MagicAtom's embodied Agent framework takes first place globally on MolmoSpaces

What happened: Magic Atom's embodied Agent framework EchoPolicy-0 ranked first globally on AI2's MolmoSpaces Combined benchmark, covering all four v1 tasks. Its self-developed MagicBrain framework ranked fourth. Key numbers: EchoPolicy-0 scored 75.4% average robot manipulation success across 3915 evaluation episodes. Why it matters: The results validate two complementary routes for embodied brains: Agent scheduling with trajectory selection, and visual guidance for action generation. Magic Atom plans to combine both into a full feedback loop and deploy in real industrial settings.

MagicAtom's embodied Agent framework ranks first globally on MolmoSpaces benchmark.

MagicAtom's embodied Agent framework ranks first globally on MolmoSpaces benchmark. — impact context not separately stored; see What changed.