arXiv subjects

Yuhan Ma

Publications and source records attributed to Yuhan Ma.

18 recordsLinked to original sources

arXiv2026

SecureClaw: Clawing Back Control of LLM Agents

Tool-using large language model (LLM) agents face two distinct security failures: unauthorized external actions and exposure of sensitive plaintext inside the runtime before any final output check can intervene. Existing defenses usually protect one boundary, either the planner/runtime or the action sink, and therefore do not by themselves secure both surfaces. We present SecureClaw, a dual-boundary architecture that places authorization at the effect sink and plaintext confinement at the read boundary. Sensitive reads pass through a trusted gateway that replaces raw values with opaque handles and, in the evaluated deployment, bounded summaries as an explicit declassification interface. Writes that change external state follow a PREVIEW$\rightarrow$COMMIT protocol in which only a trusted executor may commit the exact canonical request authorized by policy. The runtime can still plan over summaries and symbolic references, but cannot directly dereference secrets or perform side effects. Across AgentDojo, AgentLeak, and Agent Security Bench (ASB), SecureClaw is the only defense we evaluate in a common harness that simultaneously retains usable task utility and achieves 0\% attack success rate (ASR) on ASB, 0.64\% ASR on AgentDojo, and 3.23\% overall leak on AgentLeak's attacked parity lane, which measures final-output and internal-relay leakage.

cs.CR↗

arXiv2026

FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing

Two-server secure inference allows a client to query a hosted large language model (LLM) without revealing prompts or embeddings. Recent GPU systems based on function secret sharing (FSS) make linear layers efficient, but fixed-point nonlinearities and helper operations remain a bottleneck because each operator is typically implemented as a bespoke protocol with its own comparisons, wrap-around corrections, and preprocessing material. We present FuseFSS, a compiler that replaces per-operator protocol design with a single compilation pipeline. For each scalar fixed-point operator, a compact specification lists its interval partition, low-degree arithmetic pieces, and required predicate bits. The compiler emits two batched FSS evaluations on the public masked value: one packed comparison that returns all predicate bits, and one vector interval lookup that returns the active coefficients and constants. Compared to the current state-of-the-art FSS-based GPU secure inference, FuseFSS preserves accuracy while achieving a $1.24\times$--$1.50\times$ end-to-end speedup and reducing online communication by $9\%$--$16\%$ on BERT and GPT-style models; preprocessing is also lighter, with $14\%$--$23\%$ lower key-generation time and $20\%$--$24\%$ smaller keys.

cs.CR↗

arXiv2026

Design and Verification of a New Current Source for Tsinghua Tabletop Kibble Balance

This paper presents an ultra-stable current source tailored for the one-mode, two-phase (OMTP) measurement scheme in Tsinghua tabletop Kibble balances. To achieve simultaneous high resolution and nA/A-level stability, a composite 'coarse-fine' control topology is proposed, utilizing a dual-DAC architecture and an active digital feedback loop. Experimental results show that the Allan deviation reaches 1 nA/A at an integration time of approximately 3 minutes, representing a tenfold improvement in measurement speed compared to commercial-source-based setups. Furthermore, the design offers a significant cost advantage, providing a satisfying option for high-precision, cost-effective mass realization.

physics.ins-det↗

arXiv2026

Observation antibunching with classical light in a linear interferometer

Understanding the boundary between classical and nonclassical phenomena is important for both fundamental researches in quantum optics and applications in quantum information. One of the most interesting research directions in this field is exploring nonclassical effects with classical light. In this paper, we will show that it is possible to observe antibunching with thermal light in a Hanbury Brown-Twiss interferometer by treating single-photon detectors as photon-number-resolving detectors to perform photon-number projection measurements. Both temporal and spatial antibunching is observed via the correlation of two detectors detecting one and zero photon, respectively. By comparing the measured results of thermal and laser light, it is found that the observed antibunching arises from the combined effect of photon statistics of thermal light and photon-number projection measurement.The classical and nonclassical nature of the observed antibunching is analyzed. The results are helpful to understand the connection between classical and nonclassical correlation and may find applications in multiphoton interference and quantum imaging.

