I am a final year Ph.D. candidate at the University of Southern California coadvised by Erdem Bıyık (LiRA Lab) and Stephen Tu. I am interested in efficient robot learning from various forms of natural human feedback, such as gaze, hand demonstrations, videos, and preferences. I am also interested in reinforcement learning for efficient adaptation of agents to new tasks and environments. I had the pleasure to intern at Google Gemini to work on post-training RL for reasoning models and with Google Research working on meta-RL on dynamic long-horizon MDPs. I also interned at Meta Reality Lab building task-oriented dialogue systems using LLMs.
Prior to joining USC, I spent a summer (during the pandemic) as a visiting scholar at CMU working with Changliu Liu. I graduated from the University of Michigan with a Masters in Robotics and Bachelors in Computer Science where I worked with Wilka Carvalho and Honglak Lee. I had the pleasure to intern at Amazon, Invisible.ai, Google Ads, Luminar and Socratic.
Feel free to say hi: anthony dot liang at usc dot edu
Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons
HAND Me the Data: Fast Robot Adaptation via Hand Path Retrieval
