Research Interest. My current research focus is to develop theory and algorithms for reliable machine learning (ML), foundation model reliability, and related applications, specifically I work on:
Algorithmic reliability: designing adaptive and interpretable learning principles that help ML models detect and generalize on out-of-distribution (OOD) samples, such as semantic/covariate shifts, adversarial and noisy samples.
Language model reliability: understanding the blindspots of Large Language Models (LLMs) and Multimodal LLMs through uncertainty estimation, such as model hallucination detection and mitigation, malicious prompt prevention, etc.
[Radio Lab Recruiting]. I am looking for self-motivated Research Assistants, Visiting Students and Interns in the fall! Please check the Recruitment Page for details. For PhD applicants, the NTU CCDS recruits students at Spring and Fall semesters per year, please mind the deadlines (e.g. 7/31 for Spring'26 intake and 1/31 for Fall'26 intake) and check the website for more information, directly apply here and mention my name in the system. For the other positions, there are no strict deadlines but depend on the Lab's availability.
Before you email me, I strongly suggest reading my Advising Statement to get to know me and the Lab.
Due to the large volume of interest and limited availability, I am currently prioritizing applicants with strong ML research experience, ideally with publications or relevant project experience in theory and algorithms for LLM safety. Please direct all emails to radiolab.ntu.recruiting (at) gmail.com!
News
[11/25/2025] Our lab receives the Google Cloud Research Credits and NSCC Young Investigator Seed Project Award.