture capable of producing both laser-Compton x-rays and VHEEs are the use of X-band RF accelerator structures which have been demonstrated to operate with over 100 MeV/m acceleration gradients. The operation of these structures in a distributed charge mode in which each radiofrequency (RF) cycle of the drive RF pulse is filled with a low-charge, high-brightness electron bunch is enabled by the illumination of a high-brightness photogun with a train of UV laser pulses synchronized to the frequency of the underlying accelerator system. The UV pulse trains are created by a patented pulse synthesis approach which utilizes the RF clock of the accelerator to phase and amplitude modulate a narrow band continuous wave (CW) seed laser. In this way it is possible to produce up to 10 $μ$A of average beam current from the accelerator. Such high current from a compact accelerator enables production of sufficient x-rays via laser-Compton scattering for clinical imaging and does so from a machine of "clinical" footprint. At the same time, the production of 1000 or greater individual micro-bunches per RF pulse enables > 10 nC of charge to be produced in a macrobunch of < 100 ns. The design, construction, and test of the 100-MeV class prototype system in Irvine, CA is also presented.

physics.acc-ph↗

arXiv2024

Human-centric Reward Optimization for Reinforcement Learning-based Automated Driving using Large Language Models

One of the key challenges in current Reinforcement Learning (RL)-based Automated Driving (AD) agents is achieving flexible, precise, and human-like behavior cost-effectively. This paper introduces an innovative approach that uses large language models (LLMs) to intuitively and effectively optimize RL reward functions in a human-centric way. We developed a framework where instructions and dynamic environment descriptions are input into the LLM. The LLM then utilizes this information to assist in generating rewards, thereby steering the behavior of RL agents towards patterns that more closely resemble human driving. The experimental results demonstrate that this approach not only makes RL agents more anthropomorphic but also achieves better performance. Additionally, various strategies for reward-proxy and reward-shaping are investigated, revealing the significant impact of prompt design on shaping an AD vehicle's behavior. These findings offer a promising direction for the development of more advanced, human-like automated driving systems. Our experimental data and source code can be found here

cs.AI↗

arXiv2024

Token Statistics Transformer: Linear-Time Attention via Variational Rate Reduction

The attention operator is arguably the key distinguishing factor of transformer architectures, which have demonstrated state-of-the-art performance on a variety of tasks. However, transformer attention operators often impose a significant computational burden, with the computational complexity scaling quadratically with the number of tokens. In this work, we propose a novel transformer attention operator whose computational complexity scales linearly with the number of tokens. We derive our network architecture by extending prior work which has shown that a transformer style architecture naturally arises by "white-box" architecture design, where each layer of the network is designed to implement an incremental optimization step of a maximal coding rate reduction objective (MCR$^2$). Specifically, we derive a novel variational form of the MCR$^2$ objective and show that the architecture that results from unrolled gradient descent of this variational objective leads to a new attention module called Token Statistics Self-Attention (TSSA). TSSA has linear computational and memory complexity and radically departs from the typical attention architecture that computes pairwise similarities between tokens. Experiments on vision, language, and long sequence tasks show that simply swapping TSSA for standard self-attention, which we refer to as the Token Statistics Transformer (ToST), achieves competitive performance with conventional transformers while being significantly more computationally efficient and interpretable. Our results also somewhat call into question the conventional wisdom that pairwise similarity style attention mechanisms are critical to the success of transformer architectures. Code will be available at https://github.com/RobinWu218/ToST.

cs.LG↗

arXiv2024

Smart-LLaMA: Two-Stage Post-Training of Large Language Models for Smart Contract Vulnerability Detection and Explanation

