Weidong Huang (黄维东)

Research Engineer, Beijing Institute of General Artificial Intelligence

I am a Research Engineer at the Robotics Lab of the Beijing Institute of General Artificial Intelligence (BIGAI), where I work on building real-world reinforcement learning pipelines for humanoid robots. Previously, I received my M.S. degree from Beihang University, where I was supervised by Yaodong Yang. I have interned at Tencent AI Lab and was a visiting scholar at Peking University and Tsinghua University. I obtained my B.S. degree from South China Normal University.

My research interests lie in advancing reinforcement learning toward complex real-world settings, with the goal of compressing environment dynamics into reusable and generalizable intelligence. I focus on reinforcement learning, world models, and humanoid learning, and I enjoy building robotic systems that can safely and efficiently explore and continuously learn in the real world.

Publications and preprints

Papers sorted by recency. Representative papers are highlighted.

Towards Bridging the Gap between Large-Scale Pretraining and Efficient Finetuning for Humanoid Control
Weidong Huang, Zhehan Li, Hangxin Liu, Biao Hou, Yao Su, Jingwen Zhang
Under Review
project page
ECO: Energy-Constrained Optimization with Reinforcement Learning for Humanoid Walking
Weidong Huang, Jingwen Zhang, Hangxin Liu, Yaodong Yang, Yao Su
Under Review
project page
World Models Should Prioritize the Unification of Physical and Social Dynamics
X Zhang, C Ma, Y Huang, Weidong Huang, S Qi, SC Zhu, X Feng, Y Yang
Conference on Neural Information Processing Systems (NeurIPS), 2025
project page / arXiv
Differentiable Information Enhanced Model-Based Reinforcement Learning
X Zhang, X Cai, B Liu, Weidong Huang, SC Zhu, S Qi, Y Yang
AAAI Conference on Artificial Intelligence (AAAI), 2025 (Oral)
arXiv
SafeDreamer: Safe Reinforcement Learning with World Models
Weidong Huang, Jiaming Ji, Borong Zhang, Chunhe Xia and Yaodong Yang
International Conference on Learning Representations (ICLR), 2024 Poster
project page / OpenReview / arXiv
OmniSafe: An Infrastructure for Accelerating Safe Reinforcement Learning Research
Jiaming Ji, Jiayi Zhou, Borong Zhang, Juntao Dai, Xuehai Pan, Ruiyang Sun, Weidong Huang, ...
Journal of Machine Learning Research (JMLR)
code ⭐ 1K+ / OpenReview / arXiv
Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmark
Jiaming Ji, Borong Zhang, Xuehai Pan, Jiayi Zhou, Weidong Huang, Juntao Dai and Yaodong Yang
Conference on Neural Information Processing Systems (NeurIPS), 2023
code ⭐ 500+ / arXiv

Select open-source projects

LIFT: Towards Bridging the Gap between Large-Scale Pretraining and Efficient Finetuning for Humanoid Control
Weidong Huang, Zhehan Li, Hangxin Liu, Biao Hou, Yao Su, Jingwen Zhang
code / project page
Under Review. Scalable RL framework integrating large-scale SAC pretraining and physics-informed world-model finetuning
SafeDreamer: Safe Reinforcement Learning with World Models⭐ 90
Weidong Huang, Jiaming Ji, Borong Zhang, Chunhe Xia and Yaodong Yang
code / project page / OpenReview / arXiv
ICLR 2024 Poster. Tackling zero-cost performance within SafeRL, finding optimal policy while satisfying safety constraints
OmniSafe: An Infrastructure for Accelerating Safe Reinforcement Learning Research⭐ 1K+
Jiaming Ji, Jiayi Zhou, Borong Zhang, Juntao Dai, Xuehai Pan, Ruiyang Sun, Weidong Huang, ...
code / arXiv
I lead the model-based RL module development, contributing over 13,000+ lines of code
Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmark⭐ 500+
Jiaming Ji, Borong Zhang, Xuehai Pan, Jiayi Zhou, Weidong Huang, Juntao Dai and Yaodong Yang
code / arXiv
I contributed to the development, focusing on the design and validation of safe RL tasks based on visual inputs

Contact

You are welcome to contact me regarding my research! My email is bigeasthuang [at] gmail 。com