Astribot publishes Lumo-2 world-action model research
Astribot published Lumo-2, a latent world-action model for predictive, aligned and scalable robot learning, alongside a technical report on arXiv.
Evidence notes
- The report introduces a latent world-action model that reasons over world dynamics in latent space before generating robot actions.
- Astribot says Lumo-2 improves over Lumo-1 through predictive latent dynamics, progressive modality pre-alignment and broader multi-source data scaling.
- The official research page highlights real-world evaluations covering temporal reasoning, physical understanding, long-horizon tasks and dexterous manipulation.
Company context
Astribot develops AI humanoid and mobile-manipulator systems that combine robot hardware, manipulation, teleoperation data collection, developer tooling and embodied models for research and emerging real-world applications.