Tsinghua's Embodied Model Tops Global Rankings. Outperforming GPT-6 and NVIDIA Without Hacks or Extra Data

Xingdong Jiyuan's self-developed world action model VPP2 has topped the RoboDojo simulation leaderboard with an average success rate of 32.26% and a score of 39.26. It leads in three key capabilities. The model achieved this breakthrough using only standard datasets, without additional data or enhancement strategies, demonstrating strong generalization and execution capabilities. VPP2 employs a phased training strategy to decouple video prediction from action learning, enabling both predictive and action generalization, offering a new paradigm for embodied intelligence development.