On October 30, Qiyuan Robot launched the PrimeSim world model and reached the top of the WorldSimProbe LIBERO list. PrimeSim is a world model simulator for robot training. It constructs a virtual world where robots can interact, trial and error, and learn continuously, so that strategy verification and reinforcement learning can be carried out efficiently. According to the published evaluation results, PrimeSim ranked first in three of the five tests in the LIBERO comprehensive list; using it for reinforcement learning, the strategy success rate increased from 61.7% to 73.2%.

On October 30, Qiyuan Robot launched the PrimeSim world model and reached the top of the WorldSimProbe LIBERO list. PrimeSim is a world model simulator for robot training. It constructs a virtual world where robots can interact, trial and error, and learn continuously, so that strategy verification and reinforcement learning can be carried out efficiently. According to the published evaluation results, PrimeSim ranked first in three of the five tests in the LIBERO comprehensive list; using it for reinforcement learning, the strategy success rate increased from 61.7% to 73.2%.