Neutrino telescopes are large-scale detectors designed to observe Cherenkov radiation produced from neutrino interactions in water or ice. They exist to identify extraterrestrial neutrino sources and to probe fundamental questions pertaining to the elusive neutrino itself. A central challenge common across neutrino telescopes is to solve a series of inverse problems known as event reconstruction, which seeks to resolve properties of the incident neutrino, based on the detected Cherenkov light. In recent times, significant efforts have been made in adapting advances from deep learning research to event reconstruction, as such techniques provide several benefits over traditional methods. While a large degree of similarity in reconstruction needs and low-level data exists, cross-experimental collaboration has been hindered by a lack of diverse open-source datasets for comparing methods. We present NuBench, an open benchmark for deep learning-based event reconstruction in neutrino telescopes. NuBench comprises seven large-scale simulated datasets containing nearly 130 million charged- and neutral-current muon-neutrino interactions spanning 10 GeV to 100 TeV, generated across six detector geometries inspired by existing and proposed experiments. These datasets provide pulse- and event-level information suitable for developing and comparing machine-learning reconstruction methods in both water and ice environments. Using NuBench, we evaluate four reconstruction algorithms - ParticleNeT and DynEdge, both actively used within the KM3NeT and IceCube collaborations, respectively, along with GRIT and DeepIce - on up to five core tasks: energy and direction reconstruction, topology classification, interaction vertex prediction, and inelasticity estimation.

Significant experimental efforts have been devoted to studying (anti)neutrino-nucleus interactions [1, 2] in the energy range relevant for next-generation neutrino oscillation experiments, such as Hyper-Kamiokande [3] and DUNE [4]. In parallel, theoretical models describing these interactions have been developed [2]. The outcomes of both experimental and theoretical advances are incorporated into Monte Carlo (MC) event generators, which simulate (anti)neutrino-nucleus collisions under realistic conditions [5-10]. MC generators are often tuned to reproduce experimental observations, relying on adjustable parameters that are fitted using available data [11]. However, this tuning process cannot fully compensate for the fundamental limitations of the underlying models, especially those relying on complex approximations, such as nuclear modeling. Consequently, there is a growing interest in alternative approaches to traditional MC event generation--methods that can learn directly from experimental data and dynamically refine their predictions.

The first type of 13 interaction plays a pivotal role in the oscillation analyses carried out by the T2K and Hyper-Kamiokande experiments, and the other is important for the DUNE experiment. We consider various kinematic distributions of the charged lepton. The models we present successfully reproduce the peak structure in data distributions. Once these models are developed, they generate events significantly faster than "classical" generators. We also anticipate that these models can be adapted to more realistic scenarios after retraining them on experimental data. Essentially, they can serve as pre-trained models that can be fine-tuned for specific applications. Our study opens the door for future developments, including considering complete event topologies and realistic neutrino fluxes. Furthermore, these deep neural network models can be repurposed to simulate related processes by utilizing advanced deep learning techniques such as transfer learning [15].

Convolutional neural networks (CNNs) have seen extensive applications in scientific data analysis, including in neutrino telescopes. However, the data from these experiments present numerous challenges to CNNs, such as non-regular geometry, sparsity, and high dimensionality. Consequently, CNNs are highly inefficient on neutrino telescope data, and require significant pre-processing that results in information loss. We propose sparse submanifold convolutions (SSCNNs) as a solution to these issues and show that the SSCNN event reconstruction performance is comparable to or better than traditional and machine learning algorithms. Additionally, our SSCNN runs approximately 16 times faster than a traditional CNN on a GPU. As a result of this speedup, it is expected to be capable of handling the trigger-level event rate of IceCube-scale neutrino telescopes. These networks could be used to improve the first estimation of the neutrino energy and direction to seed more advanced reconstructions, or to provide this information to an alert-sending system to quickly follow-up interesting events.