Learning from evolving data
Detecting and adapting to concept drift across single and multiple data streams for dependable real-time prediction.
Trustworthy machine learning · Streaming data
Yiliao (Lia) Song
Lecturer in Artificial Intelligence at Adelaide University
I develop trustworthy machine learning methods for data that changes over time. My work spans concept drift, robust real-time prediction, federated learning, and the reliability of AI-generated content and agentic systems.
Research
Learning reliably in a changing world
My research asks how AI systems can remain accurate, robust, and worthy of trust when their data, users, and environments do not stand still.
Trustworthy AI systems
Building robust, privacy-aware, and explainable learning systems for high-stakes and non-stationary settings.
Reliable generative AI
Studying machine-generated text detection, federated learning, and whether dependencies in agentic workflows can be trusted.
Latest
News & highlights
A selection of recent awards, projects, publications, and invited talks.
Preprints
New work on reliable and collaborative AI agents
Memory Control Flow Attacks on LLM Agents and Vision-Sharing Collaboration for Vision-Language Navigation.
Invited talk
Adelaide–Vietnam Electricity Roundtable
Presented “Multiple-Step Ahead Wind Energy Forecast.”
Award
Australasian AI Award
Received the Emerging Research Contributor award from the Australian Computer Society.
Grant
ARC Discovery Project
Chief Investigator on “Trustworthy Model Reprogramming: Learning with Imperfect Pre-trained Models.”
Invited talk
Technical University of Munich
Presented “When Machine Learning Meets User Data: Challenges Beyond the I.i.d. Assumption.”
Publication
Paper accepted at NeurIPS 2025
“Can Dependencies Induced by LLM-Agent Workflows Be Trusted?”
Publications
Papers accepted at ACL and KDD 2025
New work on multimodal metaphor understanding and one-shot federated learning.
More updates from 2025
- “A Multistream Concept Drift Handling Framework via Data Sharing” accepted by IEEE Transactions on Cybernetics.
- Invited to present at the AGS SA/NT Symposium 2025.
- Received the School of Computer and Mathematical Sciences inter-department award.
- “Machine Learning in the Australian Equity Market” accepted by Pacific-Basin Finance Journal.
- Awarded a DAAD AINeT Fellowship.
- Selected for the DAAD Postdoc-NeT-AI Virtual Networking Week on Natural Language Processing.
- Paper published in Intelligent Systems with Applications.
- Paper accepted by IEEE Transactions on Information Forensics and Security.
- Two papers accepted at ICLR 2025: FLOW and R-detect.
About
Biography
I am a Lecturer in Artificial Intelligence in the School of Computer Science and Information Technology at Adelaide University, and a researcher at the Australian Institute for Machine Learning. I am also a Visiting Scholar at the Australian Artificial Intelligence Institute, University of Technology Sydney.
I received my PhD in Computer Science from UTS in 2020, supervised by Distinguished Professor Jie Lu AO, Associate Professor Haiyan Lu, and Associate Professor Guangquan Zhang. Before joining Adelaide, I was a Research Fellow at RMIT University and an Australian Laureate Postdoctoral Fellow at UTS.
2024–presentLecturer, Adelaide University
2022–2023Research Fellow, RMIT University
2020PhD, University of Technology Sydney