Abstract:
Large language models are increasingly used to support scientific research, but challenges remain in developing reliable AI research partners. Scientific literature exploration needs to move beyond semantic similarity, since meaningful discoveries often emerge from connecting seemingly unrelated work through shared problems or transferable methods. Novelty assessment requires grounding scientific claims in structured evidence and prior work. Research reproducibility is hindered by incomplete descriptions and omissions of implementation details in papers. In this talk, I will discuss our recent work addressing these challenges, including graph-distilled generative retrieval enabling literature discovery beyond similarity-based search; bi-level graph reasoning for evidence-grounded novelty assessment; and agent-based reproducibility evaluation revealing gaps between algorithm understanding and reproduction. Finally, I will introduce our ongoing work on a personal AI research assistant that maintains an evolving graph of research questions, hypotheses, and experimental evidence to support human-guided idea refinement, experimental planning, evaluation, and scientific writing. Our ultimate goal is to move beyond isolated AI tools towards sustained human-AI collaboration in scientific discovery.
Bio:
Professor Yulan He is a Professor in Natural Language Processing in the Department of Informatics at King’s College London. Her research focuses on machine learning and natural language processing for text understanding, with particular interests in improving the robustness of large language model reasoning, agentic AI, model interpretability, trustworthy AI, and the applications of AI for in science, education, and health. Her five-year UKRI Turing AI Fellowship project was awarded , “Event-Centric Framework for Natural Language Understanding”, received the Best Research Project (Research Excellence) Award at the RAi UK AI & Robotics Awards 2026, where she also received a Best Research Paper Award. Her other honours include the SWSA Ten-Year Award and the CIKM Test-of-Time Award. She earned her PhD from the University of Cambridge, and her MEng and BASc degrees from Nanyang Technological University, Singapore.
There will be a networking opportunity following this talk. If you would like to attend, please complete the registration form. Registration will close on Wednesday 18 November at 12.00.