![]()
Chenxiao Gao
Georgia Institute of Technology
I am a Machine Learning Ph.D. student at Georgia Institute of Technology, advised by Prof. Bo Dai. I am fortunate to collaborate with Prof. Na Li and the students in her group at Harvard University. Before starting my Ph.D., I received my bachelorβs and masterβs degrees from Nanjing University, where I conducted research in the LAMDA group.
My research focuses on reinforcement learning (RL) and generative modeling, with applications to LLM agents and robotics. My work spans three directions:
- Efficient reinforcement learning: Developing RL algorithms that improve efficiency through self-supervised learning and advances in generative model architectures.
- Generative models and agents for robotics: Developing foundational generative models and agentic systems for robotic learning and decision-making.
- LLM and agent post-training: Designing efficient post-training algorithms for large language models and interactive agents.
Feel free to contact me if you are interested in my research!
News
-
Our new preprint, Learning Expressive and Compositional Motion Representation via Spectral Skills, won the Best Paper Award at the IROS 2026 Workshop on Compositional and Modular Learning in the Era of Scaling Robotics! Explore the project page.
-
We released RLE-Bench, a benchmark evaluating coding agents as robot learning engineers. Read our research blog and the Harvard SEAS news coverage.
-
I moved to Boston and joined Prof. Na Liβs group at Harvard University as a Research Fellow!
Education
Experience
Academic Service
- Reviewer for conferences: ICML 2025-2026, NeurIPS 2025, ICLR 2025-2026, IJCAI 2025, AAAI 2025-2026, UAI 2025
- Reviewer for journals: TMLR
- Teaching Assistant: CX4240 - Computing for Data Analysis
Software / Projects
- RLE-Bench
A benchmark for evaluating coding agents as robot learning engineers.
- Flow-RL
A modular JAX framework for reinforcement learning with diffusion and flow policies.
- Spectral-RL
Spectral representations for reinforcement learning and downstream policy optimization.
Visitors
- Visits
- β
- Countries / regions
- β