{
  "schema_version": 1,
  "seed_label": "2026-08-11-r1-mini-sanity-v1",
  "source_revision_audited": "4aa9c07de93210c36f6e7118009e58ec8d4c7b6c",
  "uniform_sample_size": 32,
  "semantic_review": {"pass": 32, "fail": 0, "uncertain": 0},
  "full_artifact_scan": {
    "rows": 6236,
    "duplicate_ids": 0,
    "duplicate_content": 0,
    "strict_answer_under_15_words": 0,
    "explicit_visual_artifacts": 0,
    "step_contract_failures": 0
  },
  "rows_removed_after_review": 0,
  "post_review_rows": 6236,
  "post_review_jsonl_sha256": "3ba6fd47cba972f4d304afb0805ee5da6d4d7909eed0a5e2574a9d0ae126cd5e",
  "method": "Uniform random row sampling without replacement followed by structured review of question self-containment, reasoning relevance and coherence, answer completeness, and identifiable mathematical errors.",
  "limitation": "A 32-row semantic review is a sanity check, not a proof that every retained response is correct."
}