Research & Projects
Research & Projects
Embedded Files
This page curates select publications and projects. See Google Scholar for the full list of publications.
AI Safety and Alignment
AI Safety and Alignment
AI Wellbeing: Measuring and Improving the Functional Pleasure and Pain of AIs: Large language models exhibit measurable, consequential preferences over the experiences they undergo in interaction with users. We formalize this as functional wellbeing and develop multiple independent metrics. [PhilML Workshop at ICML 2026, presented at NISS Ai, Statistics & Data Science Webinar 2026]
AudioLLM for Southeast Asia
AudioLLM for Southeast Asia
MERaLiON-AudioLLM is an open-weights Speech-Text Large Language Model tailored for the multilingual and multicultural landscapes in Singapore and Southeast Asia. MERaLiON-AudioLLM is finetuned on 260,000 hours of speech and audio data. [ACL Demo 2025, Hugging Face 2025]
MoWE-Audio: Multitask AudioLLMs with Mixture of Weak Encoders [ICASSP 2025, first author]
AudioBench: A Universal Benchmark for Audio Large Language Models [NAACL 2025]
Spatial Reasoning with Vision-Language Model
Spatial Reasoning with Vision-Language Model
SPHERE: Unveiling Spatial Blind Spots in Vision-Language Models Through Hierarchical Evaluation (website) [ACL 2025, first author]
Domain Adaptation and Generalization
Domain Adaptation and Generalization
Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias [CVPR 2024, first author]
Source-Free Domain Adaptation Guided by Vision and Vision-Language Pre-Training [IJCV 2024, first author]
Rethinking the Role of Pre-Trained Networks in Source-Free Domain Adaptation [ICCV 2023, first author]
Few-Shot Adaptation of Pre-Trained Networks for Domain Shift [IJCAI 2022, first author]
Rethinking the Role of Pre-Trained Networks in Source-Free Domain Adaptation.pdfFew-Shot Adaptation of Pre-Trained Networks for Domain Shift.pdfTime Series Anomaly Detection
Time Series Anomaly Detection
An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series [TNNLS 2021]
Predictive Classification of Future Operations [Patent 2021]
ABACUS: Unsupervised Multivariate Change Detection via Bayesian Source Separation [SDM 2019, first author]
Pruning and Nonparametric Multiple Change Point Detection [ICDM Workshops 2017, first author]
ABACUS- Unsupervised Multivariate Change Detection via Bayesian Source Separation.pdfPage updated
Google Sites
Report abuse