GenRobot AI Spotlights Touch, Data at CoRL

Physical AI’s hardest problem may be contact, not cognition: robots need aligned visual, motion, tactile, and force data before better policies can reliably control the real world.

• –DEX 2.0 focuses on fine-grained hand motion and tactile interaction for dexterous manipulation.
• –EGO 2.0 captures first-person multimodal data with wide vision, synchronized cameras, and precise hand tracking. [Product details](https://www.genrobotai.com/products/ego2)
• –The event’s surrounding talks connected tactile world models, coding agents, VLA systems, and online exploration—signaling a shift toward closed-loop robot learning.
• –For developers, standardized embodied datasets may prove more valuable than another isolated model checkpoint.
• –The unresolved test is transfer: whether human-captured signals improve force control and recovery on unfamiliar robots.