Calorimetry | Liangyu Wu - Hunting in the Invisible Worldhttps://liangyu5wu.github.io/tags/calorimetry/CalorimetryHugo Blox Builder (https://hugoblox.com)en-usThu, 23 Apr 2026 00:00:00 +0000https://liangyu5wu.github.io/media/icon_hu7729264130191091259.pngCalorimetryhttps://liangyu5wu.github.io/tags/calorimetry/Phenomenological Detector Design and Optimization in Vertically-Integrated Differentiable Full Simulations with Agentic-AIhttps://liangyu5wu.github.io/publication/det_opt_agentic/Thu, 23 Apr 2026 00:00:00 +0000https://liangyu5wu.github.io/publication/det_opt_agentic/<p>This work demonstrates the first application of AI agents to phenomenological detector design in high-energy physics, leveraging a novel bilevel optimization framework that simultaneously optimizes detector geometry and reconstruction parameters. Through agent-driven workflow execution using the SciFi framework with Claude Code Opus 4.6, we successfully reduced an 11-dimensional optimization problem to tractable subspaces, identifying optimal configurations for a dual-readout crystal ECAL. Future work will explore fully open-weight LLM backbones and extend the framework to broader detector subsystems and physics objectives, while investigating the integration of domain-specific knowledge into agentic reasoning capabilities.</p>