Turn the dial. Revisit a paper. Notice what it changed, what it tested, and what it left for the next curious person.
2017
Foundations
Attention Is All You Need
Vaswani et al.
The Transformer organized sequence modeling around attention, dispensing with recurrence and convolutions in its proposed architecture. The paper evaluated the design on machine translation.
REPORTED CONTRIBUTION
A sequence architecture with self-attention and parallelizable training, evaluated against translation baselines.
READ WITH THIS LIMIT
Translation results do not establish every later capability associated with Transformer-based models.
OUR QUESTION, LOOKING BACK
When we add more context, which information actually changes an answer?
Eight selected research milestones, not a complete history or a survey of the latest frontier. Dates are first arXiv submissions, not journal publication dates. Summaries describe the cited papers; the open questions are our editorial interpretation. A later date does not establish that one method replaces another.
