Papers
11
Total Citations
275
H-Index
7
About
Chenjia Bai is a prominent reinforcement learning researcher whose work spans exploration strategies, multi-goal learning, offline RL, and robot locomotion. He is perhaps best known for his comprehensive survey on exploration in deep reinforcement learning — covering both single-agent and multiagent domains — which has garnered an impressive 158 citations and stands as an authoritative reference for researchers tackling sample inefficiency in complex environments like game AI, autonomous vehicles, and robotics. His contributions to multi-goal reinforcement learning are equally notable, with a series of works addressing hindsight experience replay, hindsight bias, and attentive goal generation that collectively advance how agents learn from sparse-reward settings. Bai has also made meaningful strides in safety-aware offline RL, proposing monotonic quantile networks to optimize worst-case return criteria — a critical concern in safety-sensitive applications. More recently, his research has extended into physical robot control, with notable work on risk-averse quadrupedal locomotion and humanoid balance on challenging terrain, demonstrating a clear trajectory toward real-world deployment. Across his body of work, Bai consistently bridges theoretical rigor with practical impact, making him a valuable voice in modern reinforcement learning research.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Guided goal generation for hindsight multi-goal reinforcement learning20 citations · 2019
- 4Addressing Hindsight Bias in Multigoal Reinforcement Learning18 citations · 2021
- 5Monotonic Quantile Network for Worst-Case Offline Reinforcement Learning16 citations · 2022
- 6Generating attentive goals for prioritized hindsight reinforcement learning14 citations · 2020
- 7
- 8Robust Quadrupedal Locomotion via Risk-Averse Policy Learning7 citations · 2024
- 9
- 10Cross-Domain Policy Adaptation via Value-Guided Data Filtering3 citations · 2023