Xinyang Tong
Papers
2
Total Citations
6
H-Index
2
About
Xinyang Tong is an emerging researcher at the forefront of embodied artificial intelligence, with a specialized focus on quadruped robotics, vision-language-action (VLA) models, and reinforcement learning. Their work addresses one of the most pressing challenges in modern robotics: enabling legged robots to perform versatile, real-world tasks by integrating multimodal large language models with physical control systems. Tong's most notable contribution, "MoRE: Unlocking Scalability in Reinforcement Learning for Quadruped Vision-Language-Action Models" (2025, 4 citations), introduces a pioneering Mixture of Robotic Experts framework that enhances scalability and task versatility in quadruped locomotion. Complementing this, their work on "Quart-Online" (2025, 2 citations) tackles the critical problem of inference latency in deployed multimodal models — a practical bottleneck that conventional parameter reduction fails to solve — proposing innovative solutions that preserve model performance without sacrificing real-time responsiveness. Though early in their research career, Tong's contributions are already shaping the conversation around deployable, intelligent quadruped systems. Their research sits at a compelling intersection of robotics, large language models, and reinforcement learning, making them a researcher to watch as embodied AI rapidly matures.
Research Focus
Key Achievements
Top Papers
- 1
- 2