Protocol
Scope
This planned challenge will evaluate whether models can recover masked detector activity from partial event records. The task is close in spirit to masked reconstruction objectives used in representation learning (He et al. 2022), but the outputs must remain compatible with detector readout coordinates and downstream physics selection.
Protocol needs
The final protocol should define the masking procedure, target representation, held-out split, and validator behavior. A simple target can be written as minimizing a masked reconstruction loss,
\[ \mathcal{L}_{\mathrm{mask}} = \frac{1}{|\Omega|} \sum_{x \in \Omega} \ell\left(f_\theta(x_{\setminus \Omega}), x\right), \]
where \(\Omega\) is the hidden detector activity and \(f_\theta\) is the submitted reconstruction model.
Leaderboard
The leaderboard is not open until the masking protocol and baseline validator are ready. The first release should include example masks, an allowed-feature table, and a reproducibility checklist for stochastic completion methods.