quant-ph↗

arXiv2026

PTCBENCH: Benchmarking Contextual Stability of Personality Traits in LLM Systems

With the increasing deployment of large language models (LLMs) in affective agents and AI systems, maintaining a consistent and authentic LLM personality becomes critical for user trust and engagement. However, existing work overlooks a fundamental psychological consensus that personality traits are dynamic and context-dependent. To bridge this gap, we introduce PTCBENCH, a systematic benchmark designed to quantify the consistency of LLM personalities under controlled situational contexts. PTCBENCH subjects models to 12 distinct external conditions spanning diverse location contexts and life events, and rigorously assesses the personality using the NEO Five-Factor Inventory. Our study on 39,240 personality trait records reveals that certain external scenarios (e.g., "Unemployment") can trigger significant personality changes of LLMs, and even alter their reasoning capabilities. Overall, PTCBENCH establishes an extensible framework for evaluating personality consistency in realistic, evolving environments, offering actionable insights for developing robust and psychologically aligned AI systems.

cs.CL↗

arXiv2025

EchoVLA: Robotic Vision-Language-Action Model with Synergistic Declarative Memory for Mobile Manipulation

Recent progress in Vision-Language-Action (VLA) models has enabled embodied agents to interpret multimodal instructions and perform complex tasks. However, existing VLAs are mostly confined to short-horizon, table-top manipulation, lacking the memory and reasoning capability required for mobile manipulation, where agents must coordinate navigation and manipulation under changing spatial contexts. In this work, we present EchoVLA, a memory-aware VLA model for mobile manipulation. EchoVLA incorporates a synergistic declarative memory inspired by the human brain, consisting of a scene memory that maintains a collection of spatial-semantic maps and an episodic memory that stores task-level experiences with multimodal contextual features. The two memories are individually stored, updated, and retrieved based on current observations, task history, and instructions, and their retrieved representations are fused via coarse- and fine-grained attention to guide base-arm diffusion policies. To support large-scale training, we further introduce MoMani, an automated benchmark that generates expert-level trajectories through multimodal large language model (MLLM)-guided planning and feedback-driven refinement, supplemented with real-robot demonstrations. Comprehensive simulated and real-world results demonstrate that EchoVLA substantially improves overall performance, e.g., it achieves the highest success rates of 0.52 on manipulation/navigation tasks and 0.31 on mobile manipulation tasks in simulation, exceeding the strong baseline $\pi_{0.5}$ by +0.20 and +0.11, respectively.

cs.RO↗

arXiv2025

Chimera: Harnessing Multi-Agent LLMs for Automatic Insider Threat Simulation

Insider threats pose a persistent and critical security risk, yet are notoriously difficult to detect in complex enterprise environments, where malicious actions are often hidden within seemingly benign user behaviors. Although machine-learning-based insider threat detection (ITD) methods have shown promise, their effectiveness is fundamentally limited by the scarcity of high-quality and realistic training data. Enterprise internal data is highly sensitive and rarely accessible, while existing public and synthetic datasets are either small-scale or lack sufficient realism, semantic richness, and behavioral diversity. To address this challenge, we propose Chimera, an LLM-based multi-agent framework that automatically simulates both benign and malicious insider activities and generates comprehensive system logs across diverse enterprise environments. Chimera models each agent as an individual employee with fine-grained roles and supports group meetings, pairwise interactions, and self-organized scheduling to capture realistic organizational dynamics. Based on 15 insider attacks abstracted from real-world incidents, we deploy Chimera in three representative data-sensitive organizational scenarios and construct ChimeraLog, a new dataset for developing and evaluating ITD methods. We evaluate ChimeraLog through human studies and quantitative analyses, demonstrating its diversity and realism. Experiments with existing ITD methods show substantially lower detection performance on ChimeraLog compared to prior datasets, indicating a more challenging and realistic benchmark. Moreover, despite distribution shifts, models trained on ChimeraLog exhibit strong generalization, highlighting the practical value of LLM-based multi-agent simulation for advancing insider threat detection.