With the rapid development of blockchain technology, smart contract security has become a critical challenge. Existing smart contract vulnerability detection methods face three main issues: (1) Insufficient quality of datasets, lacking detailed explanations and precise vulnerability locations. (2) Limited adaptability of large language models (LLMs) to the smart contract domain, as most LLMs are pre-trained on general text data but minimal smart contract-specific data. (3) Lack of high-quality explanations for detected vulnerabilities, as existing methods focus solely on detection without clear explanations. These limitations hinder detection performance and make it harder for developers to understand and fix vulnerabilities quickly, potentially leading to severe financial losses. To address these problems, we propose Smart-LLaMA, an advanced detection method based on the LLaMA language model. First, we construct a comprehensive dataset covering four vulnerability types with labels, detailed explanations, and precise vulnerability locations. Second, we introduce Smart Contract-Specific Continual Pre-Training, using raw smart contract data to enable the LLM to learn smart contract syntax and semantics, enhancing their domain adaptability. Furthermore, we propose Explanation-Guided Fine-Tuning, which fine-tunes the LLM using paired vulnerable code and explanations, enabling both vulnerability detection and reasoned explanations. We evaluate explanation quality through LLM and human evaluation, focusing on Correctness, Completeness, and Conciseness. Experimental results show that Smart-LLaMA outperforms state-of-the-art baselines, with average improvements of 6.49% in F1 score and 3.78% in accuracy, while providing reliable explanations.

cs.CR↗

arXiv2024

Enhancing Federated Domain Adaptation with Multi-Domain Prototype-Based Federated Fine-Tuning

Federated Domain Adaptation (FDA) is a Federated Learning (FL) scenario where models are trained across multiple clients with unique data domains but a shared category space, without transmitting private data. The primary challenge in FDA is data heterogeneity, which causes significant divergences in gradient updates when using conventional averaging-based aggregation methods, reducing the efficacy of the global model. This further undermines both in-domain and out-of-domain performance (within the same federated system but outside the local client). To address this, we propose a novel framework called \textbf{M}ulti-domain \textbf{P}rototype-based \textbf{F}ederated Fine-\textbf{T}uning (MPFT). MPFT fine-tunes a pre-trained model using multi-domain prototypes, i.e., pretrained representations enriched with domain-specific information from category-specific local data. This enables supervised learning on the server to derive a globally optimized adapter that is subsequently distributed to local clients, without the intrusion of data privacy. Empirical results show that MPFT significantly improves both in-domain and out-of-domain accuracy over conventional methods, enhancing knowledge preservation and adaptation in FDA. Notably, MPFT achieves convergence within a single communication round, greatly reducing computation and communication costs. To ensure privacy, MPFT applies differential privacy to protect the prototypes. Additionally, we develop a prototype-based feature space hijacking attack to evaluate robustness, confirming that raw data samples remain unrecoverable even after extensive training epochs. The complete implementation of MPFL is available at \url{https://anonymous.4open.science/r/DomainFL/}.

cs.LG↗

arXiv2024

Ctrl123: Consistent Novel View Synthesis via Closed-Loop Transcription

Large image diffusion models have demonstrated zero-shot capability in novel view synthesis (NVS). However, existing diffusion-based NVS methods struggle to generate novel views that are accurately consistent with the corresponding ground truth poses and appearances, even on the training set. This consequently limits the performance of downstream tasks, such as image-to-multiview generation and 3D reconstruction. We realize that such inconsistency is largely due to the fact that it is difficult to enforce accurate pose and appearance alignment directly in the diffusion training, as mostly done by existing methods such as Zero123. To remedy this problem, we propose Ctrl123, a closed-loop transcription-based NVS diffusion method that enforces alignment between the generated view and ground truth in a pose-sensitive feature space. Our extensive experiments demonstrate the effectiveness of Ctrl123 on the tasks of NVS and 3D reconstruction, achieving significant improvements in both multiview-consistency and pose-consistency over existing methods.

cs.CV↗

arXiv2024

Decoding at the Speed of Thought: Harnessing Parallel Decoding of Lexical Units for LLMs

