Chen et al. (Moto-GPT team)

Yi Chen is a researcher jointly affiliated with The University of Hong Kong and ARC Lab at Tencent PCG. He is the lead and corresponding author of Moto (colloquially Moto-GPT), a language-action model that introduces latent motion tokens as a bridging representation for learning robot manipulation from video data without requiring action labels during pre-training. The paper was accepted as an oral presentation at ICCV 2025 and demonstrated that its 98M-parameter GPT backbone outperforms much larger vision-language-action models such as RT-2-X and OpenVLA on the SIMPLER benchmark, making it a prominent pseudo-label baseline in subsequent robot learning research.