invariant-computing-lab/rari-neurobrain-v7

NeuroBrain V7: saved finite estimators

Actual, unchanged fitted coefficients from V7; not a newly trained language model. This is distinct from the previously supplied OMEGA bank.

Run python infer.py MODELS.json example_features.json --key pca/12/reconstruction after installing NumPy 2.x. Input shape is [n,1212] in the original mean/std/last [8,101,4] descriptor order. example_expected.json contains source-derived predictions for all 522 configurations on one seen development fixture. Reload consistency is a numerical software test, not scientific validation.

The bank contains 3 maps ×29 widths ×6 tasks. The chosen V7 group is PCA/12. “quantum_phase” names cos(XP)^2; it is not a quantum device. Classical phase is cos(XP+0.7)^2.

No license grant, new public repository, clinical claim or unconstrained raw-signal support is implied. Source-owner approval is needed for public distribution.

Tested numerical environment

CP05 release environment is Python 3.13.5 with NumPy 2.3.5. A separate NumPy 2.2.4 run passed 16/17 tests but failed the strict 1e-12 saved-fixture comparison at an observed 3.60e-12 difference. The tolerance was not relaxed and the failure is preserved; NumPy 2.2.4 is not certified by this release. The clean wheel environment shares the explicitly identified preinstalled NumPy 2.3.5 distribution, rather than claiming all dependencies were downloaded or rebuilt offline.

Repeatable reload check

Run python validate.py. This checks all 522 saved predictions at the original strict 1e-12 absolute tolerance, plus rejected-input contracts; it does not retrain or supply new holdout evidence.

Verified packaging revision — 2026-10-20

Preserved model weights reloaded; saved or synthetic example checks passed. No training, held-out evaluation, clinical or hosted verification.

Runtime state: PASS. Rights and release approval are separate. See RELEASE_VERIFICATION.json. No paper or conference submission is included as a new software result.