arXiv · 2611.16450
A Recipe for Physical Action Model Training
Abstract
Choosing robot model capacity and demonstration budgets also requires decisions about data composition, representation, optimization, and deployment. We connect these decisions through a common conditional-distribution target. Nested learning oracles give exact risk accounting; local information geometry, conditional mutual information, and empirical convergence provide conditional training prescriptions. Known-target numerical checks expose capacity-data coupling, representation-induced mixture mismatch, and an example where deployment KL increases while negative log-likelihood decreases. An analysis of 206 public PushT trajectories evaluates fixed-feature information-design criteria without fitting a policy. Under asymmetric evaluation targets, the selected allocations reduce this criterion by 4.5-6.2% relative to target matching; balanced targets show no benefit. These results clarify which quantities can support training recommendations and when an apparently favorable predictive metric has a different interpretation.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Shaojiang Wang, Geer Yang. 2026-11-12. A Recipe for Physical Action Model Training. https://searcharxiv.com/abs/2611.16450
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.