Chenzhuang Du

Papers

Total Citations

H-Index

About

Chenzhuang Du is a researcher advancing the frontiers of multi-modal learning and reinforcement learning. His work focuses on developing more efficient and robust AI systems that can learn from multiple data sources—such as vision, language, and sound—while also mastering complex decision-making tasks. Du’s most impactful contribution, "Improving Multi-Modal Learning with Uni-Modal Teachers" (2021), introduces a novel framework that leverages uni-modal teachers to guide multi-modal student models, addressing the common pitfall of joint training objectives that often lead to suboptimal representations. This paper has garnered 25 citations, reflecting its influence on the field. In reinforcement learning, Du’s "Intrinsically Motivated Self-supervised Learning in Reinforcement Learning" (2022) explores how intrinsic motivation can enhance self-supervised auxiliary tasks, aiming to improve sample efficiency and semantic representation learning in vision-based RL. While still early in its impact, this work hints at Du’s broader interest in bridging unsupervised learning and RL. His research is notable for its practical orientation toward real-world robotic applications, where robust multi-modal perception and efficient learning are critical. Du’s contributions are shaping how AI systems integrate diverse sensory inputs and learn from limited interactions.