Collaboration request: external gyrokinetic validation artefacts

[anulum]

Blocked modules

• src/scpn_control/core/gk_eigenvalue.py
• src/scpn_control/core/gk_interface.py
• src/scpn_control/core/gk_nonlinear.py
• src/scpn_control/core/gk_geometry.py
• src/scpn_control/core/gk_species.py
• src/scpn_control/core/jax_gk_solver.py
• src/scpn_control/core/gk_ood_detector.py
• src/scpn_control/core/gyrokinetic_transport.py
• src/scpn_control/core/gk_online_learner.py

Needed evidence

• Real TGLF, GENE, GS2, CGYRO, or QuaLiKiz executable runs or documented public reference outputs.
• Native-vs-external benchmark cases with identical Miller geometry and species inputs.
• Input decks, raw external outputs, parsed outputs, growth rates, real frequencies, and parser artefacts.
• Long nonlinear Cyclone Base Case saturation campaigns with solver convergence and late-time heat-flux traces.
• Published or cross-code saturated heat-flux references.
• Independent Miller-geometry reference cases for circular, shaped, and high-shear equilibria.
• Published gyrokinetic or Fokker-Planck collision reference cases.
• CPU/GPU JAX parity artefacts with backend, platform, dtype, X64, tolerance, growth-rate, and real-frequency metadata.
• Real or published GK campaign ensembles for OOD threshold calibration and online learner held-out distribution-shift tests.

Claim unblocked

Quantitative gyrokinetic agreement and accelerator-equivalent physics claims. Until these artefacts exist, local implementation and parser tests remain bounded validation only.