Speaker
Description
The FCC-ee aims to measure flavor, Higgs and top Physics with unprecedented precision, making detector optimization a critical component of the FCC-ee program. Traditional detector studies typically optimize individual detector subsystems or low-level performance metrics, such as tracking resolution or flavor-tagging efficiencies, rather than the ultimate physics reach. We present an agentic framework for end-to-end detector optimization that directly minimizes the expected uncertainty on Higgs couplings.
Our workflow combines fast detector simulation, transformer-based flavor tagging, and full physics analysis within an automated optimization loop. Candidate detector configurations, including geometry, timing resolution, and detector granularity, are evaluated using Delphes-based simulation, transformer taggers trained on particle-flow objects, and a Higgs coupling analysis. The resulting coupling precisions define a physics-informed loss function that enables the agent to drive the optimization while satisfying realistic detector and engineering constraints. This framework enables systematic exploration of detector design choices using physics performance as the optimization objective and provides a scalable strategy for AI-assisted detector design studies at the FCC-ee.
