robotics
The World Model Planned Through Its Policy’s Noise
LeWAM combines four prediction modes in one latent model and constrains planning to reduce exploitation of model error.
Summary
LeWAM combines four prediction modes in one latent model and constrains planning to reduce exploitation of model error.
LeWAM trains forward, backward, inverse-dynamics and policy prediction together on a decoder-free JEPA representation. Linear probes read robot and object state better than from a forward-only comparison while the latent remains resistant to visual distractors. At matched size, closed-loop action results matched a regular flow-matching policy trained on the same encoder, with a world model included. For planning, the authors steer samples through the policy head’s noise space rather than optimize raw actions, improving the tested world-action models’ closed-loop performance. The distinction matters because an optimizer can discover actions that exploit prediction errors instead of helping the real task. The evidence remains bounded by the reported evaluations.
Why it matters
LeWAM combines four prediction modes in one latent model and constrains planning to reduce exploitation of model error.
Limits and context
- Linear probes read robot and object state better than from a forward-only comparison while the latent remains resistant to visual distractors.
Key claims
LeWAM combines four prediction modes in one latent model and constrains planning to reduce exploitation of model error.
Qualification: Linear probes read robot and object state better than from a forward-only comparison while the latent remains resistant to visual distractors.
Evidence: source-2026-10-09-010
Sources
- arXiv preprint 2610.12407, original v1arXiv / submitting research authors · primary research
Corrections
No corrections have been recorded for this story.