Posttraining

posttraining.dev

Posttraining

Notes on AI systems after pretraining.

Essays on post-training, evaluation, agents, alignment, productized AI, and the engineering details behind useful models.

Read essays Subscribe Support

Latest Essays

View all
  • distillation
  • post-training
  • alignment
  • evals

On-Policy Distillation

A careful introduction to on-policy distillation as on-policy student rollouts with dense teacher supervision.

  • meta
  • post-training
  • ai

Welcome to Posttraining

Why this site exists and what kinds of AI systems writing will live here.