Lin Li

Lin Li is a Researcher in the Department of Computer Science at the University of Oxford, where he conducts postdoctoral research in the Oxford Applied and Theoretical Machine Learning (OATML) Group under Professor Yarin Gal's supervision. His work bridges theoretical advances in AI safety with real-world applications across healthcare, robotics, and economics, positioning him at the forefront of robust and trustworthy artificial intelligence development.

His academic journey includes:

• PhD in Machine Learning from King’s College London (supervised by Prof. Michael Spratling), focusing on model robustness
• MSc in Computing from Imperial College London (advised by Prof. Wayne Luk)
• Bachelor of Music in Finance from Xiamen University (advised by Prof. Zheng Qiao)

Li's research centers on critical AI safety challenges including hallucination mitigation, jailbreaking defenses, and safety alignment in large language models. He pioneers LLM-based agent systems for complex decision-making and societal simulation while developing practical applications in healthcare diagnostics, robotic dexterity, and economic modeling. His work uniquely combines adversarial machine learning techniques with domain-specific implementations to address real-world deployment challenges.

Analysis of his publication record reveals a strategic evolution from foundational adversarial robustness research toward safety-critical applications. Recent work demonstrates increasing emphasis on healthcare domains—particularly women's health and ophthalmology—while maintaining strong contributions to core AI robustness through novel benchmarks like OODRobustBench and techniques such as semantic entropy regularization. His publications consistently bridge theoretical innovation with practical implementation across computer vision, natural language processing, and multimodal systems.

Recognition includes:

• USERN Prize nomination (2025) for early-career research excellence

Li actively shapes the field through program committee roles for MICCAI 2025, CVPR 2025, and AAAI 2025, while reviewing for top venues including ICML, NeurIPS, and IEEE Transactions journals. He actively seeks motivated students for collaborative research in AI safety and robustness through the OATML Group, which provides interdisciplinary infrastructure for theoretical and applied machine learning research with strong industry partnerships including Tencent's Robotics X Lab.