AI and the World's Hard Problems · Physics · arXiv · 2026-10-23
Track finding and particle identification performance with particle physics foundation models
High-energy physics experiments produce large volumes of sparse data that must be reduced before analysis. The authors evaluated a foundation-model-based reconstruction on several downstream tasks, using spacepoints from a Time Projection Chamber to find charged-particle tracks. The model significantly outperformed traditional methods and could be adapted to tasks not related to its original training objective.