SkillOpt

// about

SkillOpt is Microsoft research code that optimizes a single skill document. A separate optimizer model reads scored rollouts and proposes bounded add, delete or replace edits. In the default path an edit is kept only when it strictly improves a held-out validation score, and a text learning-rate budget limits how much can change per step. It has a WebUI and lets you add new backends and benchmarks.

Clone the repository and follow the README to configure a backend and benchmark.