cs.CR↗

arXiv2025

PatchFuzz: Patch Fuzzing for JavaScript Engines

Patch fuzzing is a technique aimed at identifying vulnerabilities that arise from newly patched code. While researchers have made efforts to apply patch fuzzing to testing JavaScript engines with considerable success, these efforts have been limited to using ordinary test cases or publicly available vulnerability PoCs (Proof of Concepts) as seeds, and the sustainability of these approaches is hindered by the challenges associated with automating the PoC collection. To address these limitations, we propose an end-to-end sustainable approach for JavaScript engine patch fuzzing, named PatchFuzz. It automates the collection of PoCs of a broader range of historical vulnerabilities and leverages both the PoCs and their corresponding patches to uncover new vulnerabilities more effectively. PatchFuzz starts by recognizing git commits which intend to fix security bugs. Subsequently, it extracts and processes PoCs from these commits to form the seeds for fuzzing, while utilizing code revisions to focus limited fuzzing resources on the more vulnerable code areas through selective instrumentation. The mutation strategy of PatchFuzz is also optimized to maximize the potential of the PoCs. Experimental results demonstrate the effectiveness of PatchFuzz. Notably, 54 bugs across six popular JavaScript engines have been exposed and a total of $62,500 bounties has been received.

cs.CR↗

arXiv2025

On walk domination: Between different types of walks and $m_3$-path

Given two non-adjacent vertices $ u $ and $ v $, we say a $uv$-walk $ W $ dominates a $uv$-walk $ W' $ if every internal vertex of $ W' $ is adjacent to some internal vertex of $ W $ or belongs to $ W $. A class of walks $\mathbf{A}$ dominates a class of walks $\mathbf{B}$ if for every pair of non-adjacent vertices $u,v$ in the graph, every $uv$-walk in $\mathbf{A}$ dominates every $uv$-walk in $\mathbf{B}$. This paper investigates the domination relationships among various types of walks connecting two non-adjacent vertices in a graph. In particular, we focus on a problem proposed by Tondato (2024). We study the domination between different walk types (shortest paths, toll walks, weakly toll walks, $l_k$-paths for $k\in \left\{2,3\right\}$) and $m_3$-paths. Furthermore, we show how these relationships give rise to characterizations of graph classes.

math.CO↗

arXiv2024

Who is Undercover? Guiding LLMs to Explore Multi-Perspective Team Tactic in the Game

Large Language Models (LLMs) are pivotal AI agents in complex tasks but still face challenges in open decision-making problems within complex scenarios. To address this, we use the language logic game ``Who is Undercover?'' (WIU) as an experimental platform to propose the Multi-Perspective Team Tactic (MPTT) framework. MPTT aims to cultivate LLMs' human-like language expression logic, multi-dimensional thinking, and self-perception in complex scenarios. By alternating speaking and voting sessions, integrating techniques like self-perspective, identity-determination, self-reflection, self-summary and multi-round find-teammates, LLM agents make rational decisions through strategic concealment and communication, fostering human-like trust. Preliminary results show that MPTT, combined with WIU, leverages LLMs' cognitive capabilities to create a decision-making framework that can simulate real society. This framework aids minority groups in communication and expression, promoting fairness and diversity in decision-making. Additionally, our Human-in-the-loop experiments demonstrate that LLMs can learn and align with human behaviors through interactive, indicating their potential for active participation in societal decision-making.

