Pretraining Next-Token Prediction: 预训练的发动机 Scaling Laws: 模型、数据、算力的三角博弈 Training Tech Outline Train LLM from scratch How to make custom dataset? Alignment 从 GPT 到 ChatGPT：让模型学会「听话」的对齐之路 Detailed explanation of RLHF technology Parent MOC Large Language Model - MOC