A5091386446- Manufacturing Process and Optimization
- Robot Manipulation and Learning
- Robotic Path Planning Algorithms
- Digital Rights Management and Security
- Scientific Computing and Data Management
- Computability, Logic, AI Algorithms
- Graphene research and applications
- Anomaly Detection Techniques and Applications
- Adversarial Robustness in Machine Learning
- Modular Robots and Swarm Intelligence
- Brain Tumor Detection and Classification
- Access Control and Trust
- Surface and Thin Film Phenomena
- Digital and Cyber Forensics
- Advanced Manufacturing and Logistics Optimization
- Virus-based gene therapy research
- Multi-Agent Systems and Negotiation
- Web Application Security Vulnerabilities
- Cell Image Analysis Techniques
- Topological Materials and Phenomena
- Medical Imaging Techniques and Applications
Osaka University
2023-2025
Southern University of Science and Technology
2024
National Yang Ming Chiao Tung University
2006
Scientific discovery is poised for rapid advancement through advanced robotics and artificial intelligence. Current scientific practices face substantial limitations as manual experimentation remains time-consuming resource-intensive, while multidisciplinary research demands knowledge integration beyond individual researchers' expertise boundaries. Here, we envision an autonomous generalist scientist (AGS) concept combines agentic AI embodied to automate the entire lifecycle. This system...
Although LLM-based agents, powered by Large Language Models (LLMs), can use external tools and memory mechanisms to solve complex real-world tasks, they may also introduce critical security vulnerabilities. However, the existing literature does not comprehensively evaluate attacks defenses against agents. To address this, we Agent Security Bench (ASB), a comprehensive framework designed formalize, benchmark, of including 10 scenarios (e.g., e-commerce, autonomous driving, finance), agents...
Self-supervised learning models are vulnerable to backdoor attacks. Existing attacks that effective in self-supervised often involve noticeable triggers, like colored patches, which human inspection. In this paper, we propose an imperceptible and attack against models. We first find existing triggers designed for supervised not as compromising then identify ineffectiveness is attributed the overlap distributions between augmented samples used learning. Building on insight, design using...
Recent advances in code-specific large language models (LLMs) have greatly enhanced code generation and refinement capabilities. However, the safety of LLMs remains under-explored, posing potential risks as insecure generated by these may introduce vulnerabilities into real-world systems. Previous work proposes to collect security-focused instruction-tuning dataset from vulnerabilities. It is constrained data sparsity vulnerable code, has limited applicability iterative post-training...
To enhance the accuracy of robotic assembly planning by understanding graphical instruction manual, this paper proposes a novel two-step error correction method. While constructing Assembly Task Sequence Graph (ATSG) from we performed an focusing on component, symbol, speech bubble, and model number included in manual. The component symbol information were used to check correctness manipulated components, needed motion tool single step task. bubble was remove repeated drawn components...