Open Models Almanac
Methodspost-trainingreinforcement learning algorithm

single mention · post-training

Group-centered batch-wide advantage normalization

Subtracts each group's mean reward and normalizes centered rewards by their standard deviation across the batch.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See reinforcement learning algorithm for the methods filed beside it.

Also called Advantage normalization.

Filed under post-training › reinforcement learning algorithm