Professor Yuan Miao

Professor Yuan Miao received his PhD from Tsinghua University, Automation Department, and became an associate professor within the Department of Computer Science and Technology. His academic career expanded to the University of Melbourne in Australia, Nanyang Technological University in Singapore, and Victoria University (VU) in Australia. He is now a professor in the College of Arts, Business, Law, Education, and IT at VU, and the Head of the Information Technology Program.

Yuan's recent research focuses on Large Language Models (LLMs, Generative AI) and their practical applications. His research has discovered several barriers that affect the practical applications of LLMs in professional processes. For example, limited memory length in ChatGPT, Gemini, etc (thus the inability to cover ongoing operations or a project's lifespan), generated explanations that contradict the conclusions, lack of local knowledge, errors in text and data hybrid reasoning (up to 38% of entries contain errors in complex inferences). His team has developed solutions, such as cognitive map graphs and rational intelligence models, to form customised AI systems integrating LLMs for practical solutions.

Sample research areas applying LLMs:

• Human knowledge modelling with LLMs (fuzzy cognitive map, knowledge graph, decision support)
• Multimodal computer interaction (chatbots)
• Rational LLMs with local knowledge (not fine tuning)
• Data-driven storytelling and its applications in smart systems, social media mining, and sentiment analysis
• Healthcare and active living (active ageing), e.g. dementia detection
• Cybersecurity (e.g. rational compliance, explainable vulnerability analysis)
• Robots powered by rational intelligence (LLMs; physical robots and intelligent software agents)

Yuan has published 138 peer-reviewed papers (ORCID). Two of his articles in fuzzy cognitive map modelling have been among the top 10 most cited works since 2000 in the area (Google Scholar 2000–2016). His team developed an adversarial dataset based on SQuAD 2.0, which led to the accuracy drop of BERT and ELECTRA (2020) from close to 90% to less than 50%. Yuan's research has been awarded grant support from industries, including Microsoft, Amazon, Oracle, and Google, and government funding bodies, including the Australia Research Council, Data61, and Singapore's NRF.

PhD applicants, check the VU website for scholarships, timelines, and documents. For the supervisor form, a 2-3 page proposal is required (including research problems, existing research progress, innovation of the proposed approach, and significance) in an area aligned with our research interests (the application area is flexible based on your choice), and an academic CV. We will meet via Zoom before the supervisor form can be confirmed and provided.