Pre-training vs Fine-tuning in LLMs

Question: What is the difference between pre-training and fine-tuning a language model? Key points: Pre-training teaches a model general language capability ("how to speak") via self-supervised learning (e.g. next-token prediction) over massive, broad text corpora. Fine-tuning specializes that pre-trained model for a narrower task or domain ("how to be a medical assistant") using smaller, task/domain-specific labeled or curated data, often combined with instruction-tuning or RLHF for alignment.