Research papers, from first principles

Start with the problem the authors faced. Work through a small example. Then read the part of the paper that gives the idea its name.

Paper	The question it helps answer
Attention Is All You Need · 2017	How can one piece of text combine information from other pieces?
Recurrence papers · 2018–2025	Can a model do more internal work using the same learned rule?
DeepSeek-V4.1-Flash · 2026	How can a large model avoid preparing and reading the same information repeatedly?

New to the subject? Begin with Attention Is All You Need. It explains the basic operation the other papers build on.

Each reading introduces terms when they become useful, links to the corresponding source section, and ends with a question to check your understanding. Extra calculations are optional.