Open Models Almanac
Methodspost-trainingreinforcement learning algorithm

single mention · post-training

Per-token weight normalization by rollout group

Assigns equal weight to trainable tokens within a group and scales weights so the group's total equals its rollout count.

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.

Filed under post-training › reinforcement learning algorithm