Recently, RoboTwin unveiled its latest leaderboard, where the Zhiyuan GO-3 Nano Preview emerged as the frontrunner, boasting an average success rate of 85.01%. According to reports, RoboTwin's evaluation primarily focuses on dual-arm collaborative manipulation and the ability to generalize across diverse environments in complex scenarios. By designing various scenarios and tasks, it rigorously assesses a model's performance in target recognition, action coordination, and adaptability to environmental changes.
