- Papers accepted at NeurIPS 2025, EMNLP 2025, ICML 2025, and other top conferences.
- Research achievements span understanding and mitigating memorization in generative models, low-bit quantized neural networks, and more.
Research Experience
Involved in multiple research projects, covering but not limited to preventing harmful reasoning, an information-theoretic approach to discrete diffusion, unlearning in large reasoning models, etc.
Background
Research interests include deep learning and the application of information-theoretic approaches to various machine learning problems.