43. Large language models and foundation models

Study large pretrained models built from broad datasets, including scaling behavior, tokenization, pretraining objectives, instruction tuning, and emergent capabilities. Compare language, vision, audio, and multimodal foundation-model designs.