Xin Lang Cai Jing·2026-02-12 08:43Core Insights - Alibaba's Gaode has launched two foundational models, ABot-M0 for embodied operation and ABot-N0 for embodied navigation, achieving state-of-the-art (SOTA) capabilities in both areas, marking a significant milestone in the robotics industry [1][3]. Group 1: Embodied Operation Model (ABot-M0) - ABot-M0 addresses key challenges in robotics, including data fragmentation, inconsistent action representation, and insufficient spatial understanding, which have hindered the scalable application of robotic technology [1][3]. - The model has undergone systematic reconstruction focusing on "data unification, algorithm innovation, and spatial perception," enhancing its generalization ability across diverse robotic forms and task scenarios [1][3]. - In benchmark tests such as Libero, Libero-Plus, and RoboCasa, ABot-M0 achieved an average task success rate that meets SOTA, with Libero-Plus reaching 80.5%, nearly a 30% improvement over the advanced industry solution pi0 [1][3]. Group 2: Embodied Navigation Model (ABot-N0) - ABot-N0 aims to unify various navigation tasks, overcoming the fragmentation seen in current embodied navigation research, which often relies on isolated architectures for specific tasks [2][4]. - The model integrates five major navigation tasks: Point-Goal, Object-Goal, Instruction-Following, POI-Goal, and Person-Following, effectively breaking through traditional task segmentation barriers [2][4]. - ABot-N0 has set world records in seven authoritative benchmark tests, including CityWalker, SocNav, R2R-CE/RxR-CE, HM3D-OVON, BridgeNav, and EVT-Bench, showcasing its innovative capabilities [2][5].