Ziyi Liu

Hi! I am a visiting student at IIIS, Tsinghua University, advised by Huazhe Xu. Previously, I was at USC, where I worked with Gaurav Sukhatme and Joseph Lim.

I received my master's degree in Electronic Engineering from the University of Calgary. Before that, I completed my B.Eng. in Software Engineering at Zhejiang University of Technology.

My main research goal is to advance robots’ ability to autonomously learn, adapt, and engage with the complexities of the real world. I strive to push the limits of robotics by developing innovative machine learning and adaptive control algorithms.

Email  /  CV  /  Google Scholar  /  Twitter  /  Github

We introduce a language-conditioned robotic sketch-painting benchmark designed to evaluate whether VLA models can ground language instructions into continuous drawing actions in simulation and real-world setup.

We propose Multitask discriminator Proximity-Guided IRL (MPG), a novel few-shot IRL method that addresses the challenge of learning a reward function and RL policy from too few demonstrations, which cannot fully specify the task in an environment with variations, by making use of a multi-task dataset of expert trajectories.

Our approach BOSS (BOotStrapping your own Skills) learns to accomplish new tasks by performing "skill bootstrapping," where an agent with a set of primitive skills interacts with the environment to practice new skills without receiving reward feedback for tasks outside of the initial skill set. This bootstrapping phase is guided by large language models that inform the agent of meaningful skills to chain together. Through this process, BOSS builds a wide range of complex and useful behaviors from a basic set of primitive skills.

© 2026 Ziyi Liu. A fork of Jon Barron's website.