Jinxin Liu

Hi, I am Jinxin Liu (刘金鑫), a fifth-year (joint) Ph.D student at Zhejiang University and Westlake University (Machine Intelligence Lab, MiLAB), supervised by Prof. Donglin Wang. Before that, I obtained my bachelor's degree of Engineering from Chongqing University of Posts and Telecommunications, majoring in Communication Engineering.

The central goal of my research is to build decision-making agents that are capable of performing complex tasks in a variety of unstructured environments. I strive to achieve this by developing deep reinforcement learning (RL) algorithms and deploying sample-efficient RL to real-world tasks. My research helps people acquire general and robust decision-making behaviors in both simulated and real-world tasks. Especially, I study three main topics:

• Deep Reinforcement Learning (RL): focusing on general and ready-to-be-deployed RL algorithms, i.e, imitation learning, reward-free RL, unsupervised RL (learning skills), and RL in real-world tasks (games).
• Planning and Inference: offline RL, offline-to-online RL, embodied agent, and design from data.
• Distribution Shift: RL dynamics/embodyment adaptation, multi-goal RL, sim2real, and sample-efficient RL.