Hechang Chen

Papers

Total Citations

H-Index

About

Hechang Chen is a leading researcher in reinforcement learning (RL), with a focus on improving agent generalization, multi-agent coordination, and continual learning. His work addresses critical challenges in making RL agents robust and adaptable in dynamic, real-world environments. In his 2023 paper "Learning Generalizable Agents via Saliency-Guided Features Decorrelation," Chen tackles the problem of agents overfitting to task-irrelevant visual features, such as background noise, by introducing a method that decorrelates salient features—an approach that enhances generalization across unseen environmental variations. This work, with 3 citations, is foundational for developing more reliable visual-based RL systems. Chen also advances multi-agent RL through his 2022 paper "Distributional Reward Estimation for Effective Multi-Agent Deep Reinforcement Learning," which mitigates reward uncertainty in complex settings like robotics and autonomous driving, earning 2 citations. His 2025 work "Continual Diffuser (CoD)" pioneers continual offline RL by integrating diffusion models with experience rehearsal, enabling agents to master sequential tasks without catastrophic forgetting. With a growing citation impact, Chen’s contributions are shaping the future of adaptive, scalable RL systems, making him a key figure in the field.