About Me
I am a second-year Computer Science PhD student at the University of Wisconsin-Madison, advised by Prof. Junjie Hu, and I'm fortunate to be supported by the NSF Graduate Research Fellowship (GRFP). I previously earned my bachelor's degree in Computer Science from the University of Massachusetts Amherst.
Research Interests
I study self-improving agentic systems, and interested in how agents can learn from experience, adapt their reasoning, and improve over time in open-ended settings.
- Open-ended, evolutionary search: discovering and refining strategies through iterative search
- Recursive self-improvement: improving an agent's own reasoning and learning processes
- Continual learning & memory: accumulating and reusing knowledge across interactions
- Agent safety & reliability: maintaining robust behavior as agents become more adaptive
Publications
2026
When Safety Fails Before the Answer: Benchmarking Harmful Behavior Detection in Reasoning Chains
We introduce a step-level taxonomy and an annotated dataset HarmThoughts, that traces how harm emerges and propagates through a model's reasoning.
2026
Fluent but Unfeeling: The Emotional Blind Spots of Language Models
We introduce EXPRESS, a benchmark of 251 fine-grained, self-disclosed emotion labels from Reddit, and show that even strong LLMs struggle to match how people actually describe their own emotions.
Awards
Personal
I love graphic design and calligraphy! Check out my work here.