TheMachine Press

A daily newspaper for the age of artificial intelligence.

Morning editionPermanent story

robotics

One Hundred Hours of Human Touch Improved the Robot's World Model

Shared tactile arrays transferred contact dynamics while real-robot supervision stayed fixed at five hours.

Published Updated Story ID: mp-2026-09-18-009
Read the complete editionStory JSON

Summary

Shared tactile arrays transferred contact dynamics while real-robot supervision stayed fixed at five hours.

DexTouch-WM places compatible flexible tactile arrays on human and robot hands, retargets human motion into the robot action space, and trains a model to predict future video and bilateral touch. With five hours of robot data held constant, increasing human interaction from zero to 100 hours improved held-out robot-domain visual, geometric and contact prediction even though the human and robot task sets did not overlap. The model also served as a surrogate environment for policy evaluation and generated synthetic trajectories for physical policy learning.

Why it matters

Shared tactile arrays transferred contact dynamics while real-robot supervision stayed fixed at five hours.

Limits and context

  • With five hours of robot data held constant, increasing human interaction from zero to 100 hours improved held-out robot-domain visual, geometric and contact prediction even though the human and robot task sets did not overlap.

Key claims

  1. Shared tactile arrays transferred contact dynamics while real-robot supervision stayed fixed at five hours.

    Qualification: With five hours of robot data held constant, increasing human interaction from zero to 100 hours improved held-out robot-domain visual, geometric and contact prediction even though the human and robot task sets did not overlap.

    Evidence: source-2026-09-18-009

Sources

  1. arXiv preprint 2609.20649arXiv · primary research

Corrections

No corrections have been recorded for this story.