Papers
6
Total Citations
1,067
H-Index
6
About
Justin Fu is a leading researcher in reinforcement learning (RL), with a focus on making RL practical for real-world robotic systems. His primary contributions lie in offline reinforcement learning, where he has helped pioneer methods that allow agents to learn effective policies from static, pre-collected datasets—eliminating the need for costly or risky online interaction. His highly cited tutorial, "Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems" (795 citations), has become a foundational resource for the field. Fu has also advanced sample-efficient robotic learning through one-shot manipulation skills, combining online dynamics adaptation with neural network priors (134 citations). His work on inverse reinforcement learning bridges language and vision, enabling robots to follow instructions from natural language goals (68 citations). Additionally, he has contributed to semi-supervised RL for skill generalization and to benchmarks for off-policy evaluation (25 citations), providing critical tools for assessing policy performance offline. Through these efforts, Fu has shaped how researchers approach data-driven decision-making in robotics, making him a key figure in the transition of RL from simulation to real-world deployment.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4Generalizing Skills with Semi-Supervised Reinforcement Learning34 citations · 2016
- 5Benchmarks for Deep Off-Policy Evaluation25 citations · 2021
- 6