AI model trained on 120M physics events improves collision data classification

Researchers have developed a new foundation model for event classification in high-energy physics, utilizing a Graph Neural Network architecture. This model was pretrained on 120 million simulated proton-proton collision events across 12 physics processes to learn general representations of collision data. Fine-tuning the model demonstrated significant improvements in accuracy and computational efficiency for various classification tasks, including those involving new physics processes and real-world ATLAS Open Data, showcasing its generalizability. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT This foundation model could accelerate research in high-energy physics by improving the efficiency and accuracy of event classification.

RANK_REASON This is a research paper detailing a new model for physics analysis. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Graph Neural Network
• ATLAS Open Data
• Delphes

• paper
• other