Magic Atom's EchoPolicy-0 Tops MolmoSpaces with 75.4% Success Rate, Validating Zero-Shot Embodied Agent Generalization

1 reports archived1 independent sourcesupdated 10/13/2026, 19:05:39

Synthesis & Latest Updates

1 Sources Cross-Validated

Magic Atom's VLM Agent framework EchoPolicy-0 ranked first on the global MolmoSpaces benchmark with a 75.4% average success rate, while its sibling framework MagicBrain secured fourth place. This achievement marks a shift from 'demo-ready' to 'production-ready' embodied intelligence, validating the effectiveness of hierarchical architectures and instruction-following capabilities in complex zero-shot environments.

LATEST/Magic Atom's VLM Agent framework EchoPolicy-0 ranked first on the global MolmoSpaces benchmark with a 75.4% average success rate, while its sibling framework MagicBrain secured fourth place. This achievement marks a shift from 'demo-ready' to 'production-ready' embodied intelligence, validating the effectiveness of hierarchical architectures and instruction-following capabilities in complex zero-shot environments.

HEAT TRENDHourly heat curve

3 fully observed hours

Now

10.7

Peak

11.222:00

24h change

–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

TIMELINECoverage timeline

Total 1 reports · Latest first

量子位官网T2·Oct 13, 19:0568 pts
Magic Atom's EchoPolicy-0 Tops MolmoSpaces with 75.4% Success Rate, Validating Zero-Shot Embodied Agent Generalization
- EchoPolicy-0 ranks first globally on the MolmoSpaces Combined leaderboard with a 75.4% average success rate across all four v1 tasks.
- MagicBrain secures fourth place with a 3.4B parameter scale, achieving a 94.4% grasping success rate that significantly exceeds industry averages.
- The evaluation emphasizes zero-shot generalization, requiring robots to perform free grasping and articulated interactions in unseen indoor environments, object distributions, and geometric structures.
View Evidence Original Source