PN — A benchmark run is fully spec-driven: a RunSpec YAML validated with…

A benchmark run is fully spec-driven: a RunSpec YAML validated with Pydantic v2 plus a fixed seed makes every run reproducible; run_id is {spec.id}{spec_hash}{timestamp} where spec_hash is SHA-256[:16] of the RunSpec excluding its description.

Source: PUMA project documentation · Traceability: corpus unit PUMA-008 · Confidence: repo-backed