Yilun Hao

Papers

Total Citations

H-Index

About

Yilun Hao is an early-career researcher working at the intersection of imitation learning, reinforcement learning, and robot learning. Their most notable work addresses a fundamental challenge in learning from demonstrations: the assumption that demonstrators and imitating agents share identical dynamics. In their 2021 paper, "Learning Feasibility to Imitate Demonstrators with Different Dynamics," Hao tackles the critical real-world scenario where a learning agent must meaningfully extract and transfer behavior from demonstrations generated under different physical or environmental constraints — a problem highly relevant to sim-to-real transfer and cross-embodiment robot learning. By introducing feasibility estimation into the imitation learning pipeline, this work enables more flexible and robust policy learning when direct mimicry is not possible, broadening the practical applicability of imitation learning frameworks. With 5 citations, the work is gaining early traction within the research community. Hao's contributions speak to growing interest in making learning from demonstrations more generalizable and deployment-ready, positioning their research as a meaningful step toward adaptive autonomous agents capable of learning from imperfect or mismatched demonstration sources.