Speaker
Description
The High-Luminosity Large Hadron Collider (HL-LHC) faces a two-pronged challenge: first, data management and analysis has become more complex and needs to be prioritized. For this purpose, the ever-evolving nature of AI warrants a shift from conventional apprenticeship-based training methods to more scalable workforce development infrastructure. At the same time, quantum-based methods, which are rapidly advancing, have increasingly started to intersect with classical machine learning workflows. This requires researchers to acquire a wider range of skillset.
To address this problem, we demonstrate an AI-native pedagogical engine. Leveraging large language models (LLMs), students use hands-on learning based on CERN Open Data, such as the ATLAS (2012, 8 TeV) and CMS datasets (CERN Open Data Portal, 2024). Furthermore, learners interact with a system that integrates a RAG framework grounded in the CERN Document Server (Lewis et al., 2020). The unique feature of the server is that it has a sandboxed execution environment (SWAN/ROOT). Not only this setup allows iterative code generation, execution, and validation on real collision data, but it preserves physics consistency.
The range of learning tasks varies from understanding detector-level data to performing classification, which includes applications to Higgs boson decay channels (H → γγ). Modules based on advanced topics underscore the applications of graph neural networks (GNNs). Lastly, learners explore hybrid approaches and leverage variational quantum classifiers (VQC) and quantum support vector machines (QSVM) on LHC datasets (Biamonte et al., 2017).
Our pilot, conducted through a hackathon with early researchers, indicates that there was an improvement in the understanding of concepts. Furthermore, the learners showed a better grasp of code correction and achieved better efficiency in completing their task using an AI-based curriculum and learning system. Based on our findings, this work underscores the effectiveness of AI-based training as an engine to enable scalable HL-LHC readiness and quantum-savvy workforce development.
| Track | Outreach, Diversity and Education |
|---|