cs.AI↗

arXiv2024

Characterizations of the graphs with dominating parameters

A subset $S$ of vertices of $G$ is a \textit{dominating set} of $G$ if every vertex in $V(G)-S$ has a neighbor in $S$. The \textit{domination number} $\gamma(G)$ is the minimum cardinality of a dominating set of $G$. A dominating set $S$ is a \textit{total dominating set} if $N(S)$=$V$ where $N(S)$ is the neighbor of $S$. The \textit{total domination number} $\gamma_t(G)$ equals the minimum cardinality of a total dominating set of $G$. A set $D$ is an \textit{isolate set} if the induced subgragh $G[D]$ has at least one isolated vertex. The \textit{isolate number} $i_0(G)$ is the minimum cardinality of a maximal isolate set. In this paper we study these parameters and answer open problems proposed by Hamid et al. in 2016.

math.CO↗

arXiv2024

Enhancing Deep Learning with Optimized Gradient Descent: Bridging Numerical Methods and Neural Network Training

Optimization theory serves as a pivotal scientific instrument for achieving optimal system performance, with its origins in economic applications to identify the best investment strategies for maximizing benefits. Over the centuries, from the geometric inquiries of ancient Greece to the calculus contributions by Newton and Leibniz, optimization theory has significantly advanced. The persistent work of scientists like Lagrange, Cauchy, and von Neumann has fortified its progress. The modern era has seen an unprecedented expansion of optimization theory applications, particularly with the growth of computer science, enabling more sophisticated computational practices and widespread utilization across engineering, decision analysis, and operations research. This paper delves into the profound relationship between optimization theory and deep learning, highlighting the omnipresence of optimization problems in the latter. We explore the gradient descent algorithm and its variants, which are the cornerstone of optimizing neural networks. The chapter introduces an enhancement to the SGD optimizer, drawing inspiration from numerical optimization methods, aiming to enhance interpretability and accuracy. Our experiments on diverse deep learning tasks substantiate the improved algorithm's efficacy. The paper concludes by emphasizing the continuous development of optimization theory and its expanding role in solving intricate problems, enhancing computational capabilities, and informing better policy decisions.

cs.LG↗

arXiv2024

Text classification optimization algorithm based on graph neural network

In the field of natural language processing, text classification, as a basic task, has important research value and application prospects. Traditional text classification methods usually rely on feature representations such as the bag of words model or TF-IDF, which overlook the semantic connections between words and make it challenging to grasp the deep structural details of the text. Recently, GNNs have proven to be a valuable asset for text classification tasks, thanks to their capability to handle non-Euclidean data efficiently. However, the existing text classification methods based on GNN still face challenges such as complex graph structure construction and high cost of model training. This paper introduces a text classification optimization algorithm utilizing graph neural networks. By introducing adaptive graph construction strategy and efficient graph convolution operation, the accuracy and efficiency of text classification are effectively improved. The experimental results demonstrate that the proposed method surpasses traditional approaches and existing GNN models across multiple public datasets, highlighting its superior performance and feasibility for text classification tasks.

cs.CL↗

arXiv2024

Research on Optimization of Natural Language Processing Model Based on Multimodal Deep Learning

This project intends to study the image representation based on attention mechanism and multimodal data. By adding multiple pattern layers to the attribute model, the semantic and hidden layers of image content are integrated. The word vector is quantified by the Word2Vec method and then evaluated by a word embedding convolutional neural network. The published experimental results of the two groups were tested. The experimental results show that this method can convert discrete features into continuous characters, thus reducing the complexity of feature preprocessing. Word2Vec and natural language processing technology are integrated to achieve the goal of direct evaluation of missing image features. The robustness of the image feature evaluation model is improved by using the excellent feature analysis characteristics of a convolutional neural network. This project intends to improve the existing image feature identification methods and eliminate the subjective influence in the evaluation process. The findings from the simulation indicate that the novel approach has developed is viable, effectively augmenting the features within the produced representations.

cs.CL↗

arXiv2024

Advancements in Feature Extraction Recognition of Medical Imaging Systems Through Deep Learning Technique

