想了解机器人怎么同时走路干活？Omega-0 在真实家务上做到 81.8% 成功率，比 pi-0.5 等模型都强，值得看看。
南洋理工大学、北京大学、港科大（广州）与智源研究院联合发布 Omega-0，一种潜在预测世界动作模型，能让机器人同时行走、观察和操作。在 11 项真实家务任务中，单模型成功率达 81.8%，超过 pi-0.5、EgoVLA、GR00T-N1.7 和 psi-0。该模型采用潜在预测机制，提升了对环境变化的响应能力。研究团队强调，Omega-0 在真实家庭场景中的表现优于多个现有基线模型。
Nanyang Technological University, Peking University, and BAAI Release Omega-0 — A Whole-Body Action Model That Hits 81.8 Percent Success on Real Home Tasks
Researchers from Nanyang Technological University, Peking University, HKUST (GZ), and Beijing Academy of Artificial Intelligence (BAAI) released Omega-0, a latent-prediction world action model that lets a humanoid robot walk, look, and work at the same time. Single-model success on 11 real home tasks reached 81.8 percent, ahead of pi-0.5, EgoVLA, GR00T-N1.7, and psi-0.