Mr Yangyang Shu | UNSW Research

Fields of Research (FoR)

Machine learning, Computer vision, Semi- and unsupervised learning, Knowledge representation and reasoning

Biography

Dr Yangyang Shu is an associate lecturer in the School of Systems and Computing at the University of New South Wales, Canberra.

Before joining UNSW, he was a research fellow at the Australian Institute for Machine Learning (AIML), the University of Adelaide and worked on the project - The Centre for Augmented Reasoning (CAR).

His research focuses on building practical and reliable AI systems that are trustworthy, efficient, and adaptive. He...view more

Dr Yangyang Shu is an associate lecturer in the School of Systems and Computing at the University of New South Wales, Canberra.

Before joining UNSW, he was a research fellow at the Australian Institute for Machine Learning (AIML), the University of Adelaide and worked on the project - The Centre for Augmented Reasoning (CAR).

His research focuses on building practical and reliable AI systems that are trustworthy, efficient, and adaptive. He develops machine learning methods that improve how AI systems reason and align with human intentions, enhance the controllability and efficiency of foundation models, and enable models to learn and generalise beyond large-scale supervision. His research spans machine learning, generative AI, computer vision, and multimodal learning, with a particular interest in reliable and data-efficient foundation models. More information about his research can be found on his homepage: https://ganperf.github.io/yangyangshu.github.io/

My Qualifications

Ph.D. in Computer Science, University of Technology Sydney

My Research Activities

My current research program is organised around three interconnected themes:

Learning to Reason and Align . I investigate how AI systems can reason more reliably and better align with human intentions. My work explores rationale-guided learning, counterfactual reasoning, and robust multimodal learning to improve model generalisation, reliability, and alignment.

Efficient and Controllable Foundation Models . I develop methods to make foundation models more controllable and computationally efficient, particularly autoregressive and multimodal generative models. My research explores counterfactual learning, fine-grained generation control, efficient adaptation, and accelerated inference.

Learning Beyond Data . I study how AI systems can learn and generalise effectively under limited supervision, scarce data, and changing environments. This research includes low-supervision learning, domain adaptation, and robust representation learning.