In one line, what is pre‑training in LLMs, and how does it differ from post‑training (fine‑tuning)?

Pre‑training is the initial, large‑scale, unsupervised learning phase where a language model learns general language representations by predicting masked tokens or next words across a massive corpus. Fine‑tuning (post‑training) is a subsequent supervised phase where the pre‑trained weights are adapted to a specific downstream task—such as sentiment analysis or question answering—using labeled data. The key difference is that pre‑training captures broad linguistic knowledge, while post‑training tailors that knowledge to a particular application. In practice, pre‑training is computationally expensive but reusable across many tasks; fine‑tuning is cheaper and task‑specific.