The docs use a fixed vocabulary that each page glosses on first mention: System 1 decision model, decision model, model slot, front (tool, browser, rail), rail, head (choice, noul, score), arm, tick, replay, episode, labelled set. A reader landing mid-page has no lookup.
Write docs/glossary.md: one line per term, then a link to the page that defines it. Keep the first-mention glosses on the pages as they are; the glossary is the lookup. Link it from the Docs list in README.md.
Where to start: docs/decision-models.md for the model terms, docs/architecture.md for the fronts and the slot, docs/benchmarks.md for arms and runs, evals/README.md for episodes, ticks and labelled sets.
The docs use a fixed vocabulary that each page glosses on first mention: System 1 decision model, decision model, model slot, front (tool, browser, rail), rail, head (
choice,noul,score), arm, tick, replay, episode, labelled set. A reader landing mid-page has no lookup.Write
docs/glossary.md: one line per term, then a link to the page that defines it. Keep the first-mention glosses on the pages as they are; the glossary is the lookup. Link it from the Docs list inREADME.md.Where to start:
docs/decision-models.mdfor the model terms,docs/architecture.mdfor the fronts and the slot,docs/benchmarks.mdfor arms and runs,evals/README.mdfor episodes, ticks and labelled sets.