Papers
1
Total Citations
5
H-Index
1
About
Liyuan Zheng is a rising researcher in multi-agent reinforcement learning (MARL), with a focus on autocurricular training—a paradigm where agents co-evolve to learn emergent behaviors without explicit supervision. Their key contributions lie in bridging game theory and MARL, particularly through the lens of Stackelberg games, to model competitive dynamics during autocurricula. In their highly cited 2023 work, *Stackelberg Games for Learning Emergent Behaviors During Competitive Autocurricula*, Zheng formalized how leader-follower asymmetries can stabilize and accelerate skill emergence in adversarial multi-agent settings, offering a theoretical foundation for training robust, physically grounded agents. This paper has garnered early attention (5 citations) for its novel integration of hierarchical decision-making into unsupervised co-evolution. Zheng’s research directly impacts robotics, where such frameworks enable agents to develop resilient strategies against adaptive opponents—a critical step toward deployable autonomous systems. By merging game-theoretic rigor with scalable learning algorithms, Zheng is shaping how we understand and engineer emergent intelligence in complex, competitive environments. Their work promises to unlock more efficient, interpretable, and robust multi-agent systems.
Research Focus
Key Achievements
Top Papers
- 1