UniBot 世界挑战赛 UniBot World Challenge
本赛事直击具身智能大模型在真实世界泛化性不足的核心瓶颈,望汇聚全球顶尖力量,在复杂多变的桌面任务中全面检验单模型多任务的通用性能,赛事围绕抓取、放置、双手协调等任务,开展统一、规模化的真实硬件泛化评测 This challenge targets a core bottleneck — the limited real-world generalization of embodied-AI foundation models — gathering top teams worldwide to comprehensively evaluate single-model, multi-task general performance across complex, varied tabletop tasks, centering on grasping, placement and bimanual coordination with unified, large-scale evaluation on real hardware.
UniBot-V1 挑战赛排行榜UniBot-V1 Challenge Leaderboard
UniBot-V1 挑战赛榜单按平均成功率与步骤分综合排序The UniBot-V1 Challenge ranking, sorted by average success rate and step score
赛事愿景Vision
加速推动人形机器人真实世界通用操作能力Accelerating general-purpose manipulation capabilities for humanoid robots in the real world
实现跨任务、跨物品、抗干扰的泛化操作性能Achieving generalizable manipulation across tasks and objects, robust to disturbances
打造人形机器人泛化操作评测体系Building a generalization benchmark for humanoid manipulation
加速人形机器人通用操作能力落地Accelerating the real-world deployment of general-purpose manipulation for humanoids
准备好接受挑战了吗?Ready to take the challenge?
提交模型、运行评测、登上 UniBot-V1 挑战赛排行榜Submit a model, run the benchmark, climb the UniBot-V1 Challenge leaderboard