This study introduces a novel unsupervised medical image feature extraction method that employs spatial stratification techniques. An objective function based on weight is proposed to achieve the purpose of fast image recognition. The algorithm divides the pixels of the image into multiple subdomains and uses a quadtree to access the image. A technique for threshold optimization utilizing a simplex algorithm is presented. Aiming at the nonlinear characteristics of hyperspectral images, a generalized discriminant analysis algorithm based on kernel function is proposed. In this project, a hyperspectral remote sensing image is taken as the object, and we investigate its mathematical modeling, solution methods, and feature extraction techniques. It is found that different types of objects are independent of each other and compact in image processing. Compared with the traditional linear discrimination method, the result of image segmentation is better. This method can not only overcome the disadvantage of the traditional method which is easy to be affected by light, but also extract the features of the object quickly and accurately. It has important reference significance for clinical diagnosis.

eess.IV↗

arXiv2024

Robust Dual-Modal Speech Keyword Spotting for XR Headsets

While speech interaction finds widespread utility within the Extended Reality (XR) domain, conventional vocal speech keyword spotting systems continue to grapple with formidable challenges, including suboptimal performance in noisy environments, impracticality in situations requiring silence, and susceptibility to inadvertent activations when others speak nearby. These challenges, however, can potentially be surmounted through the cost-effective fusion of voice and lip movement information. Consequently, we propose a novel vocal-echoic dual-modal keyword spotting system designed for XR headsets. We devise two different modal fusion approches and conduct experiments to test the system's performance across diverse scenarios. The results show that our dual-modal system not only consistently outperforms its single-modal counterparts, demonstrating higher precision in both typical and noisy environments, but also excels in accurately identifying silent utterances. Furthermore, we have successfully applied the system in real-time demonstrations, achieving promising results. The code is available at https://github.com/caizhuojiang/VE-KWS.

cs.HC↗

arXiv2023

Embodied Cognition Guides Virtual-Real Interaction Design to Help Yicheng Flower Drum Intangible Cultural Heritage Dissemination

In order to make the non-heritage culture of Yicheng Flower Drum more relevant to the trend of the digital era and promote its dissemination and inheritance, the design and application of gesture recognition and virtual reality technologies guided by embodied cognition theory in the process of non-heritage culture dissemination is studied. At the same time, it will enhance the interaction between people and NRM culture, stimulate the audience's interest in understanding NRM and spreading NRM, and create awareness of preserving NRM culture. Using embodied cognition as a theoretical guide, expanding the unidirectional communication mode through human-computer interaction close to natural behavior and cooperating with multisensory information reception channels, so as to construct an embodied and immersive interactive atmosphere for the participants and enable them to naturally form the cognition and understanding of the traditional culture in the process of interaction. The dissemination of the non-heritage culture Yicheng Flower Drum can take the theory of embodied cognition as an entry point, and through the virtual and real scenes of Yicheng Flower Drum and the immersive experience, we can empower the interaction design of non-heritage culture dissemination of the virtual and real, and provide a new method for the research of digital design of non-heritage culture.

cs.HC↗

arXiv2023

Sci-CoT: Leveraging Large Language Models for Enhanced Knowledge Distillation in Small Models for Scientific QA

Large Language Models (LLMs) have shown outstanding performance across wide range of downstream tasks. This competency is attributed to their substantial parameter size and pre-training on extensive corpus. Moreover, LLMs have exhibited enhanced reasoning capabilities in tackling complex reasoning tasks, owing to the utilization of a method named ``Chain-of-Thought (CoT) prompting''. This method is designed to generate intermediate reasoning steps that guide the inference of the final answer. However, it is essential to highlight that these advanced reasoning abilities appear to emerge in models with a minimum of 10 billion parameters, thereby limiting its efficacy in situations where computational resources are constrained. In this paper, we investigate the possibility of transferring the reasoning capabilities of LLMs to smaller models via knowledge distillation. Specifically, we propose Sci-CoT, a two-stage framework that separates the processes of generating rationales and inferring answers. This method enables a more efficient use of rationales during the answer inference stage, leading to improved performance on scientific question-answering tasks. Utilizing Sci-CoT, our 80-million parameter model is able to exceed the performance of BLOOM-176B in the ARC-Easy dataset under the few shot setting.

cs.CL↗