post-training

Type: Concept

Model improvement stage after pretraining; includes SFT, RLHF, rubrics/verifiers, and RL environments; described as partly an 'art' involving taste.

Field: Machine learning

Mentioned in 1 podcast episode