Journey: Share a benchmark
← UX reviews · Persona: Benchmark publisher
Goal: a person with real test data (exam papers, checked translations) shares it as a benchmark, with permission and credit, so that tools are measured on real work.
Done when: the benchmark is listed with its maturity level, and tools are scored on it.
Steps today (2026-11-18)
| # | Where | What the person does |
|---|---|---|
| 1 | Homepage (“Grow a benchmark”, “Review a benchmark”) | Finds the way in. |
| 2 | Propose a benchmark | Describes the data in a GitHub issue form. |
| 3 | Maintainer and publisher | Agree on permission, license, held-out parts and format (how-to). Exam papers need a permission record. |
| 4 | Maintainer | Prepares the data in the standard format (id, input, target, metadata) from the benchmark template. |
| 5 | Leaderboard | The benchmark is listed, and models and skills are scored on it. |
Pages this journey touches
Homepage benchmark sections, the propose and expert-review forms, the template-benchmark README, buddhist-benchmarks CONTRIBUTING.md, the Add a benchmark tutorial.

BuddhistAI Knowledge base