Large language models have demonstrated exceptional capability in natural language understanding and generation. However, their generation speed is limited by the inherently sequential nature of their decoding process, posing challenges for real-time applications. This paper introduces Lexical Unit Decoding (LUD), a novel decoding methodology implemented in a data-driven manner, accelerating the decoding process without sacrificing output quality. The core of our approach is the observation that a pre-trained language model can confidently predict multiple contiguous tokens, forming the basis for a \textit{lexical unit}, in which these contiguous tokens could be decoded in parallel. Extensive experiments validate that our method substantially reduces decoding time while maintaining generation quality, i.e., 33\% speed up on natural language generation with no quality loss, and 30\% speed up on code generation with a negligible quality loss of 3\%. Distinctively, LUD requires no auxiliary models and does not require changes to existing architectures. It can also be integrated with other decoding acceleration methods, thus achieving an even more pronounced inference efficiency boost. We posit that the foundational principles of LUD could define a new decoding paradigm for future language models, enhancing their applicability for a broader spectrum of applications. All codes are be publicly available at https://github.com/tjunlp-lab/Lexical-Unit-Decoding-LUD-. Keywords: Parallel Decoding, Lexical Unit Decoding, Large Language Model

cs.CL↗

arXiv2024

Chain-of-Specificity: An Iteratively Refining Method for Eliciting Knowledge from Large Language Models

Large Language Models (LLMs) exhibit remarkable generative capabilities, enabling the generation of valuable information. Despite these advancements, previous research found that LLMs sometimes struggle with adhering to specific constraints (e.g., in specific place or at specific time), at times even overlooking them, which leads to responses that are either too generic or not fully satisfactory. Existing approaches attempted to address this issue by decomposing or rewriting input instructions, yet they fall short in adequately emphasizing specific constraints and in unlocking the underlying knowledge (e.g., programming within the context of software development). In response, this paper proposes a simple yet effective method named Chain-of-Specificity (CoS). Specifically, CoS iteratively emphasizes the specific constraints in the input instructions, unlocks knowledge within LLMs, and refines responses. Experiments conducted on publicly available and self-build complex datasets demonstrate that CoS outperforms existing methods in enhancing generated content especially for the specificity. Besides, as the number of specific constraints increase, other baselines falter, while CoS still performs well. Moreover, we show that distilling responses generated by CoS effectively enhances the ability of smaller models to follow the constrained instructions. Resources of this paper will be released for further research.

cs.AI↗

arXiv2024

Evil Geniuses: Delving into the Safety of LLM-based Agents

Rapid advancements in large language models (LLMs) have revitalized in LLM-based agents, exhibiting impressive human-like behaviors and cooperative capabilities in various scenarios. However, these agents also bring some exclusive risks, stemming from the complexity of interaction environments and the usability of tools. This paper delves into the safety of LLM-based agents from three perspectives: agent quantity, role definition, and attack level. Specifically, we initially propose to employ a template-based attack strategy on LLM-based agents to find the influence of agent quantity. In addition, to address interaction environment and role specificity issues, we introduce Evil Geniuses (EG), an effective attack method that autonomously generates prompts related to the original role to examine the impact across various role definitions and attack levels. EG leverages Red-Blue exercises, significantly improving the generated prompt aggressiveness and similarity to original roles. Our evaluations on CAMEL, Metagpt and ChatDev based on GPT-3.5 and GPT-4, demonstrate high success rates. Extensive evaluation and discussion reveal that these agents are less robust, prone to more harmful behaviors, and capable of generating stealthier content than LLMs, highlighting significant safety challenges and guiding future research. Our code is available at https://github.com/T1aNS1R/Evil-Geniuses.

