build_release_notes.py

from pathlib import Path
import json
root=Path(__file__).resolve().parent
r=[json.loads((root/'results/results.json').read_text())]+[json.loads(p.read_text()) for p in sorted((root/'results/replications').glob('*/results.json'))]
s={'seeds':[x['seed'] for x in r], 'runs':len(r),
   'learned_pipeline_test_words':sum(x['learned_esre_pipeline']['test_words'] for x in r),
   'learned_pipeline_correct':sum(x['learned_esre_pipeline']['weave_correct'] for x in r),
   'pooled_baseline_correct':sum(x['learned_esre_pipeline']['conventional_pooled_group_correct'] for x in r),
   'erasure_trials':sum(sum(y['trials'] for y in x['erasure_recovery']['rows']) for x in r),
   'false_erasure_certificates':sum(sum(y['false_certificates'] for y in x['erasure_recovery']['rows']) for x in r),
   'exhaustive_fiber_tests':sum(x['exhaustive_identifiability']['trials'] for x in r),
   'incorrect_fiber_certificates':sum(x['exhaustive_identifiability']['incorrect_certificates'] for x in r),
   'frozen_closure_words':sum(sum(y['test_words'] for y in x['closure_and_frozen']['rows']) for x in r),
   'fractal_tree_words':sum(x['recursive_fracture']['test_words'] for x in r),
   'curriculum_words':sum(sum(y['tested'] for y in x['recursive_curriculum']['rows']) for x in r),
   'neural_experiments':0,
   'warning':'Exact-family implementation checks. No claims of statistical general-AI superiority.'}
(root/'results/replication_summary.json').write_text(json.dumps(s,indent=2))
print(json.dumps(s,indent=2))