UBDA Research Seminar - Sustainably Evolving Multi-Agent Systems

You are cordially invited to attend the University Research Facility in Big Data Analytics (UBDA) research seminar to be held on 26 March 2026. UBDA is the first university-level research facility in Big Data Analytics in Hong Kong for cross-disciplinary research collaborations, teaching, and learning, as well as a partnership with industry.

Abstract
This research addresses the limitations of traditional large language models (LLMs) regarding durability, cross-domain transferability, and knowledge retention. Our objective is to advance artificial intelligence from a static "human-data era" into a dynamic "interactive-experience era." To achieve this, we propose an evolutionary agent framework that synergizes the broad cognitive capabilities of large models with the specialized execution skills of small models. This framework is anchored by three core pillars: (1) small-data-driven evolutionary design for complex engineering innovation; (2) the GPU-accelerated EvoX platform for large-scale evolutionary computation; and (3) the Git-compatible EvoGit system for decentralized multi-agent collaboration and continuous knowledge accumulation. Validated across diverse industrial and scientific applications, this approach demonstrates the efficiency, scalability, and practical viability of sustainably evolving intelligent agents.

Bio:
Prof. Cheng is an Associate Professor in the Departments of Data Science and Artificial Intelligence and Computing at The Hong Kong Polytechnic University. His research focuses on the evolvability of AI systems, aiming to build intelligent systems with autonomous learning and continuous evolution capabilities. He is the founder of the open-source project EvoX, a distributed GPU-based platform for evolutionary computation. He has published over 130 papers with 19,000+ Google Scholar citations, received the IEEE CIS Outstanding Early Career Award, and is listed among Clarivate Highly Cited Researchers. He serves as Associate Editor for multiple ACM/IEEE journals and is founding Chair of the IEEE CIS Shenzhen Chapter.

For enquiries, please contact Dr Vincent Ng of UBDA (Email: vincent.ng.comp@polyu.edu.hk).