Pretraining Builds Representations, Posttraining Induces Metapolicy

Liu_eroteme · x · 2026-08-27

The post discusses the distinction between pretraining and posttraining, suggesting that pretraining primarily constructs observation and policy representations, while posttraining induces a 'metapolicy' for representation and policy selection. This framing aligns with observations in reinforcement learning.

Related event: Pretraining as Transformation, Post-training as Translation(2 posts)→

Original post →

More from Research

Research channel →