Open Models Almanac
Methodspost-trainingrollout & RL infrastructure

single mention · post-training

Token-level interruption of generation

Halts in-flight generation at a token boundary to stop rollouts promptly.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See rollout & RL infrastructure for the methods filed beside it.

Also called token-level interruption.

Filed under post-training › rollout & RL infrastructure