← 研踪 PaperTrailarXiv:2611.16450 · v1
A Recipe for Physical Action Model Training
首次提交 2026-11-12 16:03
版本更新 2026-11-12 16:03
雷达观测 2026-11-13 14:15
摘要
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.
作者备注：13 pages, 1 figure, 2 tables; ancillary code and numerical results included
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-12 16:03 · 最新观测版本
A Recipe for Physical Action Model Training
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.
查看此版本解读 →查看此版本原文 ↗