{
  "schema_version": "public-table-export-v1",
  "dataset_id": "microsoft/benchpress-score-matrix",
  "generated_at_utc": "2026-08-18T17:25:50.638821+00:00",
  "source_json": "data/llm_benchmark_data.json",
  "parquet_written": true,
  "rows": {
    "models": 283,
    "benchmarks": 712,
    "scores_all": 8713,
    "scores_paper": 4905
  },
  "paper_matrix": {
    "models": 129,
    "benchmarks": 253,
    "observations": 4905,
    "fill_rate": 0.15028954867175293,
    "m_threshold": 15,
    "b_threshold": 8,
    "iterations": 5
  },
  "canonical_artifacts": {
    "data/llm_benchmark_data.json": {
      "sha256": "3de4b35676335836e199ea21a707e5a706b660472185eed71dcb4dd803ad1f28",
      "bytes": 9057527
    },
    "data/benchmark_cost_evidence.json": {
      "sha256": "42dd2f7bf666067ef31a72fb9df4f0670aea49df149bce78fbd0e75fed0cae16",
      "bytes": 112717
    }
  },
  "files": [
    "data/models.csv",
    "data/models.parquet",
    "data/benchmarks.csv",
    "data/benchmarks.parquet",
    "data/scores_all.csv",
    "data/scores_all.parquet",
    "data/scores_paper.csv",
    "data/scores_paper.parquet",
    "data/score_matrix_paper_wide.csv",
    "data/llm_benchmark_data.json",
    "data/benchmark_cost_evidence.json",
    "data/README.md",
    "data/SCHEMA.md",
    "data/benchmark_cost_evidence.README.md",
    "data/LICENSE-CDLA-2.0.md",
    "README.md",
    "metadata.json"
  ],
  "notes": [
    "llm_benchmark_data.json is the authoritative rich score-matrix artifact.",
    "benchmark_cost_evidence.json is the authoritative raw public cost-evidence artifact.",
    "scores_all is a flat public pre-filter table derived from the canonical JSON.",
    "scores_paper is the paper-canonical filtered long table.",
    "score_matrix_paper_wide.csv is the paper-canonical model-by-benchmark matrix.",
    "CSV and Parquet files omit nested fields such as candidates and audit provenance; use the canonical JSON for lossless data."
  ]
}