pulselogicbio/cbae-compliance-algorithm: v1.1.0-gigascience — GigaScience submission

v1.1.0-gigascience — GigaScience Submission This release is the companion code for: Dowden, C.B. (2026). CBAE: a deterministic reproducibility evaluation algorithm and toolkit for biological machine learning pipelines. GigaScience [submitted]. What's new in v1.1.0 18-assertion compliance engine Full 18-assertion implementation across 5 evaluation dimensions New assertions: A5.2, A5.3, A5.4, V1, V2, V3 Criterion 4 assertions (A4.1–A4.3) now compute from real data Synthetic example dataset (no patient data required) example_data/synthetic_example.npz — Beta(3,1.5)/Beta(1.5,3), n=300, seed=42, AUC=0.881 python scripts/run_compliance.py --example → 18/18 PASS python scripts/demo_synthetic.py → standalone reviewer verification script Config corrections (v1.1.0) Phase 2: 16 tasks / 1,216 patients (corrected) Phase 3: 4 tasks / 1,542 patients (corrected) Deployment envelope declared in datasets.yaml (required for A5.3) Out-of-envelope NULL examples documented (required for A5.4) Verified output Assertions passed: 18/18 Verdict: PASS ✅ Previous releases v1.0.2 — SHA-256 Hashing Fix + UTC Compliance (PLOS CB submission) v1.0.1 — Initial public release