1. QLASS: Boosting Language Agent Inference via Q-Guided Stepwise Search - ICML 2025
2. Learning Versatile Skills with Curriculum Masking - NeurIPS 2024
Research Experience
Research experiences mainly lie in the intersection of reinforcement learning and large language models. Projects include: Unsupervised Pre-training for Reinforcement Learning, Language Agent Self-improvement, Conditional Generative Modeling for Decision Making.
Education
Received a bachelor's degree from Shanghai Jiao Tong University (SJTU). Worked at Microsoft Research Asia in 2025, mentored by Dr. Li Dong. Spent a summer at UCLA NLP in 2024, advised by Prof. Kai-Wei Chang and collaborated with Johnson Lin.
Background
Research Interests: Agentic reasoning, reinforcement learning; Professional Field: Artificial Intelligence; Brief Introduction: A first-year CIS PhD student at the University of Pennsylvania, advised by Prof. Jiatao Gu.