{"schema_version":1,"total":1,"offset":0,"limit":100,"observations":[{"id":"public:340c76070a29b7a2454f34d0","benchmark_id":"epoch-gpqa-diamond::snapshot-2026-09-18","subject":{"source_id":"deepseek-r1-0528-qwen3-8b","name":"deepseek-r1-0528-qwen3-8b","model_id":"deepseek-r1-0528-qwen3-8b::default","variant":null,"harness":null},"value":0.0928030303030303,"unit":"fraction","basis":"measured","source":{"url":"https://epoch.ai/data/benchmark_data.zip","retrieved_at":"2026-09-18T23:49:00Z","published_at":null,"file":"data/raw/benchmarks/daily-evidence/2026-09-18-epoch-hub/epoch-gpqa_diamond.csv.gz","sha256":"8fc9a2c260248f1a95b5282d3af733894df6c39ebea9b50528231bfce1000748","locator":"csv; source row 35; deepseek-r1-0528-qwen3-8b; field Best score (across scorers)"},"protocol":"GPQA Diamond run by Epoch AI (inspect-ai), separate identity from the AA and OpenRouter runs; value = Best score (across scorers). mean_score, stderr, Started at and the Epoch run id stay in the protocol; CC BY 4.0, attribute Epoch AI.; mean_score=0.0928030303030303; stderr=0.010663485573693416; Started at=2026-08-28T07:48:47.000Z; id=dszRh7ELyCEukWnwYnUxMp","comparison_key":null,"join_note":"Reviewed identity map 2026-09-16: label names deepseek-r1-0528-qwen3-8b without an effort; the catalog has exactly one configuration, the default (deepseek-r1-0528-qwen3-8b::default)"}],"divergences":[],"cell":null,"coverage":{"model":null,"benchmark":{"total_models":844,"available":64,"unknown":780,"not_tested":0,"not_published":0,"source_unreachable":0,"contested":0,"measured":64,"self_reported":0,"observations":313,"unmatched_observations":249}},"collection":{"benchmark_id":"epoch-gpqa-diamond::snapshot-2026-09-18","status":"collected","source_url":"https://epoch.ai/data/benchmark_data.zip","reason":"313 source results parsed; configurations remain separate; unmatched model identities are retained."},"coverage_note":"Denominators are catalog model configurations × versioned registry entries; capability_available excludes Efficiency (cost) boards. Basis counts count covered cells, not runs; measured and self_reported may overlap. Unmatched source identities remain in observations and never inflate catalog coverage."}