arXiv subjects

Yanqing Li

Publications and source records attributed to Yanqing Li.

4 recordsLinked to original sources

arXiv2026

Automated jailbreak attack targeting multiple defense strategies

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks. However, their safety remains a critical concern due to their susceptibility to adversarial prompt-based attacks. In this paper, we present UNIATTACK, an adversarial testing framework designed from a defense-oriented perspective to systematically construct effective black-box attack prompts. Unlike prior approaches that rely on static templates or iterative model-specific tuning, UNIATTACK extracts minimal but high-impact attack features from diverse existing attacks, optimizes them via a specialized attacker LLM, and composes them into flexible templates through automated refinement process. This feature-centric construction enables one-shot attacks that generalize across multiple models and safety categories, providing a practical tool for assessing LLM robustness. Our evaluation results shows that compared to the baselines, UNIATTACK achieves an average attack success rate (ASR) improvement of 64.63\%-248.82\% on models deployed with multi-layered defense mechanisms and it only takes 0.03\%-4.96\% cost of the baselines. UNIATTACK artifact is available at https://anonymous.4open.science/r/UniAttack-Artifact-30F1.

cs.CR↗

arXiv2026

MagicAgent: Towards Generalized Agent Planning

The evolution of Large Language Models (LLMs) from passive text processors to autonomous agents has established planning as a core component of modern intelligence. However, achieving generalized planning remains elusive, not only by the scarcity of high-quality interaction data but also by inherent conflicts across heterogeneous planning tasks. These challenges result in models that excel at isolated tasks yet struggle to generalize, while existing multi-task training attempts suffer from gradient interference. In this paper, we present \textbf{MagicAgent}, a series of foundation models specifically designed for generalized agent planning. We introduce a lightweight and scalable synthetic data framework that generates high-quality trajectories across diverse planning tasks, including hierarchical task decomposition, tool-augmented planning, multi-constraint scheduling, procedural logic orchestration, and long-horizon tool execution. To mitigate training conflicts, we propose a two-stage training paradigm comprising supervised fine-tuning followed by multi-objective reinforcement learning over both static datasets and dynamic environments. Empirical results show that MagicAgent-32B and MagicAgent-30B-A3B achieve superior performance across diverse open-source benchmarks (\emph{e.g.}, $75.1\%$ on Worfbench and $86.9\%$ on BFCL-v3), as well as strong results on our in-house MagicEval benchmarks, substantially outperforming existing sub-100B models and surpassing leading ultra-scale models, including GPT-5.2, Kimi-K2 and GLM-4.7.

cs.AI↗

arXiv2022

A stochastic agent-based model to evaluate COVID-19 transmission influenced by human mobility

The COVID-19 pandemic has created an urgent need for mathematical models that can project epidemic trends and evaluate the effectiveness of mitigation strategies. To forecast the transmission of COVID-19, a major challenge is the accurate assessment of the multi-scale human mobility and how they impact the infection through close contacts. By combining the stochastic agent-based modeling strategy and hierarchical structures of spatial containers corresponding to the notion of places in geography, this study proposes a novel model, Mob-Cov, to study the impact of human traveling behaviour and individual health conditions on the disease outbreak and the probability of zero COVID in the population. Specifically, individuals perform power-law type of local movements within a container and global transport between different-level containers. Frequent short movements inside a small-level container (e.g. a road or a county) and a large population size influence the local crowdedness of people, which accelerates the infection and regional transmission. Travels between large-level containers (e.g. cities and nations) facilitate global spread and outbreak. Moreover, dynamic infection and recovery in the population are able to drive the bifurcation of the system to a "zero-COVID" state or a "live with COVID" state, depending on the mobility patterns, population number and health conditions. Reducing total population and local people accumulation as well as restricting global travels help achieve zero-COVID. In summary, the Mob-Cov model considers more realistic human mobility in a wide range of spatial scales, and has been designed with equal emphasis on performance, low simulation cost, accuracy, ease of use and flexibility. It is a useful tool for researchers and politicians to investigate the pandemic dynamics and plan actions against the disease.

physics.soc-ph↗

arXiv2022

A Miniature 3-DoF Flexible Parallel Robotic Wrist Using NiTi Wires for Gastrointestinal Endoscopic Surgery

Gastrointestinal endoscopic surgery (GES) has high requirements for instruments' size and distal dexterity, because of the narrow endoscopic channel and long, tortuous human gastrointestinal tract. This paper utilized Nickel-Titanium (NiTi) wires to develop a miniature 3-DoF (pitch-yaw-translation) flexible parallel robotic wrist (FPRW). Additionally, we assembled an electric knife on the wrist's connection interface and then teleoperated it to perform an endoscopic submucosal dissection (ESD) on porcine stomachs. The effective performance in each ESD workflow proves that the designed FPRW has sufficient workspace, high distal dexterity, and high positioning accuracy.

cs.RO↗