cs.CL↗

arXiv2024

Multiperson Detection and Vital-Sign Sensing Empowered by Space-Time-Coding RISs

Passive human sensing using wireless signals has attracted increasing attention due to its superiorities of non-contact and robustness in various lighting conditions. However, when multiple human individuals are present, their reflected signals could be intertwined in the time, frequency and spatial domains, making it challenging to separate them. To address this issue, this paper proposes a novel system for multiperson detection and monitoring of vital signs (i.e., respiration and heartbeat) with the assistance of space-time-coding (STC) reconfigurable intelligent metasurfaces (RISs). Specifically, the proposed system scans the area of interest (AoI) for human detection by using the harmonic beams generated by the STC RIS. Simultaneously, frequencyorthogonal beams are assigned to each detected person for accurate estimation of their respiration rate (RR) and heartbeat rate (HR). Furthermore, to efficiently extract the respiration signal and the much weaker heartbeat signal, we propose an improved variational mode decomposition (VMD) algorithm to accurately decompose the complex reflected signals into a smaller number of intrinsic mode functions (IMFs). We build a prototype to validate the proposed multiperson detection and vital-sign monitoring system. Experimental results demonstrate that the proposed system can simultaneously monitor the vital signs of up to four persons. The errors of RR and HR estimation using the improved VMD algorithm are below 1 RPM (respiration per minute) and 5 BPM (beats per minute), respectively. Further analysis reveals that the flexible beam controlling mechanism empowered by the STC RIS can reduce the noise reflected from other irrelative objects on the physical layer, and improve the signal-to-noise ratio of echoes from the human chest.

eess.SP↗

arXiv2023

Hyper-Relational Knowledge Graph Neural Network for Next POI

With the advancement of mobile technology, Point of Interest (POI) recommendation systems in Location-based Social Networks (LBSN) have brought numerous benefits to both users and companies. Many existing works employ Knowledge Graph (KG) to alleviate the data sparsity issue in LBSN. These approaches primarily focus on modeling the pair-wise relations in LBSN to enrich the semantics and thereby relieve the data sparsity issue. However, existing approaches seldom consider the hyper-relations in LBSN, such as the mobility relation (a 3-ary relation: user-POI-time). This makes the model hard to exploit the semantics accurately. In addition, prior works overlook the rich structural information inherent in KG, which consists of higher-order relations and can further alleviate the impact of data sparsity.To this end, we propose a Hyper-Relational Knowledge Graph Neural Network (HKGNN) model. In HKGNN, a Hyper-Relational Knowledge Graph (HKG) that models the LBSN data is constructed to maintain and exploit the rich semantics of hyper-relations. Then we proposed a Hypergraph Neural Network to utilize the structural information of HKG in a cohesive way. In addition, a self-attention network is used to leverage sequential information and make personalized recommendations. Furthermore, side information, essential in reducing data sparsity by providing background knowledge of POIs, is not fully utilized in current methods. In light of this, we extended the current dataset with available side information to further lessen the impact of data sparsity. Results of experiments on four real-world LBSN datasets demonstrate the effectiveness of our approach compared to existing state-of-the-art methods.

cs.AI↗

arXiv2023

Passive Human Sensing Enhanced by Reconfigurable Intelligent Surface: Opportunities and Challenges

Reconfigurable intelligent surfaces (RISs) have flexible and exceptional performance in manipulating electromagnetic waves and customizing wireless channels. These capabilities enable them to provide a plethora of valuable activity-related information for promoting wireless human sensing. In this article, we present a comprehensive review of passive human sensing using radio frequency signals with the assistance of RISs. Specifically, we first introduce fundamental principles and physical platform of RISs. Subsequently, based on the specific applications, we categorize the state-of-the-art human sensing techniques into three types, including human imaging,localization, and activity recognition. Meanwhile, we would also investigate the benefits that RISs bring to these applications. Furthermore, we explore the application of RISs in human micro-motion sensing, and propose a vital signs monitoring system enhanced by RISs. Experimental results are presented to demonstrate the promising potential of RISs in sensing vital signs for manipulating individuals. Finally, we discuss the technical challenges and opportunities in this field.

eess.SP↗