Post-training is the umbrella term for everything that turns a 'raw' pre-trained model into a useful, safe, instruction-following assistant. It typically includes supervised fine-tuning (SFT), RLHF, DPO or Constitutional AI-style synthetic feedback stages — OpenAI's InstructGPT and Anthropic's HH (Helpful & Harmless) line are the canonical references. At frontier labs, post-training is now where most of the perceived quality gap between models is decided, often more so than Pre-training. Heavy use of Synthetic Data and tight evaluation loops are central to this stage.
MEVZU N°124ISTANBULYEAR I — VOL. III
Glossary · Intermediate · 2022
Post-training
The stage after pre-training that turns a raw model into a helpful, safe, instruction-following assistant.
- EN — English term
- Post-training
- TR — Turkish term
- Sonrası-Eğitim