# Benchmark Reports
Benchmark artifacts generated by Phase 17 tooling are written here.
`docs/04-quality/reports/<timestamp>-<platform>-<lang>-<profile>/`
Each run directory contains:
- `manifest.json`
- `<lang>.json`
- `<lang>.stdout.log`
- `<lang>.stderr.log`
- `summary.md`
Use the runner wrappers:
- `scripts/bench/run-rust.sh`
- `scripts/bench/run-node.sh`
- `scripts/bench/run-python.sh`
Additional report assets:
- `2026-03-04-launch-pack/`: first publish-ready benchmark pack (pre-parity-hardening snapshot).
- `2026-03-04-launch-pack-parity/`: parity-hardened refresh with cross-language comparability validation.
- `baselines/`: accepted baseline JSON files for regression checks.
Historical naming note:
- Pre-Phase 18c benchmark artifacts may label `S-TRAV-002` as `neighbors_2hop`.
- That scenario now maps to `traverse_depth_2`, meaning `traverse(start, min_depth=2, max_depth=2, ...)`.
- Pre-CP30.5 archived run payloads may include old numeric-token field names in
`scenario_params`. They are immutable historical benchmark artifacts and are excluded
from current public API vocabulary audits; new runs should use label-ID metadata where
numeric labels are recorded.
Regression comparison outputs are typically stored alongside runs or under:
`docs/04-quality/reports/<pack-or-run>/compare*/`