We need attention for discovery

The Standard Model of particle physics accurately predicted the existence of the Higgs particle and has passed all experimental tests. However, it falls short in addressing fundamental physics questions such as the origin of matter in the Universe, the nature of dark matter, and neutrino masses. Thus, it is widely believed that undiscovered elementary particles exist.
The Large Hadron Collider (LHC) is among the most powerful instruments in physics, but the massive amount of LHC data produced thus far (100,000+ TB) requires innovative analysis strategies. Our proposal hypothesizes that recently developed Generative Pre-Trained Transformer models can be applied to the LHC data. This approach has shown success in natural language processing and could open a new era of unsupervised data analysis. The key innovation of transformers is their ”attention mechanism,” which effectively models long-range correlations in LHC data.
The GPT4LHC model will detect weak non-linear correlations, empowering searches with technology like ChatGPT. As a proof-of-concept, we will demonstrate its potential by rediscovering the Higgs boson in a novel way. Success in this project could pave the way for a new era of LHC discoveries.