Papers
5
Total Citations
51
H-Index
3
About
Paul Weng is a prominent researcher specializing in reinforcement learning, with particular expertise in reinforcement learning from human feedback (RLHF), robotic control, and sample-efficient learning methods. His work bridges the gap between theoretical machine learning and practical robotics applications, addressing some of the field's most pressing challenges. Weng's most influential contribution is his comprehensive survey on Reinforcement Learning from Human Feedback (2023), which has garnered 34 citations and has become a key reference for researchers navigating the growing intersection of human preferences and autonomous learning systems. This work contextualizes RLHF within the broader landscape of preference-based reinforcement learning, providing valuable conceptual clarity to the field. Beyond RLHF, Weng has made meaningful contributions to robotic learning efficiency, developing novel data augmentation techniques to reduce sample requirements in deep reinforcement learning and pioneering hyperparameter auto-tuning methods for self-supervised robotic systems. His work on decomposed deep reinforcement learning further demonstrates his commitment to solving high-dimensional control challenges through structural innovation. His most recent work on diverse query generation for RLHF reflects his continued dedication to making reinforcement learning more practical, adaptive, and aligned with human intent.
Research Focus
Key Achievements
Top Papers
- 1A Survey of Reinforcement Learning from Human Feedback34 citations · 2023
- 2Hyperparameter Auto-Tuning in Self-Supervised Robotic Learning11 citations · 2021
- 3
- 4Decomposed Deep Reinforcement Learning for Robotic Control2 citations · 2020
- 5