About Me
I am an Assistant Professor at the Hong Kong University of Science and Technology (HKUST) and Head of the Large Language Model Division at the Hong Kong Generative AI Research and Development Centre (HKGAI). Trained as both a legal scholar (Ph.D. in Law, CUHK) and a computer scientist (Ph.D. candidate in CS, HKUST), I am one of the few researchers globally who combines direct engineering access to a sovereign LLM with formal doctoral training in both disciplines. My research focuses on building AI systems that are legally aligned, ethically accountable, and technically robust — with a particular emphasis on Hong Kong’s unique dual legal system as a natural laboratory for plural-norm AI governance.
我是香港科技大学助理教授，同时担任香港生成式人工智能研发中心（HKGAI）大语言模型部门主管。我同时接受过法学（香港中文大学法学博士）和计算机科学（香港科技大学计算机博士候选人）的正式博士训练，是全球少数兼具主权大语言模型直接工程经验与跨学科博士背景的研究者。我的研究聚焦于构建合法合规、伦理可问责、技术稳健的AI系统——尤其以香港独特的双轨法律体系作为多元规范AI治理的天然实验室。
科研经费
顶会论文
博士生
产业合作伙伴
Research Vision
As AI systems increasingly operate within and across legal jurisdictions, the question of how to align these systems with diverse — sometimes conflicting — legal norms becomes urgent. My research program addresses this challenge at the intersection of AI and law, developing the theoretical foundations, technical methods, and governance frameworks needed to ensure that large language models can be trusted to operate within the rule of law. From AI safety and RLHF to legal reasoning and multimodal intelligence, my group pursues a unified vision: trustworthy AI that respects human values and legal boundaries.
随着AI系统日益在不同法律管辖区内及跨区运行，如何使其与多元甚至相互冲突的法律规范保持一致已成为紧迫课题。我的研究聚焦于人工智能与法律的交叉领域，致力于发展理论基础、技术方法和治理框架，以确保大语言模型能够在法治框架内可信运行。从基于人类反馈的强化学习（RLHF）与AI安全到法律推理和多模态智能，我的团队追求统一愿景：构建尊重人类价值与法律边界的可信赖AI。
Research Pillars
Legal Alignment & AI Governance
Building AI systems aligned with plural legal norms. Key projects include SafeLawBench, LegalReasoner, PrivaCI-Bench, and the Trustworthy Legal Reasoning survey.
Learn More →AI Safety & RLHF
Developing safe reinforcement learning from human feedback for multimodal LLMs. Key projects include Safe RLHF-V, PKU-SafeRLHF, SafeMT, and AM3Safety.
Learn More →Efficient & Robust LLMs
Advancing model compression, efficient fine-tuning, and inference optimization. Key projects include Sub-MoE, Nested LoRA, AIRA, and Outlier-Aware Merging.
Learn More →Multimodal Intelligence & Embodied AI
Pushing the boundaries of vision-language models, 3D understanding, and robotic manipulation. Key projects include ManipDreamer3D, GSRender, Motion-R1, and EgoTwin.
Learn More →Computational Law & Social Science
Bridging computational methods with legal and social science inquiry, including AI manipulation, deepfake governance, and corpus inequality studies.
Learn More →
Featured Publications
SafeLawBench: Towards Safe Alignment of Large Language Models
A comprehensive benchmark for evaluating the legal safety alignment of LLMs, bridging AI safety research with legal compliance requirements.
Safe RLHF-V: Safe Reinforcement Learning from Multi-modal Human Feedback
Extending safe RLHF to the multimodal domain, enabling alignment of vision-language models with human safety preferences.
Sub-MoE: Efficient Mixture-of-Expert LLMs Compression via Subspace Expert Merging
A novel approach to compressing Mixture-of-Expert LLMs through subspace expert merging, achieving significant efficiency gains.
Reimagining Legal Fact Verification with GenAI: Toward Effective Human-AI Collaboration
Pioneering the use of generative AI for legal fact verification, demonstrating effective human-AI collaboration in the legal domain.
Motion-R1: Enhancing Motion Generation with Decomposed Chain-of-Thought and RL Binding
A novel framework that enhances motion generation quality through decomposed chain-of-thought reasoning and reinforcement learning.
Recent News
Team
I lead an active research group of 30+ PhD students, 7 engineers, and research assistants spanning AI safety, legal reasoning, efficient LLMs, and multimodal systems.
PhD Students
Chuxue CAO · Jialiang CHEN · Long CHEN · Xiaowei CHI (co-supervision) · Chi-Min CHAN · Yakun CUI · Mengfei DU · Hao GU · Shengyao GUO · Chengyi JU · Qintao HUANG (co-supervision) · Yidan HUANG · Zhizhuo KOU · Lujun LI (co-supervision) · Ruoxi LI · Xinyao LIAO · Guoying LU · Siyu PENG · Xingqun QI (co-supervision) · Jiapeng SUN · Pengcheng WEN · Jian YANG (co-supervision) · Yanting ZHANG · Xiaojie ZHANG · Yuyao ZHANG · Zhenyuan ZHANG · Yujin ZHOU · Han ZHU · Qiyuan ZHU · Junqi ZHU
Engineers
Jieyang FENG · Xiangyu LONG · Zhiqiang QIAN · Xinyi WANG · Shuchao WU · Boqin YIN · Zhenghao ZHU
Research Assistants
Xueyan LI
Institutional Collaborators
BAAI
ByteDance
China Telecom International
Du Xiaoman Financial
HKGAI
MetaX
Peng Cheng Laboratory
Phoenix TV
Shanghai AI Laboratory
Tencent Hunyuan
Tsinghua I-AIIG
WeBank
WiseLaw