A Chinese robotics company, Star-Epoch, has developed a world action model (WAM) called VPP2 that has achieved the top ranking on the RoboDojo benchmark, surpassing leading models like GPT-6-ASTRA and Nvidia's GR00T-N1.7. VPP2's success is attributed to its novel approach of decoupling and reordering video prediction and action learning, focusing on robust video prediction before training action execution. This method allows VPP2 to demonstrate strong generalization, precise manipulation, and memory capabilities, even performing well in zero-shot tasks on real-world robots without additional fine-tuning. AI
IMPACT Sets a new benchmark for embodied AI, potentially accelerating the development of more capable and general-purpose robots.
RANK_REASON New world action model (WAM) release from a non-Tier-1 lab that claims SOTA performance on a recognized benchmark, surpassing Tier-1 competitors. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
