Open Models Almanac
Methodspost-trainingagentic post-training

single mention · post-training

Per-group randomization of system prompt, coding harness, and user request

Randomizing the system prompt, coding harness, and user request per group to reduce policy overfitting.

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

Also called randomised per group.

Filed under post-training › agentic post-training