Human vs. Agents in HEP Physics
Hacktron in Bonn is an initiative to bring together computer scientists and physicists for a unique competition centered around a real particle physics analysis, potentially using open data from the Large Hadron Collider (LHC). The idea is to explore, for the first time, how far autonomous AI-driven analysis workflows can compete with traditional human-led approaches in a realistic high-energy physics setting.
The competition consists of two parallel tracks. In the first track, participants develop AI-agent-based solutions in which the full physics analysis is carried out autonomously by intelligent agents. In the second track, teams perform the analysis in the conventional way, designing and executing the workflow themselves using established analysis methods and tools.
After an intensive 10-day coding and analysis period, the results from both approaches are compared in terms of physics performance, robustness, efficiency, reproducibility, and scientific insight. The ultimate goal is to produce a scientific publication covering both the physics measurement itself and a systematic evaluation of AI-agent performance relative to human-driven analyses.
The project aims not only to advance particle physics analysis techniques, but also to provide a realistic benchmark for autonomous scientific AI systems in one of the most demanding data-analysis environments in science.