Yao Tong

I am Yao Tong (童遥), a Ph.D. student in Computer Science at the National University of Singapore, supervised by Prof. Reza Shokri. I am currently a visiting researcher at ETH Zurich (SPY Lab), working with Prof. Florian Tramer. I received my B.E. in Computer Science and Engineering from The Chinese University of Hong Kong, Shenzhen, where I was advised by Prof. Baoyuan Wu, who inspired my first steps into research.

My research focuses on understanding the behaviors and capabilities of large language models (LLMs), the underlying mechanisms that give rise to them, and how these properties relate to and shape model trustworthiness.

• Model behaviors and capabilities. I study phenomena such as intrinsic biases and generalization, and how they are shaped by different stages of the learning pipeline, including model initialization, data distributions, training paradigms (e.g., SFT and RL), and inference-time methods.
• Trustworthiness. I study how these properties relate to model trustworthiness from two complementary perspectives: (i) as sources of security and privacy risks, and (ii) as signals that can be leveraged for detection and protection, such as model fingerprinting and data usage auditing.

More recently, I have become interested in agentic memory and personalization systems, particularly in how long-term memory and adaptive interaction introduce new failure modes, privacy risks, and auditing challenges.