Graduate Researcher — NTU Robot Learning Lab
- Conducted research on programmatic reinforcement learning, investigating how program code can serve as a structured and generalizable representation for control policies.
- Published our work, Hierarchical Programmatic Option Framework, at NeurIPS 2024.
- Worked on imitation learning from suboptimal demonstrations, focusing on richer supervision signals beyond scalar rewards or weights.
- Proposed using natural language as an expressive supervision signal for imitation learning in embodied control.
- Explored diffusion steering through high-level noise prediction with in-context reinforcement learning.