Evaluation matrix

Experiments

Seed, playground mock-eval, and frontend-eval runner experiments. Counts are row totals, not Index means.

Every experiment in the store, including frontend-eval. Origin mix is a count, not an aggregate score.
ExperimentStateOrigin mixRepetitionsRunsCreated
Phase 1 harness comparison

exp_preview_primary

Seed-only comparison across six synthetic tasks.

completed12 seed111/122026-08-05
Reliability retry check

exp_preview_retry

Seed-only timeout behavior check on baseline.

partially_completedNo runs10/02026-08-06
Playground mock eval channel

exp_playground_mock

Process-local playground runs (origin mock-eval). Excluded from Index publication aggregates.

running6 mock-eval16/62026-08-12
Front-end eval 0.1.0 reference

exp_frontend_eval

Runner experiment for frontend-eval 0.1.0. Starts empty. Reference adapter runs are origin runner.

draft14 runner17/142026-08-12