FAQ

Questions, answered

Common questions and a glossary so the jargon is never a barrier. For data, the leaderboard, and the starter kit, see Resources.

Questions

Frequently asked

Do I need a background in fusion or plasma physics?

No. The starter kit, the dFL visualizer, and four reference baselines are designed so newcomers can reach a first submission without fusion expertise. The data is fully documented and the synthetic-diagnostic kit converts raw signals into physics-friendly features for you.

Can I compete without a GPU?

Yes. Every reference baseline trains to within a few percent of leading numbers in under two hours on a single commodity GPU or CPU. The compute-light track recognizes accessible solutions, and GPU/cloud credits are available for resource-constrained teams.

What’s the difference between the compute-light and open tracks?

Compute-light submissions must train end-to-end in under two hours on one GPU/CPU; the open track removes all compute constraints. Both share the same data, submission format, and leaderboard — compute-light is a recognition, not a separate competition.

Why is reconstruction without magnetic sensors important?

Reactor-class devices (SPARC, ARC, CFETR) will subject magnetic sensors to neutron fluxes that degrade them severely. Equilibrium reconstruction stays essential, so an inference path from coil currents and electron profiles to ψ(R,Z) becomes a primary pipeline — and a backup for present machines.

What exactly do I submit?

For each test shot and EFIT timestamp: the 65×65 flux map, the LCFS contour, and five scalars (β_N, l_i, q₉₅, R_axis, Z_axis) — packaged as .npz or NetCDF4 with a manifest naming your harmonization layer. Submissions are capped at 5/day and 100 total.

Can I use external data or pre-trained models?

Yes — other public tokamak archives, OMFIT-produced equilibrium tables, and publicly available pre-trained vision or scientific foundation models are all permitted, provided you disclose them in your methods report.

How is cross-machine generalization judged?

Award #2 uses G_ratio = S_model(MAST) / S_model(DIII-D) — the fraction of DIII-D performance retained under zero-shot transfer to MAST — among entries that pass the harmonization gate and reach R²ψ > 0.6 on DIII-D.

When does the data become available?

Training and validation data are released publicly in June 2026, opening Phase 1. Six sample shots ship in the starter kit ahead of that for exploration.

Speak the language

Glossary

Tokamak: A toroidal magnetic-confinement device using combined toroidal and poloidal fields to confine a hot plasma.
Spherical tokamak: A tokamak with very low aspect ratio, where the plasma wraps closely around a thin central column (e.g. MAST).
EFIT: Equilibrium Fitting — the de-facto standard tokamak equilibrium-reconstruction code; provides the ground-truth flux maps.
Grad–Shafranov equation: The 2D nonlinear PDE governing axisymmetric MHD equilibria — the universal physics a model should ideally learn.
Poloidal flux ψ(R,Z): The 2D scalar field whose level sets are magnetic flux surfaces in the (R,Z) plane — the primary prediction target.
LCFS: Last Closed Flux Surface — the outermost flux surface not intercepted by a wall; the plasma boundary.
Thomson scattering: A laser diagnostic measuring electron temperature Te and density ne from scattered-light spectra.
PCA: Principal Component Analysis — compresses the 4,225-pixel flux map to ~20–50 coefficients (component 1 ≈ 92% of variance).
R²: Coefficient of determination, 1 − SS_res/SS_tot; bounded above by 1.
SSIM: Structural Similarity Index — a perceptual image-quality metric robust to small shifts.
Hausdorff distance: The greater of two directed set distances; here applied to predicted vs. true LCFS contours.
dsep: The X-point gap signal: > 0 indicates a diverted plasma, < 0 a limited plasma.