frozen dataset: frozen-20260308 / generated_at: 2026-03-08T16:02:02Z

SALT Verification Console

현재 경로: /audit/reproduce

목적: SALT가 표준이론(거시: ΛCDM, 미시: SM) 대비 예측력에서 밀리지 않는지 검증

결론 Results

SALT vs 표준이론 승/무/패

근거 Evidence

실측 vs SM/ΛCDM vs SALT 비교

재현 Reproduce

동일 데이터/동일 식/동일 판정 재실행

원칙: 공개 판정은 Evaluation(frozen) 기준으로만 산출출처 공개예측식 공개

Audit / Reproduce

아래 순서대로 실행하면 동일 데이터셋 기준으로 동일 판정을 재현할 수 있습니다.

3-Step Reproducible Experiment

  1. 예측 엔진 실행 + 평가 산출물 생성 (`run_model_eval.py`)
  2. frozen snapshot 고정 및 manifest 생성
  3. prediction lock / frozen manifest 무결성 검증
.venv/bin/python tools/evaluation/run_model_eval.py
.venv/bin/python tools/evaluation/verify_prediction_lock.py
.venv/bin/python tools/evaluation/verify_frozen_manifest.py --manifest data/frozen/current/manifest.json

Cosmic Submission Mode

.venv/bin/python tools/evaluation/cosmic_feature_sidecar.py --write
# fill redshift_z/luminosity_distance_mpc in data/processed/cosmic_observation_features.json
# optional auto-enrich from source endpoints:
.venv/bin/python tools/evaluation/enrich_cosmic_features.py
# or edit results/reports/cosmic_feature_fill_queue.csv then:
.venv/bin/python tools/evaluation/apply_cosmic_feature_queue.py
.venv/bin/python tools/evaluation/cosmic_feature_sidecar.py --check
COSMIC_SUBMISSION_MODE=1 MICRO_SUBMISSION_MODE=1 .venv/bin/python tools/evaluation/run_model_eval.py

rerun commands:

  • .venv/bin/python tools/micro/run_micro_cycle.py
  • .venv/bin/python tools/realtime/run_realtime_cycle.py

model_eval commands:

  • /home/jwater/Development/salt-verification-console/.venv/bin/python3 /home/jwater/Development/salt-verification-console/tools/evaluation/run_cosmic_predictors.py
  • /home/jwater/Development/salt-verification-console/.venv/bin/python3 /home/jwater/Development/salt-verification-console/tools/evaluation/expand_cosmic_frozen_samples.py
  • /home/jwater/Development/salt-verification-console/.venv/bin/python3 /home/jwater/Development/salt-verification-console/tools/micro/run_micro_predictors.py
  • /home/jwater/Development/salt-verification-console/.venv/bin/python3 /home/jwater/Development/salt-verification-console/tools/micro/run_micro_stats.py

prediction lock hashes:

  • micro_prediction_lock_sha256: 327bafd8582b8de649b471c6751891691e3836160f19009bde644c901e3d89f2
  • micro_sm_prediction_sha256: a10f12fc8aa5ec74537c69a620512dde95ac8fed224a387353c1fa9acc37433d
  • micro_salt_prediction_sha256: a2adfe5466d451a85f8a7395ec581ae5f9f05773b64661ab127f9716cb3bd257
  • frozen_manifest_sha256: 1c36f819961fdea39a80147168de3a430a645ea9ac798e6f2221ae62d458896c

위 해시가 일치하면 같은 입력/같은 식/같은 엔진으로 동일 실험이 재현된 것입니다.