# Provenance

This package is an anonymous reviewer artifact for SmartGridBench.

## Source Versions

- Package version: `neurips2026-review-v1`
- Package refreshed UTC: `2026-05-07T10:35:14Z`
- Package source snapshot SHA: `c9c2b6c6a2c85d97570f37a3d5c77d3d9ef136a8`
- Artifact import SHA: `c9c2b6c6a2c85d97570f37a3d5c77d3d9ef136a8`
- Evidence capture floor SHA: `1913c6e4703425f735d8cb8297cb890ba66bbeff`

The source snapshot SHA identifies the reviewed repository state used to prepare this package. The evidence capture floor SHA identifies the clean source state used by the main paper-grade experiment captures. The artifact import SHA identifies the final repository state used for the reviewer-facing evidence refresh.

## Included

- 36 canonical scenario JSON files: 31 hand-authored scenarios plus 5 generated scenarios promoted after validation and manual review.
- 5 negative schema fixtures.
- Paper-grade evidence registry and summary CSVs.
- Compact per-trial judge scores for paper-grade rows, with private URLs and absolute operator paths removed.
- Current failure-taxonomy surface and 46-row stratified manual-audit sidecars with reviewer identities and machine paths redacted.
- Current post-PR175 mitigation before/after export, with private run URLs and local generated-config paths redacted.
- Compact profiling/W&B inventory, with private run URLs, run IDs, and private artifact names redacted.
- 12-row manual judge audit sanity table.

## Excluded

- Raw upstream utility or telemetry datasets that are not redistributed in this artifact.
- VM-local logs, credentials, private run URLs, and full trajectory logs with environment-specific paths.
- Generated scenario candidates that were not promoted into the canonical 36-scenario package.
- Invalid or superseded cohorts except as explicit exclusion rows in `data/evidence_registry.csv`.

## Anonymization Notes

This package is intended for double-blind review. It avoids author names, institutional names, private run URLs, local usernames, and machine-specific absolute paths. Relative benchmark paths are preserved when useful for joining evidence tables.