Skip to content

History / Evaluating agents

Revisions

  • docs(wiki): add cross-reference and source-code hyperlinks throughout Link source-file references (src/*.rs, tests/*.rs, regret-games/src/kuhn.rs) to the files on GitHub (blob/tree/main), fixing the wrong src/deep.rs path to src/deep/deep.rs. Add page-to-page cross-reference links wherever a page names a concept documented on another page (Game trait, Trainer, Profile, exploitability, subgames, blueprints, abstraction, deep CFR, rtcfr, reference games), so a reader never has to hunt for the referenced topic. Links use page-slug form with no .md extension and no fragile anchors.

    @sqlongithub sqlongithub committed Jul 18, 2026
  • docs(wiki): add rtcfr page, task-oriented nav, and cover remaining public API Close the documentation gaps found by auditing the crate prelude against the wiki. Add an Interior-perturbed-strategies page for the previously undocumented rtcfr module (rtcfr_solve, rtcfr_solve_with, perturb_uniform), described accurately as the CFR+ average with an annealed gamma-interior output perturbation rather than a last-iterate algorithm. Name the other publicly exported helpers that had no coverage: regret_matching in Training, and UniformPolicy, the Policy trait, and namespace_key/split_namespace in Profiles. Reorganize navigation around tasks: rename Game-trait to Implementing-a-game and Evaluation to Evaluating-agents, regroup Home's index into Guides and Reference, and add a _Sidebar with the same grouping.

    @sqlongithub sqlongithub committed Jul 18, 2026