LLM Safety Alignment Center
Project Description
Research achievements in AI alignment by the team led by Researcher Yang Yaodong, especially the RICE (Robustness, Interpretability, Controllability and Ethicality) principle, HHH (Helpful, Honest, Harmless) standard, BeaverTails dataset, Safe RLHF algorithm, and the Aligner, not only lead the industry in theory but also demonstrate their outstanding practical value. These research achievements greatly strengthen the company's leading position in AI safety and value alignment and bring about a qualitative leap in the overall competitiveness of the center.
Especially for small and medium-sized enterprises and key enterprises, the LLM (Large Language Model) technology developed by the center plays a significant role in improving operational efficiency and business security. By introducing advanced LLM technology, enterprises can achieve automation and intelligence in multiple aspects such as data analysis, customer service, and product innovation, greatly improving work efficiency and decision-making quality. At the same time, the ability of LLM technology to identify potential risks and prevent security threats provides solid technical support for enterprises to protect critical data and maintain business security.
In addition, in key sectors such as education, healthcare, and finance, the unique value of LLM technology developed by the center has begun to be realized. In education, through personalized learning programs and intelligent tutoring systems, LLM technology is changing traditional teaching methods and providing students with learning experiences that better fit their individual needs. In healthcare, the application of LLM technology not only improves the accuracy of disease diagnosis and the specificity of treatment plans but also shows great potential in epidemic prediction, drug development, and other areas. In finance, through advanced risk management and intelligent investment advisory services, LLM technology is helping enterprises enhance their risk control capabilities, optimize resource allocation, and thereby improve the market competitiveness.
Lead Arranger / Chief Scientist
Doctor Yang Yaodong, Researcher and Doctoral Supervisor of the Institute for Artificial Intelligence at Peking University, Executive Director of Center for AI Safety and Governance, Recipient of Funding Scheme for High-level Overseas Chinese Students' Return, National High-level Young Talent Program, Young Elite Scientist Sponsorship Program (YESS) of The China Association for Science and Technology (CAST), and Peking University Boya Young Fellow. He specializes in the construction of general multi-agent systems, game interactions, and value alignment issues, with research areas including reinforcement learning, game theory, and multi-agent systems. He graduated from the University of Science and Technology of China with a bachelor's degree, and subsequently obtained master's and doctoral degrees from Imperial College London and University College London (his thesis was nominated for the ACM SIGAI Doctoral Dissertation Award). He has served as an assistant professor at the Department of Informatics at King's College London. He has published over a hundred papers in top conferences and journals in AI, with over four thousand citations on Google Scholar. He has secured funding exceeding 30 million RMB for projects from the National Natural Science Foundation of China, the Ministry of Science and Technology of the People's Republic of China, Municipal Science and Technology Commissions, and school-enterprise labs. He was a finalist for the Best Paper Award at the International Conference on Computer Vision (ICCV) 2023, received the Best System Paper Award at the Conference on Robot Learning (CoRL) 2020, won the Best Blue-Sky Paper Award at the International Conference on Autonomous Agents and Multiagent Systems (AAMAS) 2021 and Yunfan Award at the World Artificial Intelligence Conference (WAIC) 2022, and received the ACM SIGAI China Rising Star Award. His work has been featured on "Focus Interview"（《焦点访谈》）on CCTV-1 and "Deep International"（《深度国际》）on CCTV-4.
