Datasetpublic

microsoft/benchpress-score-matrix

BenchPress Score Matrix This dataset contains the public model-by-benchmark score matrix used by BenchPress. The release includes the lossless audited JSON, benchmark cost evidence, flat model and benchmark metadata, one row per observed score, and the paper-canonical dense subset used in the BenchPress experiments. The source repository is microsoft/benchpress. Canonical artifacts data/llm_benchmark_data.json is the authoritative rich score-matrix artifact. It… See the full description on the dataset page: https://huggingface.co/datasets/microsoft/benchpress-score-matrix.

sourceHugging Facecdla-permissive-2.0updated 1mo agoView on Hugging Face

2likes247downloads

rootmetadata.json

metadata.json60 linesDownload Raw Back to root

1{2  "schema_version": "public-table-export-v1",3  "dataset_id": "microsoft/benchpress-score-matrix",4  "generated_at_utc": "2026-08-18T17:25:50.638821+00:00",5  "source_json": "data/llm_benchmark_data.json",6  "parquet_written": true,7  "rows": {8    "models": 283,9    "benchmarks": 712,10    "scores_all": 8713,11    "scores_paper": 490512  },13  "paper_matrix": {14    "models": 129,15    "benchmarks": 253,16    "observations": 4905,17    "fill_rate": 0.15028954867175293,18    "m_threshold": 15,19    "b_threshold": 8,20    "iterations": 521  },22  "canonical_artifacts": {23    "data/llm_benchmark_data.json": {24      "sha256": "3de4b35676335836e199ea21a707e5a706b660472185eed71dcb4dd803ad1f28",25      "bytes": 905752726    },27    "data/benchmark_cost_evidence.json": {28      "sha256": "42dd2f7bf666067ef31a72fb9df4f0670aea49df149bce78fbd0e75fed0cae16",29      "bytes": 11271730    }31  },32  "files": [33    "data/models.csv",34    "data/models.parquet",35    "data/benchmarks.csv",36    "data/benchmarks.parquet",37    "data/scores_all.csv",38    "data/scores_all.parquet",39    "data/scores_paper.csv",40    "data/scores_paper.parquet",41    "data/score_matrix_paper_wide.csv",42    "data/llm_benchmark_data.json",43    "data/benchmark_cost_evidence.json",44    "data/README.md",45    "data/SCHEMA.md",46    "data/benchmark_cost_evidence.README.md",47    "data/LICENSE-CDLA-2.0.md",48    "README.md",49    "metadata.json"50  ],51  "notes": [52    "llm_benchmark_data.json is the authoritative rich score-matrix artifact.",53    "benchmark_cost_evidence.json is the authoritative raw public cost-evidence artifact.",54    "scores_all is a flat public pre-filter table derived from the canonical JSON.",55    "scores_paper is the paper-canonical filtered long table.",56    "score_matrix_paper_wide.csv is the paper-canonical model-by-benchmark matrix.",57    "CSV and Parquet files omit nested fields such as candidates and audit provenance; use the canonical JSON for lossless data."58  ]59}60