Large Language Models

3 Credits

A comprehensive introduction to Large Language Models (LLMs), a foundational technology behind recent breakthroughs in Artificial Intelligence (AI). Fundamental knowledge in LLMs, including LLM tokenization, model architecture, pre-training, fine-tuning, decoding, generation, and post-training techniques are introduced. Recent developments in LLMs, such as retrieval-augmented LLMs, knowledge-enhanced LLMs, reasoning and planning with LLMs, LLMs agent, and multi-modality LLMs are covered. Limitations of current LLMs, including harmful behaviors, hallucinations, and safety alignment concerns are examined, with discussion of mitigation strategies.