Monogate Command Center

EML-ADV-PCC4 Noisy Real-Source Holdoutprivate_reviewable

Private PCC4 noisy real-source holdout around eFrog rc_decay_stable.py. It separates model-to-model agreement from residual-to-noisy-observation and keeps prediction/runtime claims blocked.

passdeterministic_noisy_sample_gridbuilder preset

Reviewhuman gate

Keep private; use as the first noisy-data robustness guard for the EML Advantage contract.

EML-ADV-PCC5: ingest a second eFrog source and compare source-family behavior.

second real-source holdoutmissing

source-family comparisonmissing

Metricsbounded

holdout_count1

profile_count4

passing_profiles4

noisy_output_profiles1

broad_eml_advantage_claim0

runtime_performance_claim0

prediction_accuracy_claim0

Non-Claimsblocked

No broad noisy-data benchmark claim.non-claim

No broad EML advantage claim.non-claim

No prediction accuracy claim.non-claim

No runtime performance claim.non-claim

No compiler correctness or formal equivalence claim.non-claim

Claim Flagsfalse

{
  "public_ready": false,
  "safe_to_publish_publicly": false,
  "proof_claim": false,
  "runtime_performance_claim": false,
  "formal_equivalence_claim": false,
  "compiler_correctness_claim": false,
  "broad_eml_advantage_claim": false,
  "prediction_accuracy_claim": false
}

Sources And Checksmonogate

efrog/examples/rc_decay_stable.py
monogate/reports/evidence_packets/eml_advantage_pcc4_noisy_real_source_holdout.json
monogate/reports/eml_advantage_pcc4_noisy_real_source_holdout_2026_05_29.md
monogate/python/results/eml_advantage_pcc4_noisy_real_source_holdout/eml_advantage_pcc4_noisy_real_source_holdout_2026_05_29.json
monogate/python/results/eml_advantage_pcc4_noisy_real_source_holdout_packets/efrog_rc_decay_stable_noisy_real_source_holdout_v0_pcc4_noisy_real_source_holdout_2026_05_29.json

python python/scripts/eml_advantage_pcc4_noisy_real_source_holdout.py --build --strict
python -m pytest -q python/tests/test_eml_advantage_pcc4_noisy_real_source_holdout.py