AI Glossary Term

What does “Post-training” mean?

Definition

Training steps applied after pretraining, such as instruction tuning, preference optimization, and safety tuning.

Browse the full AI Glossary Explore all free AI guides

How Post-training works

Example

A base model that only continues text is taught to follow instructions, to prefer helpful answers, and to refuse harmful requests. The result feels like an assistant, though the underlying knowledge came from pretraining.

Why it matters

Post-training shapes how a model behaves and presents itself. It strongly affects usefulness and safety, even when the base is the same.

Common questions

Does post-training add knowledge?

Mostly it shapes behavior and format rather than adding large amounts of new knowledge.

What techniques are used?

Supervised fine-tuning, RLHF or DPO, and safety training.

Is it cheaper than pretraining?

Generally yes, though it requires carefully built data and testing.

DPO (Direct Preference Optimization)A training method that fine-tunes models directly on preference pairs without needing a separate reward model.Mid-trainingAn intermediate training phase between pretraining and post-training, often used for capability or domain adjustments.Parameter-Efficient Fine-Tuning (PEFT)Methods that adapt models by training a small subset of added parameters.PretrainingInitial large-scale model training on broad data before downstream adaptation.Evaluation SetA held-out dataset used to measure model quality after training.Fine-TuningContinuing training on domain-specific data to adapt a pre-trained model to a specific task.

Learn more in our free guides

AI Training Pseudo-Labeling and Self-Training ELECTRA Pretraining RoBERTa Training Recipe

Test yourself

AI Training Quiz Pseudo-Labeling and Self-Training Quiz ELECTRA Pretraining Quiz