NuBench: An Open Benchmark for Deep Learning-Based Event Reconstruction in Neutrino Telescopes

NuBench is a new open benchmark for deep learning-based event reconstruction in neutrino telescopes, comprising seven large-scale simulated datasets with nearly 130 million neutrino interactions. The benchmark enables comparison of machine learning reconstruction methods across different detector geometries and evaluates four algorithms including ParticleNeT and DynEdge on core reconstruction tasks.

• →NuBench provides the first comprehensive open benchmark for deep learning applications in neutrino telescope event reconstruction.
• →The benchmark includes nearly 130 million simulated neutrino interactions spanning six detector geometries inspired by existing and proposed experiments.
• →Four reconstruction algorithms were evaluated including ParticleNeT and DynEdge currently used by KM3NeT and IceCube collaborations.
• →The datasets support five core reconstruction tasks: energy and direction reconstruction, topology classification, interaction vertex prediction, and inelasticity estimation.
• →This standardized benchmark addresses the lack of diverse open-source datasets that previously hindered cross-experimental collaboration in neutrino physics.