Yuanfang Peng

Yuanfang Peng is a researcher at the Hong Kong University of Science and Technology who also collaborates with Microsoft Research Asia. He is the lead author of PAIR-VLA, a reinforcement learning fine-tuning framework that improves the visual robustness of Vision-Language-Action models for robotic manipulation by introducing auxiliary invariance and sensitivity objectives during PPO optimization. His work focuses on the intersection of reinforcement learning fine-tuning and visual robustness for embodied AI and robotic systems.