ai-post-training

# AI Post-Training **Domain**: the rung *after* supervised fine-tuning — turning a pretrained or SFT'd base model into an aligned, preference-tuned, or reasoning-capable model with a **reward signal**. This skill owns the post-training *decision and pipeline*: when to post-train at all, which reward signal you can produce, which algorithm family fits, and how to keep it from over-optimizing. Per-algorithm operational depth lives in [ai-llm/references/post-training.md](../ai-llm/references/post-training.md) (PPO, DPO, SimPO, KTO, GRPO, GSPO, DAPO, RLVR, RULER, ORPO — catalogue + decision tree); this skill routes there. It does **not** cover: pretraining ([ai-pretraining](../ai-pretraining/SKILL.md)), the prompt→RAG→SFT promotion ladder ([ai-architecture-advisor](../ai-architecture-advisor/SKILL.md)), or serving the result ([ai-llm-inference](../ai-llm-inference/SKILL.md)). ## Quick Reference | You have / want | Method | Deep ref | |---|---|---| | Labeled demonstrations of the target behavior | **SFT** (baseline — exhaust it first; not RL) | [ai-llm](../ai-llm/SKILL.md) | | Pairwise preferences, want the least machinery | **DPO** (or DAAs: KTO / ORPO / SimPO) | [methods](references/methods-and-pipeline.md) | | A stronger teacher model, a small student | **On-policy distillation** — try before GRPO | [methods](references/methods-and-pipeline.md) | | Preferences + reward model + online RL | **GRPO / RLOO** (critic-free, 2026 default); **PPO** is the reference algorithm, now `...