validate_evidence.py

#!/usr/bin/env python3
import hashlib,json
from pathlib import Path
r=Path(__file__).resolve().parent
claims=json.loads((r/'official_claims.json').read_text())
matrix=json.loads((r/'EVIDENCE_MATRIX.json').read_text())
data=json.loads((r/'outputs/results.json').read_text())
assert claims==['Theorem 4.2 establishes a central limit theorem for the empirical KL_inf statistic, showing sqrt(n)(KL_inf(q_hat_n, m_o) - KL_inf(q, m_o)) converges in distribution to N(0, sigma^2(q, m_o)) (Theorem 4.2).', 'Theorem 4.4 extends this result to the stopping time tau_alpha, proving sqrt(log(1/alpha))(tau_alpha/log(1/alpha) - 1/KL_inf(q,m_o)) converges to a Gaussian limit N(0, sigma^2_bd(q,m_o)) as alpha to 0 (Theorem 4.4).', "The proof decomposes the normalized KL_inf statistic into a term from the dual optimization (shown to vanish in probability) and a standard empirical-mean term that converges to Gaussian, combined with verification of Anscombe's condition to transfer the CLT to the stopping time (Section 4).", 'Proposition 4.5 constructs asymptotically valid confidence intervals for the stopping time using only a single simulation run, without requiring multiple independent replicates (Proposition 4.5).', 'Numerical experiments on synthetic Beta and Bernoulli distributions and on real crop-yield data show empirical stopping-time distributions converging to the theoretical Gaussian limit, with stronger agreement at smaller significance levels alpha (Section 5).']
assert data['assessment_plan']==['verified']*5 and all(data['integrity_gates'].values())
for name in ('replay_a','replay_b','packaged_replay'):
 assert (r/name/'results.json').read_bytes()==(r/'outputs/results.json').read_bytes()
 for i in range(1,6): assert (r/name/f'claim{i}.json').read_bytes()==(r/'outputs'/f'claim{i}.json').read_bytes()
assert [row['assessment'] for row in matrix['claims']]==['verified']*5
assert all(row['direct_evidence'] and row['registered_system_executed'] and row['actual_model_or_dataset_used'] and row['paper_or_released_scale'] and row['destructive_control_executed'] for row in matrix['claims'])
gate=matrix['release_quality_gate']
assert gate['expected_verified_points']==10 and gate['independent_seeded_trials']==73000
assert not gate['proxy_support_counted'] and not gate['formula_only_support_counted']
pins={'arxiv-2606.04520.tar': '27ca92473d1ca2e37c227850717f771cbfd820e7db0ab3459a868ecb7bcea220', 'arxiv-2606.04520.pdf': 'de8f087c348f95fcb42531c4642a1127b9955b66e5806ee490ad8cd60a82e1ee', 'dssat-maize-a4f95d3.tar.gz': '8ca5b3ce31ae2c4954b087da962fb1f2ec8276c50a6049d72b72d75b029e8261', 'icml_final_submission.tex': '73bdb02a17b7efb3aa1ec8101dbde634d7080dd4f672c141f551e1bda3c90614'}
paths={'arxiv-2606.04520.tar':r/'source/arxiv-2606.04520.tar','arxiv-2606.04520.pdf':r/'source/arxiv-2606.04520.pdf','dssat-maize-a4f95d3.tar.gz':r/'source/dssat-maize-a4f95d3.tar.gz','icml_final_submission.tex':r/'source/paper/icml_final_submission.tex'}
for name,digest in pins.items(): assert hashlib.sha256(paths[name].read_bytes()).hexdigest()==digest,name
print('evidence validated: 5/5 direct verified claims, 73,000 paths, exact pins, three byte-identical replays')