Event
Generalist publishes physical commonsense research update
Key points
- Generalist published a research update arguing that large-scale physical interaction data was yielding emergent physical commonsense in robot behavior.
- The company described handheld data-collection devices designed to preserve reactive human manipulation behavior and linked that data engine to better recovery, transfer, and low-data adaptation.
- The post also pointed back to earlier one-shot assembly and long-horizon manipulation results as evidence of broader model competence.
Company context
Generalist builds embodied foundation models for robots, initially focused on dexterous manipulation. Its work combines real-world robot interaction data, in-house data-collection hardware and models intended to generalize across tasks.
Context
- Company
- Generalist
- Segment
- Foundation Models
- Event type
- Capability
- Geography
- San Mateo, California · United States