ColliderML: High-Luminosity Detector Simulation Data for Machine Learning [pdf] (ml4physicalsciences.github.io)

CERN has announced the release of ColliderML, an expansive open dataset aimed at enhancing machine learning (ML) research in high-luminosity particle physics. ColliderML comprises one million fully simulated proton-proton collision events generated under High-Luminosity Large Hadron Collider (HL-LHC) conditions, with realistic digital reconstructions and a validated OpenDataDetector (ODD) geometry. This dataset includes a diverse range of events across ten Standard Model and Beyond Standard Model processes, positioning itself as a significant resource that bridges the gap between the complex internal simulations used by major collaborations and the higher-level data typically accessible to public researchers. This initiative is pivotal for the AI/ML community as it facilitates the development of "end-to-end" foundation models by providing granular, low-level input data, which is crucial for training advanced ML algorithms. The release not only includes comprehensive data across various collision channels but also features a lightweight Python access library and standard reconstruction pipelines. By incorporating detailed event simulations and a focus on detector-level data, ColliderML enables researchers to explore sophisticated ML applications in particle physics, such as improved track reconstruction and event tagging, helping to spur advancements in both fundamental physics discoveries and ML methodologies within the field.