sema-evalsIndependent evaluations

Sema discovery and session reuse — 20260723T073223827Z-order-20260716

Deterministic mechanism/scaffold validation of search, selection, dependency resolution, execution, and within-session reuse. Scripted outcomes are constructed and are not evidence that models discover useful patterns or that a library improves workflow performance.

Deterministic mechanism validation only. Search, selection, dependency resolution, execution, and reuse are scripted; these outcomes are not evidence that a model discovers useful patterns.

ConditionTrials End-to-endSearches Correct selectionsDependency complete Executions passedReuse hits Wire BHydration B
task-only 1 0.0% 0.0 0.0 0.0% 0.0 0.0 176.0 0.0
preselected-prose 1 0.0% 0.0 0.0 100.0% 2.0 0.0 1794.0 0.0
preselected-addressed 1 0.0% 0.0 0.0 100.0% 2.0 0.0 816.0 1584.0
discovery 1 100.0% 2.0 2.0 100.0% 2.0 0.0 1252.0 1584.0
discovery-reuse 1 100.0% 1.0 1.0 100.0% 2.0 1.0 740.0 792.0

Every displayed aggregate is recomputed from trials.public.jsonl. Wire and hydration bytes remain separate.