Chi-Chang Lee

I earned both my Bachelor’s and Master’s degrees in Engineering Science and Computer Science from National Taiwan University, where I worked closely with Prof. Yu Tsao and Prof. Hsin-Min Wang. Before 2023, my research primarily centered on machine learning for audio and speech modalities, especially Speech-to-Text systems and task-aware front-end processing for downstream audio applications. I focused on improving robustness to diverse acoustic environments (ICLR'23 & ASRU'23) and on how such robustness can be continually learned and adapted over time (Interspeech'20 and Interspeech'22).

In 2023, I have been honored to begin collaborating with the Improbable AI Lab at MIT CSAIL, led by Prof. Pulkit Agrawal, and shifted my primary research focus to Embodied AI, particularly Reinforcement Learning (RL). My current work investigates how, under a finite interaction budget, a robotic agent can learn to distinguish heuristic signals from the true task objective while still optimizing for the real but sparse success outcomes. This collaboration has led to two papers: one on the real-world deployment of locomotion policies (ICRA'24), and another on an advanced RL algorithm that achieves consistent gains on 31 robotic tasks (NeurIPS'24).

template from jonbarron