PAVE: Predictive Alignment and Value-Guided Evolution for World-Action Policies

Source: arXiv cs.AI·2026-09-03 12:00Publish time unverified

arXiv:2608.30378v2 Announce Type: replace-cross Abstract: Direct vision-language-action policies generate continuous robot actions efficiently, but standard behavior cloning leaves two complementary gaps: their representations are not explicitly required to describe how the scene evolves over multiple time scales, and deployment trajectories of unequal quality are often reused without separating useful dynamics from undesirable behavior.