Open Models Almanac
Methodspost-training

single source · post-training

Two-stage post-training pipeline (SFT then RL)

Bootstraps capabilities with supervised fine-tuning, then applies reinforcement learning to tasks the model does not yet solve at a high pass rate.

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

Also called RL reserved for tasks with low pass rate.

Filed under post-training