Proposes DDAVS, which disentangles multi-source audio semantics via a prototype-anchored query mechanism and aligns modalities through delayed bidirectional cross attention, achieving state-of-the-art audio-visual segmentation results.
研究
Research & Publications
Exemplar-Free Continual Learning for State Space Models
CVPR 2026
Highlight
Compute Transparency Champion Award
Proposes Inf-SSM, a geometry-aware regularization method for exemplar-free continual learning in state-space models. The method constrains the long-term evolution of SSM states through the extended observability subspace, reducing catastrophic forgetting while improving performance across sequential tasks.
Self-supervised Learning for Acoustic Few-Shot Classification
ICASSP 2025
Oral
Studies self-supervised pre-training for acoustic few-shot classification, where labelled audio data are scarce. The work combines CNN-based feature extractor with SSM, and utilizes contrastive learning to improve acoustic representations for few-shot classification settings.
学历
Education
Doctor of Philosophy (PhD)
National University of Singapore · School of Computing
Supervisor: Prof. Mohan Kankanhalli
2026 – Present
B.Eng. Electrical & Computer Systems Engineering (Hons)
Monash University · Melbourne, Australia
University Medal
WAM 93.05
GPA 4.0
2021 – 2024
经历
Research Experience
PhD Student
National University of Singapore · School of Computing
Present
Commenced doctoral research at the NUS School of Computing under the supervision of Prof. Mohan Kankanhalli, focusing on machine learning, continual learning, and test-time adaptation.
Visiting Research Student
Tsinghua Shenzhen International Graduate School · Tsinghua University
2025
Worked as a visiting student in A/Prof. Yansong Tang's group on audio-visual segmentation research.
Affiliate Researcher
Monash University · Melbourne, Australia
2025 – Present
Worked on continual learning and test-time adaptation research with A/Prof. Mehrtash Harandi, Dr. Trung Le, Dr. Jing Wu, and Dr. Leila Mahmoodi.
教学
Teaching
Teaching Assistant · CS4248 Introduction to Large Language Models
National University of Singapore · School of Computing
AY2026/27 Sem 1
Teaching assistant for CS4248, under Prof. Hwee Tou Ng.
散文
Writing
《梦蝶归乡》
星云 · 2020
故乡
Simanggang, Sarawak
Simanggang is a small town about a three-hour drive from Kuching, known for the Batang Lupar River and its spectacular tidal bore — locally called Benak — regarded as one of the world’s most remarkable tidal bore phenomena.
The town preserves colonial-era forts, shophouses, and heritage buildings, while the surrounding rainforest is home to saltwater crocodiles, orangutans, racket-tailed drongos, greater argus pheasants, fireflies, and more.
联系
Contact
Open to research collaborations, academic discussion, and opportunities in AI and ML.