Monogate Command Center

eFrog Holdout Registryprivate_reviewable

Four unseen holdout kernels run through source -> EML -> Forge targets -> runtime guard witness -> evidence ladder reviewer without joining the default benchmark corpus.

passnot_applicablebuilder preset

Reviewhuman gate

Keep private; use as the first repeatable unseen-kernel evidence lane.

Add tags and connect registry summaries into the EML Advantage Lab as private inputs.

holdout tagsmissing

EML Advantage Lab linkagemissing

Metricsbounded

holdout_count4

pass_count4

runtime_witnessed_cases4

runtime_witnessed_obligations6

proof_discharged_obligations0

Non-Claimsblocked

No public benchmark claim.non-claim

No proof discharge claim.non-claim

No runtime performance claim.non-claim

No full compiler correctness claim.non-claim

Claim Flagsfalse

{
  "public_ready": false,
  "proof_claim": false,
  "certified_safety_claim": false,
  "runtime_performance_claim": false,
  "formal_equivalence_claim": false,
  "broad_eml_advantage_claim": false
}

Sources And Checksefrog

efrog/examples/gaussian_stable.py
efrog/examples/rc_decay_stable.py
efrog/examples/stretched_exponential.py
efrog/examples/stable_sigmoid.py
efrog/schemas/efrog_holdout_registry_v0.schema.json
efrog/efrog/holdout_trial.py

python -m efrog --holdout-registry --out-dir /tmp/efrog_r13_holdout_registry_verify
python -m pytest -q tests/test_bridge_artifacts.py
python -m pytest -q