Australian Institute for Machine Learning
University of Adelaide
I am a third-year Ph.D. student in Computer Science at the Australian Institute for Machine Learning (AIML), University of Adelaide, advised by Prof. Javen Qinfeng Shi. My research studies multimodal representation learning, with a particular focus on contrastive learning theory, cross-modal alignment, and identifiable causal representations.
More broadly, I am interested in how language supervision shapes semantic structure in vision-language models, and how this perspective can support interpretable, reliable, and human-aligned AI systems. My work combines theoretical analysis with empirical study to better understand representation formation in modern multimodal models.
Research interests: multimodal learning, contrastive learning theory, identifiability, causal representation learning, and vision-language models.
News
Education
-
Sep 2023 - PresentUniversity of Adelaide
Ph.D. in Computer Science
Advisor: Prof. Javen Qinfeng Shi -
Sep 2012 - Jun 2019Wuhan University of Technology
M.Sc. in Instrument Science and Technology, Sep 2016 - Jun 2019
Advisor: Prof. Xiao Zhou
B.Eng. in Measurement and Control Engineering, Sep 2012 - Jun 2016
Experience
-
2025 - presentUniversity of Adelaide
Teaching Assistant - Neural Networks and Deep Learning (ARTI X300), Semester 1 2026
Guest Lecturer and Head Tutor - Statistical Machine Learning (COMP SCI 3314), Semester 2 2025
Teaching Assistant - Using Machine Learning Tools (COMP SCI 7317), Trimester 2 2025
Teaching Assistant - Concepts in AI and ML (COMP SCI 7327), Semester 1 2025 -
May 2018 - Oct 2018California PATH, UC Berkeley
Visiting Student Researcher, under Dr. Ching-Yao Chan
Honors & Awards
-
NeurIPS Scholar Award2025
-
University of Adelaide Research Scholarships2023
-
Award for Outstanding Graduates, Wuhan University of Technology2019
Academic Service
-
Reviewer for TMLR, ICLR 2026, and ICML 20262025 - 2026
Selected Publications (view all )
The Geometric Mechanics of Contrastive Learning: Alignment Potentials, Entropic Dispersion, and Modality Gap
Yichao Cai; Zhen Zhang; Yuhang Liu; Javen Q. Shi.
arXiv preprint 2026
A theoretical study of contrastive learning geometry, alignment forces, dispersion, and the emergence of modality gap.
The Geometric Mechanics of Contrastive Learning: Alignment Potentials, Entropic Dispersion, and Modality Gap
Yichao Cai; Zhen Zhang; Yuhang Liu; Javen Q. Shi.
A theoretical study of contrastive learning geometry, alignment forces, dispersion, and the emergence of modality gap.
I Predict Therefore I Am: Is Next Token Prediction Enough to Learn Human-Interpretable Concepts from Data?
Yuhang Liu; Dong Gong; Yichao Cai; Erdun Gao; Zhen Zhang; Biwei Huang; Mingming Gong; Anton van den Hengel; Javen Q. Shi.
International Conference on Learning Representations (ICLR) 2026
An investigation of whether next-token prediction alone is sufficient for learning human-interpretable concepts from data.
I Predict Therefore I Am: Is Next Token Prediction Enough to Learn Human-Interpretable Concepts from Data?
Yuhang Liu; Dong Gong; Yichao Cai; Erdun Gao; Zhen Zhang; Biwei Huang; Mingming Gong; Anton van den Hengel; Javen Q. Shi.
An investigation of whether next-token prediction alone is sufficient for learning human-interpretable concepts from data.
On the Value of Cross-Modal Misalignment in Multimodal Representation Learning
Yichao Cai*; Yuhang Liu*; Erdun Gao; Tianjiao Jiang; Zhen Zhang; Anton van den Hengel; Javen Q. Shi. (* equal contribution)
Advances in Neural Information Processing Systems (NeurIPS) 2025 Spotlight
Studies when controlled cross-modal misalignment can improve multimodal representation learning instead of only harming it.
On the Value of Cross-Modal Misalignment in Multimodal Representation Learning
Yichao Cai*; Yuhang Liu*; Erdun Gao; Tianjiao Jiang; Zhen Zhang; Anton van den Hengel; Javen Q. Shi. (* equal contribution)
Spotlight
Studies when controlled cross-modal misalignment can improve multimodal representation learning instead of only harming it.
CLAP: Isolating Content from Style through Contrastive Learning with Augmented Prompts
Yichao Cai; Yuhang Liu; Zhen Zhang; Javen Q. Shi.
European Conference on Computer Vision (ECCV) 2024
Explores language-guided disentanglement of style and content through contrastive learning with augmented prompts.
CLAP: Isolating Content from Style through Contrastive Learning with Augmented Prompts
Yichao Cai; Yuhang Liu; Zhen Zhang; Javen Q. Shi.
Explores language-guided disentanglement of style and content through contrastive learning with augmented prompts.