I received my B.Sc. degree from Kuang Yaming Honors School, Nanjing University in June 2022, majoring in Brain Science and Artificial Intelligence.

From September 2022 to June 2025, I was a member of the LAMDA Group and pursued my M.Sc. degree at the School of Artificial Intelligence, Nanjing University, researching on Reinforcement Learning under the supervision of Prof. Zongzhang Zhang.

Currently, I am a Ph.D. student at the Department of Computer Science, Peking University, supervised by Prof. He Wang. My current research interest lies in humanoid robot control.

Publications

HumanoidNDM: Closing the Training-Deployment Gap of Motion Tracking via Humanoid Neural Dynamics Model

Chao Chen*, Zhenrong Wang*, Jilong Wang, Yusen Feng, Jiangran Lyu, Li Yi, He Wang^†

CoRL 2026

We bridge the training-deployment gap in humanoid motion tracking by introducing an online fine-tuning framework.

Focus-Then-Reuse: Fast Adaptation in Visual Perturbation Environments

Jiahui Wang*, Chao Chen*, Jiacheng Xu, Zongzhang Zhang^†, Yang Yu

Neurips 2025 [PDF] [Code]

We propose an object-filter to extract task-relevant information at the object level, enabling rapid cross-domain policy deployment.

Deep Anomaly Detection via Active Anomaly Search

Chao Chen*, Dawei Wang*, Feng Mao, Jiacheng Xu*, Zongzhang Zhang^†, Yang Yu

AAMAS 2024 [PDF] [Code]

We propose an active anomaly searching mechanism based on Reinforcement Learning, achieving competitive performance in semi-supervised anomaly detection tasks.

Focus-Then-Decide: Segmentation-Assisted Reinforcement Learning

Chao Chen*, Jiacheng Xu*, Weijian Liao, Hao Ding, Zongzhang Zhang^†, Yang Yu, Rui Zhao

AAAI 2024 [PDF] [Project] [Code]

We boost the performance of reinforcement learning under visual perturbations by introducing a segmentation model.

Internal Logical Induction for Pixel-Symbolic Reinforcement Learning

Jiacheng Xu, Chao Chen, Fuxiang Zhang, Lei Yuan, Zongzhang Zhang^†, Yang Yu

KDD 2023 [PDF] [Code]

We introduce a novel framework that synergizes deep RL and rule learning to handle hybrid pixel-symbolic inputs, achieving superior performance and transferability.