Huanxuan Liao

Dataset · Other

OPERATE

Huanxuan Liao

Benchmarking Persistent Operational Agency in Source-Grounded Executable Systems This is the current public dataset for OPERATE. Filenames are versionless; immutable Hugging Face commits identify each snapshot. The matching public code lives at Xnhyacinth/OPERATE. The current catalog contains 769 Full scenarios across 502 physical sources and 141 Lite scenarios across 82 sources. Qualification scoring is 0.15.0. Live evaluations use scoring 0.21.0 (waitrelativeoutcomev1). Live realtime uses nativedtv1; the default speed card is 37 Core rows whose native tick is at most 60s and whose plant wall is at most 30 minutes. logicalpersistent is the primary treatment. realtimepersistent is a…

Dataset · Other

OPERATE-Benchmark

Huanxuan Liao

The parent v0.61 admission ledger records 2,476 terminal candidate decisions with zero unresolved. v0.62 introduces zero newly mined candidates; it preserves that historical lineage while qualifying corrected contracts for the same 769-row, 502-source denominator. Historical admission evidence is not relabelled as newly executed evidence. Benchmarking Persistent Operational Agency in Source-Grounded Executable Systems This private dataset distributes the manifest-backed artifacts for the active operatev0620 release in Xnhyacinth/OPERATE-Benchmark, retaining private maintenance history. The public current distribution is Xnhyacinth/OPERATE. The promoted Core contains 769 scenarios grounded…