Large Language Models

A topic-organized track. Seven sub-sections move from foundational ideas to current research.

Reading path

1. LLM Basics — word embeddings → the Transformer → pre-training → scaling laws → instruction tuning.
2. Reasoning & Post-training — chain-of-thought, latent-space reasoning, RLHF, RLVR.
3. Efficient Methods — parameter-efficient fine-tuning, efficient RLVR, efficient inference, long-context.
4. Factuality — hallucination and calibration.
5. Applications — RAG, agents, agentic RAG, multi-modal LLMs.
6. Evaluation — evaluating LLMs and detecting LLM-generated text.
7. Other Topics — alternative architectures (MoE / SSM / RWKV), bias, safety.

For a year-by-year view of the same models, see The Transformer Era →.

Each topic page covers

A short introduction followed by the canonical paper list for the topic. Where a paper has a name (LoRA, DPO, Medusa, …) it's used as the heading; otherwise the full title is listed with its venue and year.