robotics
One Exploratory Clip Calibrates the World Model
PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects.
Summary
PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects.
A short visual history may not reveal the physical properties that govern what happens next. PhysiCalWorld encodes one exploratory interaction from the test scene into a compact descriptor token, then conditions later action-driven rollouts on that token without fine-tuning or explicit physics labels. Across four environments with varied dynamics, the authors report better prediction and visual-planning success, including outside the training ranges. The learned token is an empirical latent description, not a measured physical parameter set.
Why it matters
PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects.
Limits and context
- A short visual history may not reveal the physical properties that govern what happens next.
- The learned token is an empirical latent description, not a measured physical parameter set.
Key claims
PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects.
Qualification: A short visual history may not reveal the physical properties that govern what happens next.
Evidence: source-2026-10-22-005
Sources
- arXiv preprint 2610.30125, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.