Open Models Almanac
Methodspost-trainingrollout & RL infrastructure

single mention · post-training

One-step-stale asynchronous response generation

Allows responses to be generated from the previous policy snapshot, making them one step stale.

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 one step stale.

Filed under post-training › rollout & RL infrastructure