Reproduce “SensOFormer: Foundation Models for Particle Accelerators” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mahindra Rautela, Alexander Scheinker
SensOFormer learns a shared representation from accelerator telemetry by masking and reconstructing readings across variable sensor sets with a set encoder and temporal transformer. For zero-shot imputation on unseen GMPS-Booster-Fermilab data at 15% masking, its B model achieves raw MAE 148.0681, versus 154.4622 for XTi, with no facility-specific fine-tuning. Supervised diagnostic tasks use task-specific heads; the study is offline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.