Context
Issue #100 identified current Distribution evidence that is not yet suitable for the canonical benchmark dashboard. The repository contains June 2026 evidence, but coverage and provenance must be audited before registration.
Goal
Produce a current structured Distribution benchmark source that can be registered without reconstructing or guessing missing measurements.
Scope
- audit existing
results/distribution_bench_2026-06-21.json and its Markdown summary;
- define the Distribution comparison identity, cases, operations, scales, and timing protocol;
- preserve measured versus reported semantics explicitly;
- add or adapt a parser only when the raw source supports the required dimensions;
- register the source in
frontend_sources.json only after provenance, timing, and statistical-alignment gates pass;
- update the coverage matrix and dashboard documentation.
Validation
- deterministic structured source;
- explicit CPU/CuPy/Torch synchronization and timing scope where applicable;
- manifest SHA256 and source date;
- parser/schema/semantic tests;
- generated-data and deployment-asset staleness;
- no values reconstructed from rounded prose.
Non-goals
- no fabricated rows;
- no registration solely to increase source counts.
Parent: #100
Context
Issue #100 identified current Distribution evidence that is not yet suitable for the canonical benchmark dashboard. The repository contains June 2026 evidence, but coverage and provenance must be audited before registration.
Goal
Produce a current structured Distribution benchmark source that can be registered without reconstructing or guessing missing measurements.
Scope
results/distribution_bench_2026-06-21.jsonand its Markdown summary;frontend_sources.jsononly after provenance, timing, and statistical-alignment gates pass;Validation
Non-goals
Parent: #100